# Jogg Pricing — Plans & Features
**MokingBird Oy**
**Status:** Current — Reflects 3-Framework launch pricing
**Last Updated:** May 2026

---

## Overview

Jogg's monetization model: **freemium → monthly subscription or lifetime purchase**.

All billing goes through the platform stores — Apple App Store (iOS) or Google Play (Android). Jogg does not handle payment card data directly.

---

## Pricing Tiers

### Free

**Price:** €0 — no credit card required

**Question access:**
- Layer 0 & 1: up to **50 free questions** per layer, per framework
- Layer 2: up to **20 free questions** per framework
- Layer 3–8: up to **10 free questions** per layer, per framework
- AI Development Layer 3 exception: up to **50 free questions**
- After limit is reached: a purchase prompt appears offering Monthly or Lifetime

**Features included:**

| Feature | Free |
|---------|------|
| Daily Jogg (spaced repetition) | Unlimited |
| Custom quiz — create | 3 per calendar month |
| Custom quiz — join | 3 per calendar month |
| Arcade | Up to Run 20 (200 questions) |
| Certificates download | Up to 3 total |
| Leaderboard | Yes (opt-in) |
| All 3 frameworks | Yes |
| All 9 layers per framework | Yes (free question quota applies) |
| Buy more questions | Yes (per layer, per framework) |

---

### Monthly Premium

**Price:** €4.99/month — cancel anytime

**Question access:**
- **Unlimited** — full question bank for all layers, all frameworks, all difficulties

**Features included:**

| Feature | Monthly |
|---------|---------|
| Daily Jogg | Unlimited |
| Custom quiz — create | Unlimited |
| Custom quiz — join | Unlimited |
| Arcade | Unlimited levels |
| Certificates download | Unlimited |
| Leaderboard | Yes (opt-in) |
| All 3 frameworks | Yes |
| All 9 layers per framework | Yes — no limits |

**Billing:** Auto-renewing via App Store or Google Play. Cancel anytime; access continues until end of billing period.

**Annual plan:** Planned — 2 months free (~€49.90/year).

---

### Lifetime Access

**Price:** €29.99 — one-time purchase

**Question access:**
- **100 questions per layer per framework** included
- Additional questions can be purchased per layer (consumable in-app purchase)
- All future updates to core features included at no extra charge

**Features included:**

| Feature | Lifetime |
|---------|---------|
| Daily Jogg | Unlimited |
| Custom quiz — create | Unlimited |
| Custom quiz — join | Unlimited |
| Arcade | Unlimited levels |
| Certificates download | Unlimited |
| Leaderboard | Yes (opt-in) |
| All 3 frameworks | Yes |
| All 9 layers per framework | Yes — 100 questions included |
| Buy more questions | Yes (per layer, per framework) |

The lifetime purchase covers all current and future core Jogg features. Major new product lines are not included.

---

## Full Comparison

| Feature | Free | Monthly (€4.99/mo) | Lifetime (€29.99) |
|---------|------|---------------------|-------------------|
| Questions per layer | Limited free quota | Unlimited | 100 per layer (buy more available) |
| Daily Jogg | Unlimited | Unlimited | Unlimited |
| Custom quiz (create) | 3/month | Unlimited | Unlimited |
| Arcade | Up to Run 20 | Unlimited | Unlimited |
| Certificates | Up to 3 | Unlimited | Unlimited |
| Leaderboard | Yes | Yes | Yes |
| All 3 frameworks | Yes | Yes | Yes |

---

## Competitive Reference

| Product | Price |
|---------|-------|
| Duolingo Plus | ~€6.99/mo |
| LeetCode Premium | ~€12/mo |
| DataCamp Individual | ~€25/mo |
| **Jogg Monthly** | **€4.99/mo** |
| **Jogg Lifetime** | **€29.99 one-time** |

---

*For pricing questions or enterprise licensing: [email protected]*

*MokingBird Oy — Jogg pricing, May 2026*

# About Jogg — Your Professional AI/ML Learning App
**by the Jogg Team | MokingBird Oy**

---

## What Is Jogg?

**Jogg** is a professional, gamified learning platform designed for people who are serious about Artificial Intelligence and Machine Learning. Whether you are a university student, a software engineer preparing for ML interviews, a researcher looking to refresh your knowledge, a power user mastering AI tools, or a professional who needs to understand AI without building it — Jogg is built for you.

Jogg is available on **iOS and Android**, built to be fast, beautiful, and deeply functional.

> *"Learn AI/ML the right way — comprehensive, adaptive, and engaging."*

---

## The Problem Jogg Solves

Learning AI/ML is genuinely hard. The field is vast, fast-moving, and the available resources are scattered across textbooks, YouTube videos, blog posts, and research papers that assume widely varying levels of prior knowledge.

Most learning apps are either:
- **Too shallow** — surface-level quizzes that don't build real understanding
- **Too fragmented** — covering isolated topics without a coherent progression path
- **Designed for children** — gamification that feels patronizing to professional learners
- **Paper-ignorant** — completely disconnected from the research literature that actually defines the field
- **One-size-fits-all** — no separation between engineers, power users, and general learners

Jogg was created to fix all of this.

---

## Who Jogg Is For

Jogg serves learners across the full AI ecosystem — from builders to users to curious everyday people:

### Students (AI Development path)
You're studying computer science, data science, or a related discipline. You want a structured way to learn the full AI/ML stack — not just isolated libraries, but a deep understanding of why things work the way they do. Jogg's AI Development framework gives you a clear path from mathematical foundations all the way to AI safety.

### Job Seekers & Interview Preppers (AI Development path)
You're applying for ML engineer, data scientist, or AI researcher positions. Jogg covers everything from linear algebra and optimization to RAG systems, transformer internals, and production deployment — exactly what modern technical interviews test.

### Working Professionals (AI Development or AI Usage path)
You're already in the industry but AI moves fast. Daily Jogg gives you a structured way to stay sharp in just 5–10 minutes a day, using scientifically-backed spaced repetition to reinforce what you know and surface what you don't.

### Researchers (AI Development path)
You understand the field deeply but want to test your understanding across the full spectrum. Jogg's Papers Quest mode lets you quiz yourself directly on the research literature, including landmark papers like "Attention Is All You Need", BERT, GPT-3, LoRA, and more.

### Power Users & AI Tool Professionals (AI Usage path)
You use AI tools every day — ChatGPT, Claude, Midjourney, Copilot, n8n, and more. You want to go from competent user to genuine master: advanced prompting, workflow automation, AI agent orchestration, evaluating outputs, and staying ahead of the curve.

### Non-Technical Professionals (AI Literacy path)
You're a teacher, healthcare worker, manager, journalist, or professional in any field where AI is becoming relevant. You don't need to build AI — but you do need to understand it: what it can and cannot do, how to use it responsibly, and how it affects your field.

### AI Enthusiasts (any path)
You're passionate about AI and want to build a genuinely rigorous foundation, not just familiarity with buzz words. Jogg meets you at your level and adapts as you grow.

---

## The 3 Learning Frameworks

Jogg is organized into **three complete learning frameworks**, each with 9 layers — 27 layers and 3,600+ questions in total. All layers within each framework are accessible from the start — no mastery gates, no prerequisites required.

### AI Development — For Engineers & Researchers

| Layer | Topic | What You Learn |
|-------|-------|---------------|
| 0 | **Foundations** | ML basics, Python, linear algebra, statistics |
| 1 | **Data Acquisition** | APIs, web scraping, SQL, data streaming |
| 2 | **Data Preprocessing** | Cleaning, feature engineering, normalization |
| 3 | **Model Training** | Neural networks, transformers, optimizers, fine-tuning |
| 4 | **Orchestration & RAG** | RAG systems, vector databases, agent orchestration |
| 5 | **Inference & Optimization** | Quantization, pruning, distillation, KV cache |
| 6 | **Integration & Applications** | API design, deployment, monitoring, CI/CD for ML |
| 7 | **Future Topics** | Multimodal AI, diffusion models, MoE, reinforcement learning |
| 8 | **AI Safety & Governance** | Bias, safety, ethics, responsible AI deployment |

### AI Usage — For Power Users & AI Tool Professionals

| Layer | Topic | What You Learn |
|-------|-------|---------------|
| 0 | **First Contact** | What AI tools are, key capabilities, major platforms |
| 1 | **Prompting Foundations** | Prompt structure, instructions, context, zero-shot |
| 2 | **Advanced Prompting** | Few-shot, chain-of-thought, system prompts, personas |
| 3 | **AI Tool Ecosystem** | ChatGPT, Claude, Gemini, Copilot, Midjourney, and more |
| 4 | **Workflow & Automation** | n8n, Zapier, Make, API integration, automated pipelines |
| 5 | **AI Agents & Orchestration** | Agents, multi-agent systems, tool use, memory |
| 6 | **Output Evaluation** | Fact-checking, hallucination detection, quality criteria |
| 7 | **Domain Applications** | AI for coding, writing, research, business, creative work |
| 8 | **Staying Current** | Model updates, new tools, evaluating new capabilities |

### AI Literacy — For Everyone

| Layer | Topic | What You Learn |
|-------|-------|---------------|
| 0 | **What Is AI?** | What AI is and isn't, history, how it works at a high level |
| 1 | **How AI Learns** | Training, data, patterns — no math required |
| 2 | **AI in Daily Life** | Tools you already use, what powers them |
| 3 | **Working with AI** | How to use AI tools effectively, basic prompting |
| 4 | **AI & Society** | Jobs, economy, education, creativity, social impact |
| 5 | **AI Safety & Ethics** | Bias, fairness, misinformation, responsible use |
| 6 | **AI in Your Field** | Healthcare, education, business, law, media applications |
| 7 | **Evaluating AI** | Critical thinking about AI claims and outputs |
| 8 | **The Future of AI** | Trends, possibilities, what to watch for |

---

## Learning Modes

Jogg isn't just one quiz app. It's several powerful learning modes in one:

### Organized Lanes
Work through each lane systematically. See your mastery percentage grow with every question. Unlock certificates as you complete each lane.

### Mixed Practice
Adaptive algorithms surface your weak spots across all lanes simultaneously — perfect for reinforcement and gap-filling.

### RUSH Mode
Timed, high-intensity challenge. Build quick recall — the kind of fluency that matters in interviews and real-world problem solving.

### Papers Quest
Quiz questions derived from 20 landmark research papers including Attention Is All You Need, BERT, GPT-3, LoRA, RAG, FlashAttention, Constitutional AI, and more. Tests your understanding of key contributions and methodological choices.

### Daily Jogg
Your daily 5–10 minute learning habit powered by **FSRS (Free Spaced Repetition Scheduler)** — the most efficient possible use of your daily study time.

### Arcade
Competitive challenge mode with speed, accuracy, and streak multipliers.

### Custom Quiz
Create your own quiz from a selected set of topics, or join a quiz shared by someone else using a 6-character code or QR scan.

---

## Paper Request Feature

Request any specific research paper and Jogg generates curated quiz questions from it. Your learning can be tailored to exactly the papers that matter most for your work or study.

---

## Gamification That Respects Professionals

### 50 Levels of Mastery

| Tier | Levels | What It Represents |
|------|--------|--------------------|
| **Bronze** | 1–12 | Beginning your journey |
| **Steel** | 13–24 | Building solid foundations |
| **Platinum** | 25–36 | Demonstrating deep knowledge |
| **Mythril** | 37–50 | Elite AI/ML mastery |

### XP System
- **Difficulty** — Beginner (10 XP), Intermediate (20 XP), Advanced (35 XP)
- **Speed bonus** — faster correct answers earn up to 1.5× XP
- **Mode multipliers** — RUSH Mode (1.35×), Papers Quest (1.30×), Mixed Practice (1.15×)
- XP is tracked **per framework** and in total

### Certificates
Complete a lane and earn a verifiable certificate issued by MokingBird Oy, verifiable at jogg.app.

### Leaderboard
Opt-in global leaderboard — available to all registered users.

---

## Privacy & Your Data

- We collect only what is necessary to operate the app
- Your data is stored securely with encryption at rest and in transit
- We do not sell your data to any third party, ever
- You can delete your account and all data at any time
- Leaderboard participation is always opt-in

---

## Technical Excellence

- **Cross-platform:** Flutter app for iOS and Android
- **Offline-first:** Questions cached locally — practice even without a connection
- **Encrypted:** AES-256 local storage encryption
- **Fast:** Sub-200ms response times for all app operations
- **Professional themes:** 5 carefully designed themes

---

## App Access Tiers

| Tier | Price | Who It's For |
|------|-------|-------------|
| **Guest** | Free | Try Jogg with ~10 sample questions — no registration required |
| **Free Account** | €0 | Core features, free question quota per layer, leaderboard access |
| **Monthly** | €4.99/mo | Unlimited questions across all frameworks and layers |
| **Lifetime** | €29.99 | One-time purchase — 100 questions per layer + buy more option |
| **Enterprise** | Contact us | Teams, universities, bootcamps — bulk licensing |

---

## About MokingBird Oy

Jogg is developed by **MokingBird Oy**, a Finnish technology company. Jogg is a registered brand of MokingBird Oy. Our mission is to build professional-grade educational tools that meet the real demands of learners in technical fields.

We are based in Finland and are fully GDPR-compliant.

**Contact:** [email protected]
**Website:** jogg.mokingbird.xyz

---

*Jogg — Your Professional AI/ML App. Built for people who take AI seriously.*

# Privacy Policy — Jogg
**Effective Date:** April 12, 2026
**Last Reviewed:** April 12, 2026

Jogg is developed and operated by **MokingBird Oy**, a company registered in Finland. In this Privacy Policy, "Mokingbird", "Jogg", "we", "our", and "us" refer to MokingBird Oy.

This policy is written for:
- website publication and in-app legal display,
- app store review and user support,
- GDPR-facing transparency.

Please read this policy carefully. By using Jogg, you agree to the practices described herein.

---

## 1. Scope

This policy applies to:
- the Jogg mobile application (iOS and Android),
- official Jogg web pages and subpages,
- account, support, and product-related interactions directly operated by MokingBird Oy.

It does not apply to third-party services you may link to from within the app. Those services have their own privacy policies.

---

## 2. What Jogg Is

Jogg is a professional AI/ML learning application that combines:
- a 3-framework, 27-layer AI/ML curriculum (AI Development, AI Usage, AI Literacy),
- quiz and practice flows across multiple learning modes,
- daily spaced repetition review (Daily Jogg),
- research paper-based learning and quiz generation,
- custom quiz creation and sharing,
- progress tracking, XP, streaks, and certificates,
- optional leaderboard features,
- optional reminders and account security controls.

---

## 3. Data Controller

**MokingBird Oy**
Operating Jogg as its product and brand.
Privacy contact: [email protected]

For general support: [email protected] (when active)

MokingBird Oy is the data controller for all personal data processed through the Jogg application.

---

## 4. Personal Data We Process

We follow a strict data minimisation principle. We only process data that is necessary for the operation and improvement of the app. Depending on how you use Jogg, we may process the following categories of data.

### 4.1 Account and Identity Data

If you create an account, we process:
- email address,
- username and display name,
- password (stored as a cryptographic hash — never in plain text),
- authentication provider identifiers (for OAuth sign-in),
- account verification status,
- account creation and last-update timestamps.

Authentication is handled through Supabase-backed account infrastructure. Supported sign-in methods include:
- email and password,
- magic link (passwordless email),
- email OTP (6-digit one-time code),
- OAuth providers: Google, Apple, Facebook, and X/Twitter where configured.

### 4.2 Profile Data

Depending on what you choose to provide or enable in your profile, we may process:
- avatar or profile image,
- bio or short description,
- learning goal, aim category, and learning-interest tags,
- any optional profile fields you choose to complete.

We do not require your date of birth, phone number, or postal address. If date of birth is collected in a future version for age verification, it will be clearly disclosed.

### 4.3 Learning and Progress Data

To operate the core app functions, we process:
- quiz sessions and question attempts,
- correct and incorrect answers,
- mode-specific activity (organized, mixed, RUSH, papers, daily, arcade, custom),
- streak count and daily activity history,
- XP totals, level, and tier state,
- lane mastery percentages and unlock progress,
- certificate status and completion history,
- question and paper quiz history,
- time-taken per question (used for XP calculation and difficulty calibration),
- Daily Jogg session history.

This data is stored in your user account and is only accessible by you, unless you explicitly opt into leaderboard features (see Section 4.5).

### 4.4 Settings and Preference Data

We may process app settings such as:
- theme selection,
- language preferences,
- sound and haptic settings,
- PIN and biometric security preferences,
- auto-lock timer settings,
- notification preferences and types,
- leaderboard sharing preferences,
- privacy-related toggles where available.

### 4.5 Leaderboard Data (Opt-In Only)

Leaderboard participation is entirely **opt-in** and requires your explicit consent.

If you choose to join leaderboard features, we may process and display limited information publicly, such as:
- username,
- ranking position,
- selected performance indicators (XP, level, tier, streak) according to your sharing settings.

Jogg supports global, country-level, and friends leaderboard views. Each has separate sharing controls inside the app.

We do not display your email address or any authentication details on the leaderboard.

You may withdraw from leaderboard participation at any time in your profile settings. Your entry is removed immediately.

### 4.6 Notification Data

If you enable reminders or notifications, we may process:
- device notification permissions,
- push-notification preferences,
- local reminder schedules,
- notification records associated with your account.

Examples of notification types include:
- Daily Jogg reminders,
- daily practice reminders,
- streak reminders and warnings,
- achievement and level-up notifications,
- quiz invitation notifications.

You can control all notification settings inside the app.

### 4.7 Anonymous Question Performance Data

To improve question quality and the learning experience, we may collect anonymous, aggregated data about how questions are answered. This collection is designed so it cannot identify you personally.

**What we collect anonymously:**
- question ID,
- whether the answer was correct or incorrect,
- time taken to answer (seconds),
- approximate user level at the time (not your identity).

**Anonymisation method:** Your user ID is transformed using a one-way cryptographic hash (SHA-256 with a salt) before any analytics data is stored. This result cannot be reversed to identify you.

**What we do NOT collect in analytics:** your name, email, IP address, or device identifiers.

This anonymous data is used exclusively to improve question difficulty calibration, identify poorly worded or ambiguous questions, and improve future question generation. It is not used for advertising or profiling.

### 4.8 Support, Legal, and Operational Data

We may process:
- support correspondence (emails, in-app feedback),
- app error context submitted via crash reports (opt-in),
- account deletion requests,
- data export actions,
- abuse-prevention and security-related signals.

### 4.9 Data We Do NOT Collect

We do not collect:
- your location or GPS data,
- your contacts or phone book,
- your browsing history outside the app,
- access to your camera or microphone (beyond any optional profile photo upload),
- your messages or content from other apps,
- behavioural tracking data for advertising purposes.

---

## 5. Why We Process Data (Purposes)

We process data in order to:
- create and manage user accounts,
- authenticate users and manage sessions,
- save and synchronise learning progress,
- deliver quizzes, question sets, and learning content,
- personalise question selection based on your interests and goals,
- support the 3-framework, 27-layer learning path,
- support daily review and streak mechanics,
- support research paper-based learning and paper-request flows,
- enable shared and custom quiz features,
- operate leaderboard features when you opt in,
- send reminders and notifications you have enabled,
- support account data export and deletion,
- protect the service and its users,
- debug, maintain, and improve reliability,
- improve question quality through anonymous performance data.

---

## 6. Lawful Bases Under GDPR

For users in the European Union or European Economic Area, we rely on one or more of the following lawful bases under the General Data Protection Regulation (GDPR).

### 6.1 Performance of a Contract (Art. 6(1)(b))
We process data necessary to provide the app and its core functions, including:
- account access and session management,
- quiz delivery and progress saving,
- certificate and history access,
- settings persistence.

### 6.2 Legitimate Interests (Art. 6(1)(f))
We may process data where necessary for our legitimate interests, including:
- service security and abuse prevention,
- product reliability and internal troubleshooting,
- anonymous quality improvement and content analysis.

We have assessed that these interests are not overridden by your rights and interests.

### 6.3 Consent (Art. 6(1)(a))
For certain optional features, we rely on your consent:
- leaderboard participation,
- optional crash reporting and analytics,
- push notifications,
- optional future integrations where consent is appropriate.

You may withdraw consent at any time without affecting the lawfulness of processing before withdrawal.

### 6.4 Legal Obligation (Art. 6(1)(c))
We may process data when required to comply with applicable law, regulation, lawful requests, or enforcement obligations.

---

## 7. Guest Mode

Jogg supports guest mode for users who wish to try the app without creating an account.

In guest mode:
- some activity may be stored locally on-device only,
- account-bound features (progress sync, certificates, leaderboard) are not available,
- continuity across devices is not available,
- no personal data is linked to a server-side account.

Guest-mode data handling differs from full account mode because guest activity is not tied to a Supabase-authenticated identity.

---

## 8. Where Data Is Stored

Jogg uses:
- **Supabase** — for account authentication and application data storage (database and storage infrastructure),
- **local on-device storage** — for cached app data and offline use,
- **encrypted local storage** — for sensitive local state and cryptographic keys.

Supabase operates on cloud infrastructure. Depending on Supabase's hosting configuration, your data may be processed in servers within or outside your country of residence. Where required by GDPR for cross-border transfers, we rely on appropriate safeguards such as Standard Contractual Clauses or equivalent mechanisms as available through Supabase's data processing agreements.

---

## 9. Security Measures

Jogg uses multiple layers of technical and organisational security measures, including:

- **Local data encryption:** AES-256-GCM encryption for local database storage (Hive),
- **Secure key storage:** Device-backed secure storage (iOS Keychain / Android Keystore) for encryption keys and refresh tokens,
- **Access tokens:** Short-lived session tokens are handled by the authentication stack; storage behavior depends on platform SDK internals and current rollout configuration,
- **Transport security:** Communication with our backend is encrypted in transit using TLS,
- **Authentication:** OAuth 2.0 with PKCE for social logins; passwords stored as cryptographic hashes,
- **Row Level Security:** Supabase database-level policies preventing cross-user data access,
- **Optional biometric protection:** Touch ID / Face ID / fingerprint for app access,
- **Optional PIN protection:** User-set PIN with auto-lock timer,
- **Account deletion:** Real authenticated account deletion via backend RPC (not cosmetic UI-only),
- **Data export:** Users can export their data at any time.

For a full description of our security practices, see our [Security Document](jogg_Security.md).

---

## 10. Data Sharing

We do **not** sell, rent, or trade your personal data to any third party for their own commercial use.

We may share data only in the following limited circumstances:

### 10.1 Service Providers
We use third-party services to operate Jogg, who process data strictly on our behalf under data processing agreements:

| Provider | Purpose | Data Shared |
|----------|---------|-------------|
| Supabase | Database, authentication, storage | Account and progress data |
| Crash/error reporting (when enabled) | App diagnostics | Diagnostic error data as configured in current rollout |

### 10.2 Legal Requirements
We may disclose data if required by law, court order, or governmental authority, or if we believe in good faith that disclosure is necessary to protect our legal rights, your safety, or the safety of others.

### 10.3 Business Transfers
In the event of a merger, acquisition, or sale of assets, user data may transfer as part of that transaction. You will be notified in advance and retain the right to delete your account before any transfer is completed.

We do not share personal data with unrelated third parties for unrelated commercial exploitation.

---

## 11. International Data Processing

Depending on the infrastructure used by Jogg and its service providers, your data may be processed in countries outside your country of residence, including countries outside the European Economic Area.

Where required by applicable law, we implement appropriate safeguards for such transfers, including relying on Supabase's data processing agreements and applicable transfer mechanisms.

---

## 12. Data Retention

We retain personal data only for as long as reasonably necessary:

| Data Category | Retention |
|---------------|-----------|
| Account data (email, username, progress) | Until account deletion |
| Learning progress and history | Until account deletion |
| Raw anonymous analytics (question answers) | 90 days |
| Aggregated anonymous analytics | Indefinite (fully anonymised) |
| Crash reports (opt-in) | 90 days |
| Support correspondence | As required by legal or operational need |
| Deleted account data | Processed via backend account-deletion flow; some records may remain temporarily in backups/logs where legally required |

Retention periods may differ depending on data type and any applicable legal obligations (e.g., fraud prevention, legal disputes).

---

## 13. Your Rights

Depending on your jurisdiction, and in particular under GDPR if you are in the EU/EEA, you have the following rights:

### 13.1 Right of Access (Art. 15)
You may request a copy of all personal data we hold about you. We provide this in a structured, machine-readable format (JSON) within 30 days.
**How to exercise:** "Download My Data" in app Settings, or email [email protected].

### 13.2 Right to Rectification (Art. 16)
You may correct inaccurate or incomplete personal data. Username and profile information can be updated directly in the app.

### 13.3 Right to Erasure (Art. 17) — "Right to Be Forgotten"
You may request deletion of all your personal data.
**How to exercise:** "Delete Account" in app Settings. Account deletion is processed through backend deletion flow. Limited records may be retained where required for security, fraud prevention, legal, or bookkeeping purposes.
Note: Some records may be retained when required for security, fraud prevention, legal, or bookkeeping reasons, as permitted by GDPR Art. 17(3).

### 13.4 Right to Data Portability (Art. 20)
You may export your personal data in JSON format at any time via app Settings.

### 13.5 Right to Restriction of Processing (Art. 18)
In certain circumstances, you may request that we restrict processing of your personal data.

### 13.6 Right to Object (Art. 21)
You may object to processing based on legitimate interests, including anonymous analytics. You can disable analytics in app Settings.

### 13.7 Right to Withdraw Consent
Where we rely on consent, you may withdraw it at any time via app Settings. Withdrawal does not affect the lawfulness of prior processing.

### 13.8 Right to Lodge a Complaint
You have the right to lodge a complaint with your national data protection authority. In Finland:

**Office of the Data Protection Ombudsman (Tietosuojavaltuutetun toimisto)**
Website: tietosuoja.fi

---

## 14. Children

Jogg is designed as a professional and educational AI/ML learning product intended for users aged **16 and over**.

If local law requires parental or guardian consent for a user under a different age threshold, that consent must be obtained before use.

We do not knowingly collect personal data from children below the applicable age threshold. If you are a parent or guardian and believe your child has provided personal information without appropriate consent, please contact us at [email protected] and we will promptly address the matter.

---

## 15. Changes to This Policy

We may update this Privacy Policy from time to time.

If we make material changes, we will:
- update the "Last Reviewed" date at the top,
- notify you via in-app notification,
- for significant changes affecting your rights, provide advance notice and, where required, seek renewed consent.

Continued use of the app after updated Terms take effect constitutes acceptance, subject to applicable law.

---

## 16. Contact

For any privacy-related questions, requests, or concerns:

**MokingBird Oy**
Privacy contact: [email protected]
Subject line: "Privacy — Jogg"
website https://jogg.mokingbird.xyz

We aim to respond to all privacy-related enquiries within 30 days.

---

## 17. Plain-Language Summary

In short:
- Jogg stores your account and learning data so the app can function,
- we do not sell your personal data,
- leaderboard participation is always opt-in,
- guest mode is available with limited scope and no server-side account,
- we use Supabase-backed authentication and storage infrastructure,
- privacy and security controls are available directly inside the app,
- we may use anonymised aggregated performance data to improve question quality,
- you can export or delete your data at any time.

---

*This Privacy Policy is governed by the laws of Finland and EU GDPR. MokingBird Oy, April 2026.*

# Terms of Service — Jogg
**Effective Date:** April 12, 2026
**Last Reviewed:** April 12, 2026

Welcome to **Jogg**, a professional AI/ML learning platform developed and operated by **MokingBird Oy**, a company registered in Finland. In these Terms of Service, "Mokingbird", "Jogg", "we", "our", and "us" refer to MokingBird Oy.

By downloading, installing, or using the Jogg application, you agree to be bound by these Terms of Service ("Terms"). If you do not agree to these Terms, do not use the service.

---

## 1. Agreement

These Terms constitute a legally binding agreement between you and MokingBird Oy governing your access to and use of Jogg.

By accessing or using Jogg, you confirm that:
- you can form a binding agreement under applicable law,
- your use complies with all applicable local, national, and EU law,
- you have provided truthful information where required,
- you have read and agree to our [Privacy Policy](jogg_privacy_policy.md).

If you use Jogg on behalf of an organisation, you confirm you have the authority to bind that organisation to these Terms.

---

## 2. What Jogg Provides

Jogg is an AI/ML learning and quiz platform that may include:
- a 3-framework, 27-layer AI/ML curriculum (AI Development, AI Usage, AI Literacy),
- organized lane-based learning with mastery tracking,
- mixed, random, and adaptive practice modes,
- Daily Jogg spaced repetition review sessions,
- arcade-style learning challenge modes,
- research paper-based quiz experiences (Papers Quest),
- user-requested paper quiz generation (Paper Request),
- custom quiz creation and sharing via code, link, and QR,
- progress tracking, XP, streaks, levels, and tier systems,
- certificate generation and verification,
- optional leaderboard participation,
- optional reminders and account security features.

Feature availability may vary by version, platform, region, or rollout stage.

---

## 3. Eligibility

You may use Jogg only if you are at least **16 years of age**, or meet any higher minimum age required by the laws of your country of residence.

If you are under the applicable minimum age, you must not use Jogg. If a parent or guardian discovers that a minor has created an account, please contact [email protected] immediately.

---

## 4. Accounts and Access Modes

### 4.1 Guest Mode
Jogg supports guest access for limited exploration without creating an account.

In guest mode:
- you can access a limited set of sample questions (approximately 10),
- activity may be stored locally on-device only,
- no progress is saved to a server-side account,
- certain features require a registered account, including: progress sync, certificates, leaderboard, history, and account management functions.

### 4.2 Registered Account Mode
By creating an account, you gain access to account-bound features including:
- persistent progress tracking and quiz history,
- certificate generation and download,
- leaderboard participation (opt-in),
- data export and account deletion,
- cross-session and cross-device continuity,
- access to premium features if subscribed.

### 4.3 Account Security
You are responsible for:
- keeping your login credentials secure and confidential,
- not sharing your account with others,
- notifying us promptly at [email protected] if you suspect unauthorised access.

We are not liable for loss resulting from unauthorised use of your account where you have failed to take reasonable steps to protect your credentials.

---

## 5. User Responsibilities

You agree to use Jogg only for lawful purposes. You must not:
- use the service to harass, harm, or impersonate others,
- attempt to gain unauthorised access to accounts, content, or backend infrastructure,
- reverse engineer, decompile, or disassemble protected parts of the app except where expressly permitted by law,
- use automated tools, bots, or scripts to access or interact with the service,
- scrape or republish substantial portions of question content or app materials without written permission,
- use shared quiz systems for spam, harassment, impersonation, or illegal content distribution,
- artificially inflate XP, leaderboard scores, or progress metrics,
- spread harmful, deceptive, or malicious material through any app feature.

We may suspend or terminate access where these obligations are breached.

---

## 6. Educational Nature of the Service

Jogg is an educational product designed to help users learn AI/ML concepts through interactive and gamified experiences.

Jogg does not guarantee:
- exam success, interview success, or academic performance,
- job placement or professional advancement,
- academic credit or accredited certification,
- the accuracy of all generated, adapted, or summarised content in every case.

Users remain responsible for independently verifying important information, especially where research papers, generated materials, or summarised content are involved. Real-world decisions with professional, academic, or safety-critical consequences should be reviewed by a qualified human expert.

---

## 7. Content and Question Systems

Jogg may include:
- human-curated questions,
- structured lane-based question banks,
- generated or generation-assisted question sets,
- research paper-linked question flows,
- adaptive and personalised question selection.

Generated or adaptive content may evolve over time. We may modify, remove, rebalance, or replace content, question pools, scoring logic, or quiz formats at any time where reasonably necessary to maintain quality or accuracy.

---

## 8. Shared Quizzes and User-Created Activity

Jogg includes custom quiz and sharing functionality, including:
- share codes (6-character),
- links and QR-based join flows,
- public or private quiz settings,
- organiser and participant flows.

If you create or share a quiz, you are responsible for using these features lawfully and respectfully. You must not use quiz-sharing systems for abuse, harassment, spam, impersonation, illegal content, or deceptive distribution.

We may remove, restrict, or disable shared quiz content or access where necessary to protect the service or its users.

---

## 9. Leaderboards and Public Display

Leaderboard participation is entirely optional and requires explicit opt-in. If you opt in, selected information (username and approved performance indicators) may be displayed to other users.

You are responsible for choosing an appropriate username and profile presentation. We may moderate, limit, or remove leaderboard entries in cases involving abuse, technical issues, or policy enforcement.

---

## 10. Payments, Subscriptions, and Pricing

### 10.1 Pricing Tiers
Jogg may operate under one or more of the following commercial models:
- **Free / Freemium:** Limited access, no payment required.
- **Premium (Monthly):** Full access billed monthly, auto-renewing.
- **Premium (Annual):** Full access billed annually at a discounted rate, auto-renewing.
- **Lifetime / One-Time Unlock:** Full access with a single payment, no recurring fees.
- **In-App Add-Ons:** Optional purchases such as extra question packs or premium features.
- **Enterprise / Institutional:** Custom licensing for organisations.

Not every pricing option is necessarily live on every platform or at every stage. Pricing rollout is completed in stages. Current and final pricing will be clearly published in-app and on our website when each option is live.

### 10.2 Billing
Paid features are processed through Apple App Store or Google Play Store. We do not directly store your payment card information. Billing, currency conversion, and platform fees are subject to those platforms' rules.

### 10.3 Auto-Renewal
Monthly and annual subscriptions automatically renew at the end of each period unless cancelled before the renewal date. You can manage or cancel subscriptions in your App Store or Google Play account settings.

### 10.4 Cancellation
You may cancel a subscription at any time. Cancellation takes effect at the end of the current billing period. No partial refunds are issued for unused time in the current period unless required by applicable law.

### 10.5 Refunds
Refund requests are subject to the policies of Apple App Store or Google Play Store. If you have a billing concern, contact [email protected] and we will do our best to assist.

### 10.6 Price Changes
We reserve the right to change subscription prices. For existing subscribers, we will provide at least **30 days' notice** before a price change takes effect. Continued use after the change takes effect constitutes acceptance of the new price.

### 10.7 Changes to Features or Pricing
We may change features, access tiers, commercial packaging, or free-versus-paid boundaries. Where we do so, we will aim to give reasonable notice and will clearly communicate material changes that affect paid subscribers.

---

## 11. Intellectual Property

### 11.1 Our Rights
Jogg, including its branding, app structure, question content, quiz systems, graphics, product design, and all associated intellectual property, is owned by or licensed to MokingBird Oy unless otherwise stated.

You are granted a **limited, non-exclusive, non-transferable, revocable licence** to use Jogg solely for your personal, non-commercial educational purposes.

### 11.2 What You May Not Do
- Copy, reproduce, or commercially redistribute our question content or app content.
- Use our content to create competing products or services.
- Use our trademarks, logos, or brand materials without our written permission.
- Scrape or republish large portions of question content without permission.

### 11.3 Third-Party Materials
Research papers and third-party materials referenced through Jogg remain subject to their own intellectual property rights and licences. Jogg's paper-based content is original educational material inspired by published works, not reproductions of those works.

### 11.4 User Feedback
If you provide feedback, ideas, or suggestions to us, we may use them to improve Jogg without any obligation to compensate you, unless otherwise agreed in writing.

---

## 12. Certificates

Certificates issued by Jogg upon completion of learning lanes are:
- issued by MokingBird Oy / Jogg as recognition of achievement within the platform,
- verifiable at jogg.app using the certificate's unique code,
- for personal development and learning motivation purposes,
- **not** academic degrees, diplomas, or formally accredited credentials recognised by any academic institution or professional body.

Employers and academic institutions are under no obligation to recognise Jogg certificates. We reserve the right to modify or discontinue the certificate programme with reasonable notice.

---

## 13. Service Availability

We do not guarantee that Jogg will always be uninterrupted, error-free, or available on all devices or regions. The service may be unavailable due to:
- scheduled maintenance,
- third-party service outages,
- technical issues,
- device or platform restrictions,
- security actions,
- events beyond our reasonable control (force majeure).

---

## 14. Account Suspension and Termination

### 14.1 By You
You may stop using Jogg at any time and delete your account via app Settings. Deletion is handled in accordance with our [Privacy Policy](jogg_privacy_policy.md).

### 14.2 By Us
We may suspend, restrict, or terminate your account without prior notice if you:
- breach these Terms,
- engage in fraudulent, abusive, or illegal activity,
- attempt to compromise the security or integrity of the service.

In cases of termination, any paid subscription may be pro-rated and refunded at our discretion in compliance with applicable law.

### 14.3 Effect of Termination
- Access to premium features ceases immediately on termination.
- Data is handled per our Privacy Policy (30-day soft delete, then permanent erasure).
- Some records may be retained where required for security, fraud prevention, or legal reasons.
- Termination does not automatically erase every record where retention is required by law or legitimate operational need.

---

## 15. Disclaimer of Warranties

TO THE EXTENT PERMITTED BY APPLICABLE LAW, JOGG IS PROVIDED ON AN "AS IS" AND "AS AVAILABLE" BASIS WITHOUT WARRANTY OF ANY KIND.

WE DO NOT WARRANT THAT:
- ALL CONTENT IS ERROR-FREE OR COMPLETE,
- GENERATED CONTENT IS ALWAYS ACCURATE,
- THE SERVICE WILL ALWAYS MEET YOUR SPECIFIC GOALS,
- UNINTERRUPTED ACCESS WILL ALWAYS BE AVAILABLE.

Nothing in these Terms excludes or limits any statutory rights you may have as a consumer under applicable law.

---

## 16. Limitation of Liability

TO THE EXTENT PERMITTED BY APPLICABLE LAW, MOKINGBIRD OY SHALL NOT BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR EXEMPLARY DAMAGES ARISING FROM YOUR USE OF JOGG.

OUR TOTAL LIABILITY TO YOU FOR ANY CLAIMS UNDER THESE TERMS SHALL NOT EXCEED THE AMOUNT YOU PAID US IN THE 12 MONTHS PRECEDING THE CLAIM.

Nothing in these Terms excludes liability that cannot be excluded under applicable law, including liability for death or personal injury caused by our negligence, fraud, or any liability that Finnish or EU consumer protection law does not permit to be excluded.

---

## 17. Privacy

Your use of Jogg is governed by our [Privacy Policy](jogg_privacy_policy.md), which is incorporated into these Terms by reference. If there is any conflict between these Terms and the Privacy Policy in matters of data processing, the Privacy Policy controls.

---

## 18. Governing Law and Dispute Resolution

These Terms are governed by and construed in accordance with the laws of **Finland** and applicable European Union law, without regard to conflict-of-law principles.

Any disputes arising from these Terms shall be subject to the jurisdiction of Finnish courts, without prejudice to your mandatory consumer protection rights under the law of your country of habitual residence.

If you are a consumer in the EU, you may also use the European Online Dispute Resolution platform:
**ec.europa.eu/consumers/odr**

---

## 19. Changes to These Terms

We may update these Terms from time to time. When we make material changes, we will:
- update the "Last Reviewed" date at the top,
- provide in-app notification,
- for significant changes, provide at least **30 days' notice** to existing users.

Continued use of Jogg after updated Terms take effect constitutes acceptance, subject to applicable law.

---

## 20. Severability

If any provision of these Terms is found invalid or unenforceable under applicable law, that provision will be modified to the minimum extent necessary to make it enforceable, or severed if modification is not possible. The remaining provisions remain in full force.

---

## 21. Entire Agreement

These Terms, together with our Privacy Policy and any other policies referenced herein, constitute the entire agreement between you and MokingBird Oy regarding your use of Jogg and supersede all prior agreements on the same subject matter.

---

## 22. Contact

For questions about these Terms:

**MokingBird Oy**
Email: [email protected]
Subject: "Terms — Jogg"
website https://jogg.mokingbird.xyz

---

## 23. Plain-Language Summary

In short:
- Jogg is an educational AI/ML learning app,
- guest mode exists but full features require an account,
- subscription and one-time purchase options exist (rollout staged by platform),
- leaderboard participation is always opt-in,
- we may update content and pricing with notice,
- you must use the service lawfully and responsibly,
- Jogg does not guarantee academic, professional, or career outcomes,
- your consumer rights under applicable law are not affected by these Terms.

---

*These Terms are governed by Finnish law and EU consumer law. MokingBird Oy, April 2026.*

# Security Overview — Jogg
**MokingBird | Developed by MokingBird Oy**
**Last Reviewed:** April 12, 2026

This document describes Jogg's security posture based on the implemented application code and current integration design.

It is intended for:
- website security overview pages,
- app store review support,
- internal launch readiness,
- user-facing trust documentation.

---

## 1. Security Positioning

Jogg is a learning application with account, progress, quiz, and content features. Security in Jogg is focused on:
- protecting user accounts and session state,
- protecting local app data through encryption,
- controlling access to authenticated features,
- reducing accidental exposure of sensitive information,
- supporting user data control (export and deletion),
- preparing the app for stronger production hardening over time.

---

## 2. Security Architecture — Four Layers

Jogg uses a four-layer security model that protects data at every stage.

```
┌────────────────────────────────────────────────────┐
│  LAYER 1: DEVICE SECURITY                          │
│  Biometric auth · Secure storage · App sandboxing  │
└────────────────────────────────────────────────────┘
                        ↓
┌────────────────────────────────────────────────────┐
│  LAYER 2: DATA ENCRYPTION                          │
│  AES-256-GCM local DB · PBKDF2 key derivation      │
│  RSA-2048 key utilities · Secure key storage       │
└────────────────────────────────────────────────────┘
                        ↓
┌────────────────────────────────────────────────────┐
│  LAYER 3: TRANSMISSION SECURITY                    │
│  TLS in transit · Cert pinning hardening (rollout) · OAuth 2.0 + PKCE │
└────────────────────────────────────────────────────┘
                        ↓
┌────────────────────────────────────────────────────┐
│  LAYER 4: CLOUD AND BACKEND SECURITY               │
│  Supabase Row Level Security · User-isolated data  │
│  Authenticated-user-only data access               │
└────────────────────────────────────────────────────┘
```

---

## 3. Implemented Security Controls

### 3.1 Authentication Infrastructure

Jogg uses Supabase-backed authentication for account and session handling.

Supported sign-in methods:
- email and password,
- magic link (passwordless),
- email OTP (6-digit code),
- OAuth via Google, Apple, Facebook, and X/Twitter where configured.

The app distinguishes between:
- authenticated account routes,
- guest-access routes,
- routes that require full account authentication.

### 3.2 Route and Session Protection

Jogg uses route guards to restrict access to protected screens and flows:
- account-required routes (quiz modes, progress, certificates),
- routes that block guest mode,
- lane access control and feature gating.

### 3.3 Local Storage Encryption

Jogg uses encrypted local storage (Hive) for important app data.

The implementation includes:
- AES-256-GCM encryption applied to local storage boxes,
- device-backed secure key storage (iOS Keychain / Android Keystore),
- local cache cleanup support.

The encryption key is generated on first app launch and stored in platform secure storage — it never resides in plain application memory or accessible files.

### 3.4 Cryptographic Utilities

The codebase includes a dedicated encryption service with:
- AES-256-GCM encryption and decryption helpers,
- RSA-2048 key-pair generation utilities,
- PBKDF2-based key derivation (100,000 iterations with random salt).

### 3.5 Token Management

Jogg uses a two-token session model:
- **Access token:** Short-lived session token managed by the auth stack; exact storage behavior depends on SDK/platform implementation.
- **Refresh token:** Refresh-session behavior is managed by the auth provider and may vary by platform/runtime configuration.

Tokens are invalidated on logout. Sessions can be managed and revoked from the app's settings.

### 3.6 Optional User Device Security

Jogg includes user-facing security settings:
- PIN protection (user-set, 4–6 digits),
- biometric unlock (Face ID / Touch ID / fingerprint) where device supports it,
- auto-lock timer controls.

After 3 failed biometric attempts, the app falls back to the device passcode or the user's PIN.

### 3.7 Data Portability and User Control

Jogg includes:
- user data export (JSON format),
- account deletion flow (real backend RPC deletion — not cosmetic UI-only deletion),
- privacy settings and cache-clearing actions.

The account deletion path removes the authenticated Supabase account, not only local UI state.

### 3.8 Leaderboard Privacy Controls

Leaderboard participation is opt-in. Users are not placed in public rankings by default.

The app includes separate toggles controlling which leaderboard-related information is shared (XP, streak, quiz metrics). Users can withdraw from the leaderboard at any time.

### 3.9 Notification Permission Control

Users explicitly control:
- push notification permissions,
- specific notification types (daily reminders, streak alerts, achievement notifications, quiz invitations).

---

## 4. Backend and Data Access Model

Jogg relies on Supabase for backend services and database access.

Security measures at the backend layer include:
- **Row Level Security (RLS):** Database-level policies that restrict each user to their own data. Cross-user data access is architecturally prevented at the database layer.
- **Server-backed identity handling:** Auth tokens are validated server-side; raw auth secrets are not stored in app code.
- **Separation of client-safe and privileged keys:** Service-role credentials are not used in the client app.
- **Environment configuration:** Structured to separate dev, staging, and production credentials.

---

## 5. Security Around Quiz and Sharing Features

Jogg includes multiple quiz and sharing surfaces:
- custom quiz creation and ownership,
- share links and 6-character share codes,
- QR-based join flows,
- organiser and participant result views.

Security-relevant controls include:
- account-bound quiz ownership (only the creator can manage their quiz),
- generated share codes (not guessable by enumeration),
- authenticated access to organiser-specific data paths,
- backend RPC usage for organiser views rather than direct client queries.

---

## 6. Privacy-Preserving Analytics

Anonymous analytics use a one-way SHA-256 hash with a salt to transform user IDs before any analytics data is stored. This means:
- no analytics record can be traced back to a specific user,
- IP addresses and device identifiers are not collected for analytics purposes,
- raw anonymous answer data is retained for 90 days, then deleted.

---

## 7. Threat Model

| Threat | Mitigation |
|--------|-----------|
| Unauthorised account access | OAuth 2.0 + PKCE, biometric option, rate limiting |
| Data theft from device | AES-256-GCM local encryption, keys in Keychain/Keystore |
| Man-in-the-middle attack | TLS 1.3 + certificate pinning direction |
| Cross-user data access | Supabase Row Level Security at DB layer |
| Session hijacking | Short-lived session tokens and provider-managed session controls |
| Lost or stolen device | Biometric gate + remote session revocation |
| Fake quiz or share abuse | Account-bound ownership, authenticated access paths |

**Residual risks (accepted):**
- Physical access by someone who has unlocked the device,
- Compromised cloud provider infrastructure (mitigated by Supabase's own security practices),
- Zero-day OS vulnerabilities (mitigated by prompt OS update guidance).

---

## 8. Current Security Strengths

Based on the implementation, the strongest current security strengths are:

1. Account-based authentication with multiple sign-in options,
2. Explicit guest-versus-authenticated separation,
3. AES-256-GCM encrypted local app storage,
4. Secure local key storage (Keychain/Keystore),
5. Optional biometric and PIN app lock,
6. Real backend account deletion (not cosmetic),
7. User data export and deletion support,
8. Opt-in public leaderboard model with granular sharing controls,
9. Privacy-preserving anonymous analytics design,
10. Security-aware environment configuration structure.

---

## 9. Areas Being Finalised for Production

Some security-related areas are scaffolded or in progress, not yet fully deployed. We are transparent about this:

- **Payment receipt validation:** Commercial architecture direction is defined; production-grade purchase verification for in-app billing is a separate rollout step.
- **Crash reporting rollout:** Analytics and error reporting infrastructure exists; full production rollout is staged.
- **Certificate pinning hardening:** Direction is implemented; final production configuration details are part of pre-launch hardening.
- **Enterprise-grade audit tooling:** Not yet required at current scale; planned for growth phase.
- **Backup integration security:** User cloud backup features (Google Drive / iCloud) are architected; security of those flows depends on completion of that feature rollout.

Public security messaging should accurately reflect what is deployed versus what is in progress.

---

## 10. Pre-Launch Security Checklist

Before full public launch, the following items should be confirmed:

1. [ ] Receipt validation for paid flows,
2. [ ] Final production certificate-pinning configuration,
3. [ ] Production secret and environment review (no hardcoded credentials),
4. [ ] Supabase RLS policy review and audit,
5. [ ] Account deletion flow end-to-end verification,
6. [ ] Data export flow verification,
7. [ ] Notification permission behaviour review,
8. [ ] OAuth redirect and callback review,
9. [ ] Backup integration security review (when backup rollout is enabled),
10. [ ] Final alignment between legal policy documents and actual telemetry and backend behaviour.

---

## 11. Responsible Disclosure

To report a security vulnerability in Jogg, please contact:

**MokingBird Oy**
Email: [email protected]
Subject: "Security — Jogg"
website https://jogg.mokingbird.xyz

We ask that you practice responsible disclosure and give us reasonable time to investigate and address the issue before any public disclosure.

---

## 12. Plain-Language Summary

Jogg uses Supabase-backed authentication, AES-256-GCM encrypted local storage, secure device-backed key handling, optional PIN and biometric protection, privacy-aware opt-in leaderboard controls, and in-app export and deletion tools.

Some advanced infrastructure — including payment verification and certain monitoring integrations — is being finalised as part of production rollout.

---

*MokingBird Oy — Security overview current as of April 2026. Updated as the platform evolves.*

# Disclaimer — Jogg
**MokingBird | Developed by MokingBird Oy**
**Last Reviewed:** April 12, 2026

---

## 1. Purpose

This disclaimer explains important limitations of Jogg as an educational product.

In this document, "Mokingbird", "we", "our", and "us" refer to **MokingBird Oy**.

---

## 2. Educational Use Only

Jogg is an educational and learning application focused on AI/ML.

It is designed to help users:
- study and review AI/ML concepts,
- practice active recall through quiz-based interaction,
- explore landmark research papers interactively,
- build structured familiarity with the full AI/ML technology stack,
- organise and share quiz experiences.

Jogg is not a substitute for:
- formal academic instruction or degree programmes,
- accredited professional certification,
- official exam preparation materials published by certifying bodies,
- original research reading and independent academic interpretation,
- expert mentoring or tutoring,
- professional, legal, financial, or technical advice of any kind.

---

## 3. No Guarantee of Outcomes

Using Jogg does not guarantee:
- exam success or passing rates,
- interview success or job placement,
- academic performance or grades,
- professional advancement or salary outcomes,
- certification or licensure by any external body,
- publication success or research validation.

Jogg is a study and practice tool. Results depend on the individual user's effort, background, and application of what they learn.

---

## 4. Research Paper Content

Jogg includes research paper learning features, including:
- curated paper-linked question sets,
- paper-request quiz generation,
- summaries and explanations connected to paper-based learning.

These features are designed to support learning about research papers — not to replace reading and critically interpreting the original source.

If you are relying on a research paper for academic, commercial, safety-critical, or scientific work, you must consult the original paper directly. Jogg's paper-based content is original educational material inspired by published works and does not substitute for the primary literature.

---

## 5. Generated and Adaptive Content

Some Jogg content may be generated, adapted, weighted, or personalised through internal content systems.

As a result:
- some content may simplify concepts for clarity,
- some explanations may omit edge cases or nuance,
- some generated or generation-assisted questions may require independent verification,
- some answers may reflect the state of the field at time of creation and may not reflect the most recent developments.

---

## 6. Rapidly Evolving Field

AI/ML is a fast-changing field. Even where Jogg content is strong and carefully curated, it may not always reflect:
- the newest research or model architectures,
- every framework or library update,
- every shift in production best practice,
- every institutional or regulatory standard.

You should independently confirm critical technical or professional information, especially in fast-moving sub-areas such as model release cycles, inference tooling, or AI governance.

---

## 7. No Professional Advice

Jogg does not provide:
- legal advice,
- financial advice,
- medical advice,
- safety certification or compliance certification,
- deployment approval for any AI system,
- research validation or academic endorsement.

Any real-world decision with professional, legal, safety, or high-stakes consequences must be reviewed independently by a qualified human expert.

---

## 8. Shared Quizzes and User Interactions

Jogg allows quiz sharing and participation through codes, links, and QR flows. We do not guarantee that all user-created or user-shared quiz content will always be complete, accurate, or suitable for every audience. Users creating and sharing quizzes are responsible for their content.

---

## 9. Leaderboard

Leaderboard ranking is a product engagement and motivation feature. It should not be interpreted as:
- a formal academic ranking,
- a professional certification or endorsement,
- a hiring benchmark,
- a complete or authoritative measure of AI/ML expertise.

---

## 10. Certificates

Certificates issued by Jogg are recognition of achievement within the Jogg platform, issued by MokingBird Oy. They are not:
- academic degrees or diplomas,
- formally accredited credentials recognised by any academic institution or professional body,
- equivalent to industry certifications issued by external bodies (e.g., cloud providers, professional associations).

---

## 11. Availability and Reliability

While we aim for reliability, Jogg may occasionally experience:
- content errors or inaccuracies,
- service outages or interruptions,
- sync issues or data loading delays,
- temporary feature unavailability.

We do not warrant uninterrupted or error-free access to the application.

---

## 12. User Responsibility

You remain responsible for:
- verifying important answers and technical claims,
- validating research paper interpretations against primary sources,
- using the app appropriately for your specific learning context,
- protecting your own account credentials,
- deciding how much to rely on app content in academic or professional contexts.

---

## 13. Limitation of Liability

To the fullest extent permitted by applicable law, MokingBird Oy shall not be liable for any loss, damage, or harm arising from:
- reliance on app content for professional, academic, or career decisions,
- inability to access the service due to technical issues,
- errors or inaccuracies in educational content,
- loss of progress due to device failure or account issues,
- any indirect, incidental, special, consequential, or punitive damages.

Nothing in this disclaimer excludes liability that cannot lawfully be excluded under applicable Finnish or EU consumer protection law.

---

## 14. Age Restriction

Jogg is intended for users **aged 16 and older**, or such higher age as required by local law. If you are under the applicable minimum age, you must not use this application.

---

## 15. Governing Law

This disclaimer and all matters relating to Jogg are governed by the laws of **Finland** and applicable European Union law.

---

## 16. Contact

For questions or concerns regarding this disclaimer:

**MokingBird Oy**
Email: [email protected]
website https://jogg.mokingbird.xyz

---

## 17. Public-Facing Summary

*Jogg is an AI/ML learning tool designed to support study and practice. It is not a substitute for original research, formal academic instruction, professional advice, or independent verification of important information. Certificates are platform achievement recognitions, not accredited credentials.*

---

*MokingBird Oy — April 2026.*

### How 27 Layers Across AI Development, AI Usage, and AI Literacy Give Every Learner Exactly the Path They Need

*by the Jogg Team | MokingBird Oy*

---

Learning AI/ML looks completely different depending on who you are.

If you are a software engineer preparing for ML system design interviews, you need to understand backpropagation, transformer architecture, and production deployment patterns. If you are a marketing manager who needs to get more out of ChatGPT and evaluate AI-generated content critically, the mathematical foundations of gradient descent are not where you should start. If you are a nurse wondering how AI will change healthcare and what it means for patient safety, neither of the above applies to you.

Most learning platforms pick one of these learners and build for them. The engineer-focused platforms build deep technical content that leaves non-technical learners behind. The beginner-friendly platforms simplify so aggressively that engineers find them useless. The tool-focused content farms cover ChatGPT prompting but nothing beneath it.

Jogg is built differently. Our **3-Framework Learning System** gives every learner a complete, 9-layer path specifically designed for their relationship to AI — without forcing everyone through the same content.

---

## The Architecture: Three Frameworks, 27 Layers, One Platform

Jogg organizes all learning content into three complete frameworks:

| Framework | Who It's For | Layers | Questions |
|-----------|--------------|--------|-----------|
| **AI Development** | Engineers & researchers | 9 | ~1,200+ |
| **AI Usage** | Power users & professionals | 9 | ~1,200+ |
| **AI Literacy** | Everyone | 9 | ~1,200+ |

Each framework has **9 layers** covering a different dimension of AI knowledge. All 27 layers are accessible from day one — no mastery gates, no prerequisites required. All three frameworks are included in every account tier (Free, Monthly, Lifetime).

---

## Framework 1: AI Development — For Engineers & Researchers

**Who it's for:** Software engineers, ML engineers, data scientists, researchers, CS students, and anyone building AI systems.

**What it covers:** The complete AI/ML engineering stack — from the mathematical foundations that underpin machine learning through data pipelines, model training, RAG systems, inference optimization, deployment, and AI safety.

### The 9 Layers

**Layer 0 — Foundations**
The bedrock. Linear algebra (vectors, matrices, eigenvalues), statistics and probability, Python for ML, and core ML concepts (supervised vs. unsupervised learning, overfitting, cross-validation). This layer does not assume a CS background — it builds it.

**Layer 1 — Data Acquisition**
Where most real ML projects spend their time. REST APIs and web scraping, SQL and NoSQL databases, data streaming architectures, file formats (Parquet, HDF5, Arrow), and data governance basics. You cannot train a good model on bad data.

**Layer 2 — Data Preprocessing**
Transforming raw data into something useful. Feature engineering, handling missing values, normalization and standardization, class imbalance, data augmentation for vision and text, and the pipeline patterns that keep preprocessing reproducible.

**Layer 3 — Model Training**
The heart of the stack. Neural network architectures, backpropagation, optimizers (SGD, Adam, AdamW), transformer architecture and attention mechanisms, fine-tuning strategies, and the training dynamics that determine whether your model converges.

**Layer 4 — Orchestration & RAG**
Modern AI systems are not single models — they are pipelines. Retrieval-Augmented Generation (RAG), vector databases and embedding search, agent orchestration frameworks, prompt engineering for system builders, and multi-model coordination patterns.

**Layer 5 — Inference & Optimization**
Getting models to production without breaking the budget. Quantization (INT8, INT4, GPTQ, AWQ), pruning and distillation, KV cache management, speculative decoding, hardware selection (GPU, TPU, edge), and the performance tradeoffs that shape real deployment decisions.

**Layer 6 — Integration & Applications**
Making models useful in real systems. REST and gRPC API design for ML services, containerization and orchestration (Docker, Kubernetes), monitoring and observability, CI/CD pipelines for ML, A/B testing, and the software engineering practices that distinguish production ML from notebook experiments.

**Layer 7 — Future Topics**
The frontier. Multimodal AI (vision-language models, CLIP, LLaVA), diffusion models and generative approaches, Mixture of Experts (MoE) architectures, reinforcement learning and RLHF, and the emerging paradigms reshaping what AI systems can do.

**Layer 8 — AI Safety & Governance**
The responsibility layer. Bias detection and mitigation, adversarial robustness, prompt injection and jailbreak defenses, Constitutional AI and alignment frameworks, regulatory frameworks (EU AI Act, GDPR intersections), and the ethical dimensions of deploying AI at scale.

---

## Framework 2: AI Usage — For Power Users & AI Tool Professionals

**Who it's for:** Professionals who use AI tools daily, prompt engineers, automation specialists, content creators, consultants, and anyone whose job increasingly involves directing AI systems rather than building them.

**What it covers:** Mastery of AI tools at a professional level — from understanding what tools exist through advanced prompting, workflow automation, agent deployment, output evaluation, and staying current in a fast-moving landscape.

### The 9 Layers

**Layer 0 — First Contact**
Orientation. What AI tools are and are not, the major platforms and their key capabilities, how to evaluate a new AI tool, and the mental models that help you understand AI outputs rather than just react to them.

**Layer 1 — Prompting Foundations**
The building blocks. How language models respond to input, the structure of effective prompts, zero-shot and instruction-following patterns, context management, and the practical differences between models that determine which tool to use for which task.

**Layer 2 — Advanced Prompting**
Where casual users stop and professionals begin. Few-shot examples, chain-of-thought prompting, system prompts and instruction hierarchies, persona assignment, role-play and simulation, and the advanced patterns that dramatically improve output quality on complex tasks.

**Layer 3 — AI Tool Ecosystem**
The landscape. In-depth coverage of ChatGPT, Claude, Gemini, Copilot, Midjourney, DALL-E, Stable Diffusion, Runway, ElevenLabs, and more. Not just what each tool does, but when to use which one and how to get professional-grade output from each.

**Layer 4 — Workflow & Automation**
From one-shot prompts to repeatable systems. n8n, Zapier, and Make for no-code automation, API integration for technical users, automated content pipelines, data extraction workflows, and the patterns that turn AI from a manual tool into a scalable process.

**Layer 5 — AI Agents & Orchestration**
The frontier of AI usage. What AI agents are and how they work, agentic frameworks (AutoGPT, CrewAI, LangChain agents), multi-agent systems, tool use and function calling, memory systems, and the practical limits of current agent technology.

**Layer 6 — Output Evaluation**
Critical skills often left unaddressed. Fact-checking AI outputs, hallucination detection patterns, quality criteria for different output types, attribution and source verification, and the systematic approach to deciding when to trust and when to verify.

**Layer 7 — Domain Applications**
AI in context. Specialized techniques for coding (GitHub Copilot, Cursor, AI pair programming), writing and content (SEO, tone, voice consistency), research (literature review, synthesis, structured analysis), business operations, and creative work.

**Layer 8 — Staying Current**
How to not fall behind. Understanding model release notes, evaluating new capabilities critically, tracking the benchmark landscape, following the field without information overload, and building a personal system for staying sharp as AI evolves faster than any single course can cover.

---

## Framework 3: AI Literacy — For Everyone

**Who it's for:** Non-technical professionals, educators, healthcare workers, policy makers, journalists, parents, students outside tech, and anyone who needs to understand AI's role in society without becoming an engineer or power user.

**What it covers:** A clear, honest, jargon-light understanding of what AI is, how it affects the world, how to use basic tools effectively, and how to think critically about AI claims and applications in your field.

### The 9 Layers

**Layer 0 — What Is AI?**
Grounding. What AI actually is (and is not), a brief honest history of the technology, how modern AI systems work at an intuitive level, what distinguishes narrow AI from AGI, and why the hype and the reality often do not match.

**Layer 1 — How AI Learns**
Demystifying training. How machine learning models learn from data, what "training" means without the mathematics, why data quality matters, what the model "knows" and where its knowledge comes from, and the fundamental difference between correlation and understanding.

**Layer 2 — AI in Daily Life**
Making the invisible visible. The AI behind search, recommendation systems, voice assistants, navigation apps, spam filters, fraud detection, and the dozens of other places AI shapes your day without announcing itself.

**Layer 3 — Working with AI**
Practical fundamentals. How to use AI tools effectively for everyday tasks, what makes a good prompt, how to interpret AI outputs, when to use AI and when not to, and the practical judgment that distinguishes effective AI users from frustrated ones.

**Layer 4 — AI & Society**
The big picture. AI's impact on employment and economic structure, AI in education, creativity, media, and entertainment, the concentration of AI capabilities in a small number of organizations, and the policy conversations shaping how AI will be governed.

**Layer 5 — AI Safety & Ethics**
The responsibility layer for non-engineers. Bias and fairness in AI systems, misinformation and deepfakes, privacy implications of AI data use, the ethics of AI in high-stakes decisions (healthcare, criminal justice, hiring), and what responsible AI use means for individuals and organizations.

**Layer 6 — AI in Your Field**
Domain-specific application. Dedicated content for healthcare, education, business and management, law and legal work, media and journalism, government and public sector, and other major professional domains where AI is changing what practice looks like.

**Layer 7 — Evaluating AI**
Critical thinking skills. How to assess AI claims in media coverage, understand what AI benchmarks do and do not measure, evaluate vendor claims about AI capabilities, and develop the informed skepticism that the current AI landscape requires.

**Layer 8 — The Future of AI**
Looking forward. Current trends and near-term trajectories, what AGI means and why estimates vary so widely, the governance landscape and where it is heading, and how to think productively about a technology whose trajectory is genuinely uncertain.

---

## How the Frameworks Work Together

The three frameworks are not siloed. An engineer who masters AI Development can explore AI Usage to understand how practitioners use the systems they build. A power user who excels at AI Usage can move into AI Literacy to understand the societal context of the tools they deploy professionally. A teacher who starts with AI Literacy and wants to go deeper can move into AI Usage to master the tools reshaping their classroom.

XP, levels, and tier progression apply across all frameworks. Your home screen XP bar is swipeable: AI Development | AI Usage | AI Literacy | All. Each framework tracks separately so you can see your depth in each dimension — and your combined total reflects the full breadth of what you know.

---

## Why Not Just One Curriculum?

Because the right curriculum depends entirely on your relationship to AI.

A nurse who needs to understand how AI is being used in diagnostic imaging does not need to understand gradient descent. Forcing her through 9 layers of engineering fundamentals to get to the content relevant to her would waste her time and bury the insights she needs.

An ML engineer who needs to pass technical screening interviews does not need to learn how to evaluate AI outputs from a non-technical perspective. His path is different.

A content creator who uses AI tools professionally every day is not a beginner, even if they have never trained a model. Their expertise is real — it just lives in a different domain than the engineer's.

One-size-fits-all AI learning fails all three of these people. Jogg's three frameworks serve all of them — with the depth they deserve.

---

*Explore the frameworks: use the curriculum section on the Jogg home screen to see all 27 layers and start wherever your learning begins.*

*Jogg — Your Professional AI/ML App.*

### The Case for a New Kind of Learning Platform for the Age of AI

*by the Jogg Team | MokingBird Oy*

---

The AI/ML learning app market is crowded. There are dozens of quiz apps, hundreds of online courses, thousands of YouTube channels, and an internet full of blog posts all claiming to teach you machine learning.

So why does Jogg exist? What makes it meaningfully different from everything else?

The short answer: because everything else was built for a different learner, a different depth, and a different purpose. Jogg was built specifically for **serious AI/ML learners who want professional-grade depth** — and no one else was building that.

Let us go through exactly what that means.

---

## The Problem with General Learning Apps

Most gamified learning apps — Duolingo, Quizlet, general knowledge quiz platforms — are brilliant at what they do. But what they do is teach **breadth at shallow depth, quickly and enjoyably**.

That is perfect for learning vocabulary in a new language or memorizing historical dates. It is the wrong model for learning AI/ML.

Here is why: AI/ML is not a collection of facts to be memorized. It is a **web of deeply interconnected concepts**, mathematical frameworks, algorithmic principles, and practical implementation patterns that build on each other in non-linear ways. Understanding how transformers work requires understanding attention mechanisms, which requires understanding dot products and softmax, which requires understanding why we need to scale attention scores by √d_k to prevent vanishing gradients.

You cannot shortcut that web. You have to build it, layer by layer.

General quiz apps do not know this and cannot accommodate it. Jogg was designed around it.

---

## The Problem with Video Courses

Massive Open Online Courses (MOOCs) — Coursera, edX, fast.ai, DeepLearning.AI — are genuinely valuable for many learners. But they have structural problems that limit their effectiveness for sustained, deep learning:

**Passive consumption:** Watching a lecture is easy. It feels like learning. But the research on learning retention is unambiguous: passive reading and watching produce far lower retention rates than active recall — answering questions, being tested, being forced to retrieve information rather than just recognize it.

**No personalization:** A MOOC course follows the same path for everyone. The 30 minutes you spend on a concept you already understand is 30 minutes you are not spending on the concept where you actually have a gap.

**No research paper integration:** MOOCs teach you to use tools. The best ones teach you the theory behind the tools. But almost none of them seriously engage with the primary literature — the actual research papers that contain the original ideas, the precise formulations, and the experimental context that defines the field.

**Completion problem:** MOOC completion rates are notoriously low — often below 15%. The passive format and lack of consequence for falling behind means most learners who start a course never finish it.

Jogg addresses all of these: active recall over passive consumption, adaptive personalization, deep integration with the research literature, and gamification that makes consistent engagement habitual.

---

## The Problem with Generic Quiz Apps in AI/ML

There are quiz apps specifically for programming, data science, and even machine learning. The best ones are honestly quite good — useful for interview prep, quick knowledge checks, and maintaining familiarity with concepts.

But they share a core limitation: **shallow question quality**.

Most AI/ML quiz apps generate questions that test surface recall:
- "What does LSTM stand for?"
- "Which of the following is NOT an activation function? (A) ReLU (B) Sigmoid (C) BatchNorm (D) Tanh)"
- "True or False: CNNs are used for image classification"

These are not useless questions. But they test recognition, not understanding. They do not require you to reason about *why* things work, *when* to use one approach over another, or *what would happen* in a specific scenario.

Jogg's questions are built to a different standard — we obsess over distractor quality, conceptual depth, and calibration to actual learning science. See our [question generation article](jogg_qustion_generation.md) for the full story.

---

## What Makes Jogg Different — The Pillars

### 1. The 3-Framework, 27-Layer System — Learning Paths for Every AI Learner

Most learning resources are built for one type of learner. A course for ML engineers. A YouTube series on prompting for power users. A news article for general audiences.

Jogg covers **all three dimensions of AI knowledge** — organized into three complete frameworks with 9 layers each, 27 layers total, 3,600+ questions:

**AI Development** — for engineers and researchers building AI systems:

| Layer | What It Covers |
|-------|---------------|
| **Foundations** | Linear algebra, statistics, Python, ML basics |
| **Data Acquisition** | APIs, scraping, SQL, streaming, file formats |
| **Data Preprocessing** | Cleaning, feature engineering, normalization, augmentation |
| **Model Training** | Neural networks, transformers, optimizers, fine-tuning |
| **Orchestration & RAG** | RAG systems, vector databases, agents, prompt engineering |
| **Inference & Optimization** | Quantization, pruning, distillation, KV cache, hardware |
| **Integration & Applications** | Deployment, APIs, monitoring, CI/CD, microservices |
| **Future Topics** | Multimodal AI, diffusion models, MoE, reinforcement learning |
| **AI Safety & Governance** | Ethics, bias, safety, responsible AI, governance |

**AI Usage** — for power users mastering AI tools:

| Layer | What It Covers |
|-------|---------------|
| **First Contact** | What AI tools are, major platforms, key capabilities |
| **Prompting Foundations** | Prompt structure, zero-shot, instructions, context |
| **Advanced Prompting** | Few-shot, chain-of-thought, system prompts, personas |
| **AI Tool Ecosystem** | ChatGPT, Claude, Gemini, Copilot, Midjourney, and more |
| **Workflow & Automation** | n8n, Zapier, Make, API integration, automated pipelines |
| **AI Agents** | Agents, multi-agent systems, tool use, memory |
| **Output Evaluation** | Fact-checking, hallucination detection, quality criteria |
| **Domain Applications** | AI for coding, writing, research, business, creative work |
| **Staying Current** | Model updates, new tools, evaluating new capabilities |

**AI Literacy** — for everyone who needs to understand AI:

| Layer | What It Covers |
|-------|---------------|
| **What Is AI?** | What AI is and isn't, history, how it works at a high level |
| **How AI Learns** | Training, data, patterns — no math required |
| **AI in Daily Life** | Tools you already use, what powers them |
| **Working with AI** | Using AI tools effectively, basic prompting |
| **AI & Society** | Jobs, economy, education, creativity, social impact |
| **AI Safety & Ethics** | Bias, fairness, misinformation, responsible use |
| **AI in Your Field** | Healthcare, education, business, law, media applications |
| **Evaluating AI** | Critical thinking about AI claims and outputs |
| **The Future of AI** | Trends, possibilities, what to watch for |

This is not three separate apps — it is one platform where engineers, power users, and curious everyday people all find a path built for them. All layers are accessible from day one; no mastery gates, no prerequisites required.

---

### 2. Research Paper Integration — Learning from First Principles

This is the feature that most clearly distinguishes Jogg from every other learning platform.

Jogg integrates **20 curated landmark research papers** directly into the learning experience through Papers Quest — and lets you request questions from any additional paper you specify.

The included papers are not incidental. They are the papers that *defined* modern AI/ML:
- *Attention Is All You Need* — every modern LLM descends from this paper
- *BERT*, *GPT-3*, *InstructGPT* — the lineage of conversational AI
- *LoRA* — the dominant approach to fine-tuning at scale
- *FlashAttention* — why modern context windows are possible
- *Constitutional AI* — the theory behind AI alignment
- *Scaling Laws* — why model size, data, and compute interact the way they do
- *Denoising Diffusion Probabilistic Models* — the foundation of image generation

These are not summaries. Jogg's Papers Quest generates questions that probe your actual understanding of each paper's methodology, contributions, and implications — the kind of questions that reveal whether you truly understand the work or just know it by name.

**Why does this matter?**

Because being able to discuss these papers fluently is the difference between someone who has learned AI/ML from the outside and someone who understands it from the inside. In interviews at top AI labs, in research discussions, in architectural decision-making at work — the researchers and engineers who can engage with the primary literature operate at a qualitatively different level.

Jogg builds that capability, systematically.

---

### 3. Adaptive Learning — Personalized to You

Jogg is not a fixed path. It is a system that learns what you know and what you do not know, and adjusts accordingly.

**Lane Mastery Tracking:** Your mastery percentage in each lane is computed from your actual performance — not just whether you answered questions, but how consistently and confidently you answer them. Questions you got wrong get weighted differently than questions you answered quickly and correctly.

**Mixed Practice Mode:** When you open Mixed Practice, Jogg's algorithm looks at your performance profile across all lanes and surfaces questions from the areas where you have the biggest gaps. It is not just "random questions from all lanes" — it is targeted gap-closing.

**Daily Jogg with FSRS:** The FSRS spaced repetition algorithm determines exactly which questions need review today based on your individual forgetting curve. This is scientifically validated to produce 30-50% better long-term retention than non-adaptive review schedules.

**Difficulty Calibration:** Every question's difficulty rating is continuously refined using Item Response Theory (IRT) — a psychometric framework that estimates true difficulty from actual response data. The difficulty labels you see are empirically validated, not just assigned by a human and never revisited.

The result: your Jogg experience is genuinely personalized. The questions you see, in the order you see them, at the difficulty levels presented to you, reflect your specific learning history and needs.

---

### 4. Professional Gamification — Motivating for Adults

Most gamified apps target children or casual learners. Their mechanics — streaks, badges, animated characters, bright colors — are designed for audiences where entertainment is the primary value.

Jogg's target audience is university students, professionals, and researchers. The gamification system was designed specifically for this audience:

**50-Level Metal Progression** — Four named tiers (Bronze, Steel, Platinum, Mythril) that represent meaningful knowledge milestones, not arbitrary points targets. Reaching Mythril genuinely means something about your AI/ML depth.

**XP Based on Real Merit** — XP reflects difficulty and speed, not just completion. Advanced questions earn 3.5× more XP than Beginner questions. Fast, fluent answers earn a time bonus of up to 1.5×. The system rewards genuine mastery, not just participation.

**Verifiable Certificates** — Lane completion certificates issued by MokingBird Oy that can be verified at jogg.app. Not gamification cosmetics — actual credentials.

**Clean, Professional Interface** — 5 carefully designed professional themes. No cartoon characters, no patronizing animations, no interface elements that feel like they belong on a children's educational platform.

The design signals clearly: this is serious software for serious learners.

---

### 5. Broad Audience Support — From Enthusiast to Researcher

AI/ML attracts an unusually diverse range of learners. Jogg is designed to serve all of them:

**The Enthusiast**
You love AI and want to understand what is actually happening inside GPT-4 and Stable Diffusion. You are not a professional ML engineer yet, but you want real depth, not just surface explanations. Jogg's Lane 0 (Foundations) meets you where you are and takes you as far as you want to go.

**The Student**
You are in a CS or data science program. You have courses, but they do not cover everything and they do not give you the practice repetition needed to actually internalize concepts. Jogg's structured curriculum maps to your coursework and goes deeper in the areas that matter for your career.

**The Interview Candidate**
You know that ML interviews at serious companies can cover anything from the math behind backpropagation to the system design of a recommendation pipeline to the content of recent NeurIPS papers. Jogg covers the full spectrum. Many users report that two months of consistent Jogg practice measurably improved their technical interview performance.

**The Working Professional**
You are already in the industry but the field moves fast. A new model architecture you are not familiar with. A deployment technique your team is adopting. A safety concept that is becoming important in your product decisions. Daily Jogg lets you stay sharp across the full AI/ML landscape in just 5-10 minutes a day.

**The Researcher**
You have deep expertise in your specific area but sometimes need to quickly refresh adjacent areas or formally test your understanding before teaching a concept. Papers Quest and the 27-layer structure let you navigate to exactly the depth you need on any topic.

**The AI Safety Specialist**
Lane 8 (AI Safety & Governance) is dedicated exclusively to the ethical, safety, and governance dimensions of AI — bias detection, prompt injection defense, alignment frameworks, responsible deployment, Constitutional AI. This is a topic that most learning platforms treat as an afterthought. Jogg treats it as a first-class curriculum component.

---

### 6. Your Interests Drive Your Learning Path

Jogg supports personalization by interests. When you set up your profile, you can specify your learning goals and focus areas:

- **Interview preparation** — algorithmic depth, system design, common interview topics
- **Academic study** — research paper depth, theoretical foundations
- **Professional upskilling** — practical deployment, optimization, production ML
- **Research** — cutting-edge topics, AI safety, research methodology
- **Enthusiast / General interest** — broad coverage, engaging progression

Based on your stated interests, Jogg adjusts question priorities, surfaces relevant papers, and tailors the difficulty progression to match your goals. Someone preparing for ML engineering interviews and someone studying for a research PhD are both well-served — but they are served differently.

---

### 7. Privacy and Data Ownership

Most learning platforms treat user data as an asset to be monetized. Jogg does not.

Your learning data — your progress, your quiz results, your weak spots, your improvement over time — belongs to you. We store it securely on your behalf. We use anonymized aggregates to improve question quality. We do not sell it, share it with advertisers, or use it for any purpose beyond operating the app for your benefit.

This is a design choice, not just a legal compliance posture. We believe that a learning platform which treats user data as an asset to be exploited is misaligned with the interests of its learners. Jogg is built to serve learners, not to profit from their data.

---

## The Honest Comparison

| Feature | Generic Quiz Apps | MOOCs/Courses | Jogg |
|---------|-------------------|---------------|------|
| AI/ML specificity | Partial | Variable | Full stack |
| Question depth | Surface recall | Lecture-following | Conceptual understanding |
| Research paper integration | None | Rarely | Core feature |
| Adaptive learning | Minimal | None | Full FSRS + IRT |
| Gamification for adults | No | No | Yes |
| Offline access | Rarely | Usually not | Yes |
| Privacy-first | Often not | Often not | Yes |
| Active recall | Yes | No | Yes |
| Interview prep depth | Partial | Partial | Comprehensive |
| Coverage breadth (full stack) | No | One area each | Yes |
| Paper request feature | None | None | Yes |
| Learning paths for all audiences | No | Partial | 3 complete frameworks |
| Non-technical AI literacy path | No | Occasionally | Yes — full 9-layer framework |
| Power user / AI tool mastery path | No | Rarely | Yes — full 9-layer framework |

---

## The Bottom Line

There is no shortage of resources for learning AI/ML superficially. Jogg exists for the learners who want more than superficial — the students who want to understand the research, the professionals who want to master the full stack, the enthusiasts who want to go beyond the hype to the real depth.

If you want to understand not just how to *use* AI tools but how they *work*, why they were designed the way they were, and where the field is going — Jogg was built for you.

If you want to be the person in the room who can discuss "Attention Is All You Need" fluently, who understands the tradeoffs in quantization strategies, who knows what RLHF actually involves at a technical level, who can speak meaningfully to AI safety and governance — Jogg will get you there.

That is what we mean when we say: **Your Professional AI/ML App**.

---

*Download Jogg on iOS or Android.*
*Questions? [email protected]*

*MokingBird Oy — Built for serious learners.*

### How Jogg's Gamified System Turns AI/ML Mastery Into a Journey You Actually Want to Take

*by the Jogg Team | MokingBird Oy*

---

Let's be honest about something uncomfortable: most people who start learning AI/ML do not finish.

They open a textbook, stare at the first chapter on linear algebra, and close it. They start an online course, watch the first five lectures with genuine enthusiasm, and abandon it by week three when life gets busy. They install PyTorch, run the first tutorial, and then… nothing. The momentum dies.

This is not a motivation problem. It is a *system* problem. Traditional learning formats were not designed to compete with everything else competing for your attention — and they were certainly not designed to make the long, slow grind of building genuine expertise feel satisfying along the way.

Jogg was. Here is how.

---

## The Psychology Behind Jogg's Design

The gamification in Jogg is not decoration. It is grounded in real research on motivation, habit formation, and learning retention.

### The Progress Principle
Harvard Business School researcher Teresa Amabile identified what she called the "progress principle": the single biggest driver of motivation in complex cognitive work is making visible progress toward a meaningful goal. When you can see yourself moving forward — however incrementally — you feel motivated to continue.

Jogg makes progress visible at every level:
- Per-question: you see immediate feedback on whether you were right and why
- Per-session: you see XP earned, questions answered, accuracy rate
- Per-lane: you see your mastery percentage grow toward 100%
- Per-day: you see your streak count and daily contribution
- Per-lifetime: you see your level, tier, and cumulative progression toward Mythril

There is always something to make progress on. There is always a number moving in the right direction.

### Variable Reward and Productive Challenge
Games are engaging partly because they use **variable reward** — you don't always know exactly what you'll get. In Jogg, you don't know in advance whether the next question will be straightforward or whether it will be a genuinely tricky expert-level question that makes you think harder than you expected. That variability keeps engagement high.

At the same time, Jogg's adaptive difficulty ensures that challenge stays in the "productive" zone — hard enough to be meaningful, not so hard that it becomes frustrating. Psychologist Mihaly Csikszentmihalyi called this the "flow state" — the optimal experience where skill and challenge are balanced.

---

## The Full Gamification System — Every Feature Explained

### The 50-Level Progression System

Jogg uses a 50-level mastery scale instead of the more common "beginner → intermediate → advanced" three-step classification. Why 50?

Because real learning is not three jumps. It is a continuous progression with dozens of meaningful milestones. When you go from level 11 to level 12, you have genuinely done more work than the previous level required, and you can feel it.

The 50 levels are organized into four tiers, each representing a meaningfully different stage of AI/ML expertise:

**Bronze (Levels 1–12): The Foundation Builder**
You are building your fundamentals. You know ML basics, you can work with data, you understand core concepts. This is where most new learners spend their first few months — and every level gained here represents real knowledge solidified.

What Bronze feels like: "I understand what a neural network is and why it works. I know the difference between supervised and unsupervised learning. I've answered questions about gradient descent that actually required me to think."

**Steel (Levels 13–24): The Practitioner**
You have a working knowledge of the full AI/ML pipeline — from data acquisition through preprocessing, model training, and deployment. You can have real technical conversations. You understand the tradeoffs.

What Steel feels like: "I understand why people use Adam over SGD in most cases, but also when SGD might be preferable. I know what attention is and how multi-head attention works. I can explain what quantization means and why it matters."

**Platinum (Levels 25–36): The Deep Practitioner**
You are now operating at a professional level. You understand research concepts, production challenges, and the nuances that separate people who have used AI tools from people who understand them.

What Platinum feels like: "I know what RAG is, how vector databases work, why KV caching matters for inference latency, and what the tradeoffs are in choosing between full fine-tuning and LoRA. I can read research papers and understand the key contributions."

**Mythril (Levels 37–50): The Master**
This is the domain of the genuinely expert. The questions at this tier test knowledge that most practitioners never encounter — edge cases, research-level nuances, the "why" behind architectural choices that most engineers take for granted.

What Mythril feels like: "I can discuss scaling laws and their implications for compute budgets. I understand constitutional AI and the theoretical basis for RLHF. I can explain why FlashAttention is IO-bound rather than compute-bound and what that means practically."

The level curve is progressive — each level requires more XP than the last, reflecting the reality that depth is harder to acquire than breadth.

---

### XP — The Fuel of Progress

Every correct answer earns you XP (experience points). Not just any amount — the XP you earn reflects the actual difficulty and conditions of your answer.

**Difficulty Base XP:**
| Level | XP |
|-------|----|
| Beginner | 10 XP |
| Intermediate | 20 XP |
| Advanced | 35 XP |
| Expert | 50 XP |

**Speed Bonus:**
Jogg rewards not just knowing the answer, but knowing it quickly — the kind of fluency that comes from genuine internalization rather than slow deliberation. If you answer faster than the expected time for a question of that difficulty, you earn a time bonus multiplier up to 1.5×.

This reflects a real principle from learning science: fluency matters. Experts don't just know the right answer — they access it quickly because the underlying concepts are deeply encoded. The speed bonus in Jogg rewards and incentivizes that fluency.

**Mode Context:**
Different learning modes have different purposes and intensities. Some modes carry additional multipliers to reflect their higher challenge or strategic value.

**The Result:**
Each question session produces a concrete XP number that honestly reflects how you performed — not just whether you got things right, but how confidently and quickly you demonstrated mastery.

---

### Streaks — The Power of Consistency

One of the most research-supported findings in habit science is this: **consistency beats intensity**. 30 minutes every day for a month produces better learning outcomes than 10 hours in a single weekend, even when the total time is equal.

Jogg's streak system is designed around this insight.

Every day you complete at least one quiz session — including Daily Jogg — your streak grows by one. Miss a day, and your streak resets. It sounds simple, and it is. But the psychology is powerful.

**What streaks do:**
- They create a commitment mechanism — you don't want to break a 14-day streak
- They make "just a few questions today" feel meaningful, because maintaining a streak is worth something
- They normalize daily learning as a habit, not an event
- They give you a visible measure of consistency to feel proud of

The streak flame icon in Jogg burns brighter as your streak grows. At high streaks, it becomes a real badge of commitment — not just to the app, but to your own learning goals.

---

### Daily Jogg — The Science of Remembering

Daily Jogg is not just a daily quiz. It is powered by **FSRS (Free Spaced Repetition Scheduler)**, one of the most advanced spaced repetition algorithms available.

Spaced repetition is one of the most well-established findings in learning science. The basic idea: you remember things better when you review them at increasing intervals, just before you are about to forget them. This is far more efficient than massed practice (cramming).

FSRS improves on older spaced repetition algorithms (like SM-2, used by Anki) by more accurately modeling individual forgetting curves. It adjusts review timing based on:
- How difficult the question is for you specifically
- How well you answered it in previous reviews
- How long it has been since your last review

The result: Daily Jogg presents you with exactly the questions you need to see today — not too early (wasteful), not too late (already forgotten). It typically takes 5–10 minutes per day, which is all the time you need when the algorithm ensures every minute is maximally productive.

**Daily Jogg XP is handled specially:** If you retry Daily Jogg on the same day, Jogg uses a delta model — it only adds the XP improvement over your previous attempt, preventing inflation while still rewarding genuine improvement.

---

### Lane Mastery and Certificates

Jogg's 9 learning lanes are not just organizational categories — they are genuine learning milestones with unlocking mechanics that reinforce the curriculum logic.

**Progressive Unlocking:**
Lane 0 (Foundations) is available to everyone immediately. Each subsequent lane requires **80% mastery** of the previous lane before it unlocks. This is not an arbitrary barrier — it reflects a real pedagogical principle: some knowledge genuinely depends on other knowledge.

You cannot meaningfully learn about transformer training (Lane 3) if you don't understand what a neural network is (Lane 0) or how data flows into a model (Lanes 1–2). The unlock system enforces productive sequencing.

**Mastery Percentage:**
Your mastery percentage in each lane is computed from your performance on that lane's questions — not just whether you answered them, but how consistently and confidently you answer them. A question you got wrong three times and then finally got right contributes differently to your mastery score than a question you nailed immediately.

**Certificates:**
Complete a lane and earn a **verifiable certificate** from MokingBird Oy / Sortify. Your certificate shows:
- Your name (as set in your profile)
- The lane completed
- The date of completion
- A unique verification code

Certificates are verifiable at jogg.app, shareable, and represent a genuine milestone in your learning journey.

---

### Papers Quest — Depth as a Game Mechanic

Papers Quest is Jogg's most unique learning mode, and it works differently from standard quiz modes.

Instead of questions organized by lane or topic, Papers Quest presents you with questions organized by research paper. Each paper has its own set of questions, and you can see your completion and accuracy for each paper individually.

**What makes Papers Quest feel like a game:**
- Each paper is a self-contained challenge — you can "complete" a paper and see your score
- Papers have difficulty ratings (Beginner → Expert), creating a natural progression
- The citation network between papers creates interesting chains — mastering "Attention Is All You Need" unlocks richer understanding of BERT, which enriches GPT-3, which enriches InstructGPT
- Paper completion badges track your progress through the literature

**What makes it educational:**
Unlike trivia games, Papers Quest questions are not about surface facts. They test genuine conceptual understanding — the kind of knowledge that stays with you because it connects to a web of related ideas.

---

### RUSH Mode — When Speed Is the Lesson

RUSH Mode is Jogg's speed challenge — questions under time pressure, with higher XP rewards for fast accurate answers.

RUSH Mode serves a real educational function beyond entertainment. It trains:
- **Automaticity** — the ability to recall correct answers without conscious deliberation
- **Confidence calibration** — knowing what you actually know versus what you only think you know
- **Exam performance** — time pressure is a real feature of technical interviews and exams

If you find that your RUSH Mode performance lags significantly behind your normal quiz performance, it tells you something important: you have understanding, but not yet fluency. You know things when you have time to think about them, but you have not yet internalized them deeply enough to access them quickly.

That gap is real and worth working on. RUSH Mode makes it visible.

---

### Arcade Mode — Challenge and Competition

Arcade mode brings a competitive edge to Jogg's learning experience. It is structured as an intensive challenge run — a set of questions where you need to hit a pass mark to succeed.

In Arcade mode:
- The stakes are higher — not reaching the pass mark means no XP reward for the run
- The questions may come from any lane and any difficulty
- Completion of a successful Arcade run earns a bonus reward

Arcade mode is for learners who have built confidence in their knowledge and want to test themselves under pressure — and for those competitive moments when you want to push your performance to the limit.

---

### Custom Quizzes and Sharing

Jogg lets you create your own quiz from a selected set of questions, then share it with anyone using a 6-character code or a QR code.

**Why this matters for learning:**
Creating a quiz is itself a learning activity. Deciding which questions represent the most important concepts in a topic — and organizing them into a coherent set — forces active engagement with the material. It is the "teaching it back" principle made interactive.

**Why this matters for community:**
Shared quizzes create a social learning dynamic. Study groups can create shared quiz sets. Professors can create curated question sets for their students. Friends can challenge each other. The 6-character code makes sharing effortless — just share the code and anyone with Jogg can join.

---

## Putting It All Together: A Day in the Life with Jogg

Here is what a productive day with Jogg actually looks like for a working professional learning AI/ML in parallel with their job:

**Morning (5 minutes):**
Open Jogg. Daily Jogg has 7 questions ready based on your spaced repetition schedule. You answer them — 5 correct, 2 wrong. You get 85 XP. Streak maintained. Day started right.

**Lunch break (10 minutes):**
You are working through Lane 3 (Model Training). You answer 12 questions on transformer architectures. You get 10 of 12 right. Lane mastery goes from 67% to 71%. 180 XP earned. One level-up notification.

**Evening (20 minutes):**
You have been meaning to properly understand LoRA for a project. You open Papers Quest and work through the LoRA paper questions. 15 questions, 12 correct. You understand low-rank adaptation at a deeper level than you did this morning.

**End of day:**
110 XP from daily, 180 from lane practice, 145 from Papers Quest = 435 XP today. Streak is now 8 days. Lane 3 mastery is 71%. You are on track to unlock Lane 4 within the week.

That is a genuinely productive day of AI/ML learning — and it fit inside the gaps of a full work day.

---

## Why This Approach Beats Traditional Learning

| Traditional Learning | Jogg |
|---------------------|------|
| No feedback until the end of a chapter/course | Immediate per-question feedback |
| Unclear how far you are from "done" | Visible mastery percentage at all times |
| Easy to skip days without consequence | Streak system makes consistency visible |
| All topics weighted equally | Adaptive difficulty focuses on your weak spots |
| Passive reading → poor retention | Active recall → significantly better retention |
| No connection to primary literature | Papers Quest and Paper Request features |
| Progress resets if you switch platforms | Your data belongs to you and stays with you |

---

## The Philosophy: Gamification in Service of Mastery

Jogg's gamification is not about keeping you hooked for its own sake. There are no dark patterns, no artificial urgency, no manipulative mechanics designed to extract more time from you than your learning warrants.

Every gamification element in Jogg exists because it supports genuine learning:
- Streaks exist because consistency is how real learning happens
- XP exists because visible progress sustains motivation
- Levels and tiers exist because milestones give achievement meaning
- Certificates exist because completion deserves recognition
- Lane unlocks exist because knowledge genuinely builds on knowledge
- Papers Quest exists because the research literature is where the real understanding lives

The goal is not to maximize time in the app. The goal is to maximize genuine understanding of AI/ML — and to make the process of getting there engaging enough that you actually stick with it long enough for it to matter.

That is what great gamified learning looks like.

---

*Ready to start your AI/ML journey? Download Jogg on iOS or Android.*

*Jogg — Your Professional AI/ML App.*

### Understanding the System That Measures Your AI/ML Mastery

*by the Jogg Team | MokingBird Oy*

---

One of the most satisfying moments in Jogg is seeing that "+35 XP" appear after you nail a tough Advanced question. But where does that number come from? How is it calculated? And why does it matter?

This article breaks down exactly how Jogg's XP system works — so you understand what you are optimizing for, why different questions earn different amounts, and how your daily effort translates into the level progression that marks your growing mastery of AI/ML.

---

## Why XP Matters (Beyond the Number)

XP is not just a game mechanic. It is a proxy for learning effort that tries to reflect something real: **not all correct answers represent the same amount of learning**.

Getting a Beginner question right about what "supervised learning" means is useful — but it reflects less effort and depth than correctly answering an Advanced question about the mathematical relationship between scaling laws, compute budgets, and optimal dataset size.

If both questions gave the same XP, the system would fail to distinguish between surface familiarity and genuine depth. Jogg's XP system is designed to make that distinction meaningful.

---

## The Core XP Formula

Every correct answer in Jogg earns XP according to this formula:

```
XP = round(baseXP × timeBonus × multiplier)
```

Let's break each component down.

---

## Component 1: Base XP by Difficulty

Every question in Jogg has a difficulty level, calibrated based on expert judgment and continuously refined through data from actual user performance. The base XP for a correct answer reflects that difficulty:

| Difficulty | Base XP | What It Typically Tests |
|------------|---------|-------------------------|
| **Beginner** | 10 XP | Core terminology, basic concepts, foundational understanding |
| **Intermediate** | 20 XP | How components interact, standard implementation patterns |
| **Advanced** | 35 XP | Tradeoffs, limitations, deeper principles, cross-concept reasoning |

**Important:** You earn XP only for **correct** answers. Incorrect answers earn 0 XP. There is no penalty XP for wrong answers — only the opportunity cost of not getting the points you could have.

This is intentional. Jogg is about building mastery, not punishing uncertainty. If you got something wrong, the more important outcome is seeing the correct answer and explanation — not being penalized.

---

## Component 2: The Time Bonus

Jogg rewards not just knowing the right answer, but knowing it with **genuine fluency** — the kind of confident, quick recall that comes from deep internalization rather than slow working-through.

When you answer faster than the expected time for a question of that difficulty, you earn a time bonus multiplier:

```
timeBonus = clamp(1.0, 1.5, 1.0 + ((expectedTime - timeTaken) / expectedTime) × 0.5)
```

In plain language:
- If you take the full expected time or longer: time bonus = **1.0×** (no bonus, but no penalty)
- If you answer in half the expected time: time bonus approaches **1.5×**
- The bonus is capped at **1.5×** — speed beyond a certain point doesn't keep multiplying

**Example:**
An Advanced question has an expected time of 60 seconds. If you answer correctly in 20 seconds:
- Time ratio: (60 - 20) / 60 = 0.667
- Time bonus: 1.0 + (0.667 × 0.5) = 1.333
- XP: round(35 × 1.333) = **47 XP** (vs. 35 XP at full time)

**The philosophy behind the time bonus:**
Learning science makes a distinction between *knowing* something and *knowing it fluently*. Fluency — the ability to retrieve correct answers quickly without effortful deliberation — is a sign that knowledge has moved from conscious recall to automatic retrieval. This reflects genuine internalization, not just exposure.

The time bonus incentivizes and rewards that fluency. It is not about rushing — it is about measuring the confidence that comes with deep mastery.

---

## Component 3: Mode Multipliers

Different learning modes in Jogg have different purposes, intensities, and strategic values. The multiplier component reflects this.

| Mode | Multiplier | Rationale |
|------|-----------|-----------|
| **Organized Lanes** | 1.0× | Structured learning — baseline XP |
| **Mixed Practice** | 1.15× | Cross-topic reasoning is harder |
| **RUSH Mode** | 1.35× | High pressure, speed required |
| **Papers Quest** | 1.30× | Research-level depth |
| **Daily Jogg** | 1.0× | Consistency rewarded separately through streaks |
| **Arcade** | Varies | Pass-based bonus model |

**Note on current implementation:** The mode multiplier system is implemented and configured. Some multipliers are actively being rolled out across all modes as part of our XP unification work. What is always active: base XP by difficulty + time bonus.

---

## Daily Jogg XP — The Delta Model

Daily Jogg uses a special XP model that is different from other modes. Understanding it helps you get the most out of your daily sessions.

**The challenge with daily XP:** If every retry of Daily Jogg awarded full XP, you could artificially inflate your daily XP by retrying the session multiple times. The same questions, answered again and again, would multiply XP beyond what the actual learning warranted.

**Jogg's solution — the delta model:**

1. When you complete Daily Jogg, your session XP is computed normally (base XP × time bonus for each correct answer, plus a daily completion bonus if you get 4 out of 5 or more correct)
2. Jogg checks whether you have already completed Daily Jogg today
3. If yes: only the **improvement** over your previous attempt is added to your total XP
4. If no: the full session XP is added

**Example:**
- First Daily Jogg attempt today: 95 XP earned → +95 XP to your total
- You retry (wanted to improve): 120 XP earned → only +25 XP added (the improvement)
- If your retry earns less: +0 XP (retrying worse does not reduce your XP)

**Daily completion bonus:** If you answer 4 out of 5 or more questions correctly in Daily Jogg, you earn an additional **+2 XP bonus** on top of your per-question XP. Small, but meaningful — it rewards accuracy, not just participation.

---

## Level Progression — What XP Buys You

XP accumulates over time and determines your current level. Jogg uses a **progressive XP curve** — each level requires more XP than the last.

The XP required to advance to the next level follows this formula:
```
XP required for next level = round(100 × currentLevel × 1.5)
```

This means:
| Level | XP to advance to next level |
|-------|-----------------------------|
| 1 | 150 XP |
| 5 | 750 XP |
| 10 | 1,500 XP |
| 20 | 3,000 XP |
| 30 | 4,500 XP |
| 40 | 6,000 XP |
| 49 | 7,350 XP |

Early levels are achievable quickly — you can see visible progress within your first few sessions. Later levels represent substantial commitment and genuine depth.

**Why a progressive curve?**
Because real expertise does not accumulate linearly. Going from knowing nothing about AI to knowing the basics takes less time and effort than going from solid practitioner knowledge to genuine expert-level mastery. The XP curve reflects this reality — it is harder to advance at higher levels, which is how it should be.

---

## Tier Bands — The Milestones That Matter

50 levels are organized into four named tiers:

| Tier | Levels | What It Signifies |
|------|--------|------------------|
| **Bronze** | 1–12 | Building foundations |
| **Steel** | 13–24 | Solid practitioner knowledge |
| **Platinum** | 25–36 | Deep, professional-level mastery |
| **Mythril** | 37–50 | Elite AI/ML expertise |

Advancing to a new tier is one of the most meaningful milestones in Jogg. Each tier boundary represents a genuine qualitative jump in your AI/ML depth — not just a number, but a reflection of the actual knowledge you have built.

When you reach Steel, you have passed through the foundational material and are operating with real practitioner knowledge. When you reach Platinum, you are at the level where you can discuss research papers fluently and handle complex technical discussions. Mythril is for the genuinely elite — the learners who have worked through the full breadth and depth of the curriculum.

---

## A Practical Guide to Maximizing Your XP

Understanding the formula gives you a practical strategy for efficient progression:

### Focus on accuracy, not speed
The time bonus matters, but accuracy matters more. An incorrect answer earns 0 XP regardless of how fast you answered. Slow down on questions you are unsure about — getting it right is worth more than being fast.

### Push into higher difficulty
Beginner questions earn 10 XP; Advanced questions earn 35 XP — with potentially up to 52 XP including the time bonus. If you find Beginner questions trivial, progress to harder questions as quickly as you can. The XP rewards reflect the actual difficulty.

### Use RUSH Mode strategically
RUSH Mode has the highest multiplier of the standard modes (1.35×) and rewards speed. Once you have solidified knowledge in a topic through Organized Lanes, switch to RUSH Mode to reinforce fluency and earn higher XP in less time.

### Never miss a Daily Jogg
Daily Jogg is XP-efficient because FSRS ensures every question is worth answering at that moment. It is also your streak mechanism — and long streaks compound your sense of momentum and commitment.

### Complete Papers Quest
Papers Quest questions tend toward the Advanced end of the difficulty spectrum, and the Papers multiplier (1.30×) applies. If you are ready for paper-level depth, Papers Quest is one of the most XP-efficient modes.

### Track your XP per framework
Your home screen XP bar is **swipeable across four views**: AI Development | AI Usage | AI Literacy | All. Each framework tracks its own XP total separately. The "All" view shows your combined XP across all three frameworks. Use the per-framework views to see where your depth is concentrated and where you have room to grow.

---

## The Transparency Principle

We want you to be able to predict your XP from first principles. Every element of the formula is public and explained here.

What you can always count on:
- **Correct answers earn XP by difficulty.** Harder questions = more XP.
- **Faster correct answers earn more XP**, up to a 1.5× bonus.
- **Daily Jogg uses the delta model** — retrying on the same day only adds your improvement.
- **Wrong answers earn 0 XP.** No penalties, just missed opportunity.

As we continue to refine and unify the XP system — including fully rolling out mode multipliers across all modes — we will update this documentation to reflect exactly what is active.

---

## What XP Represents About Your Learning

A final thought: XP in Jogg is not an arbitrary score. Every XP point represents a correct answer to a question calibrated to a specific difficulty level, answered within a specific time. Your total XP, your level, your tier — these numbers have real meaning because the questions behind them have real meaning.

When you reach Platinum in Jogg, you have answered thousands of questions across the full AI/ML curriculum, at every difficulty level, demonstrating consistent recall and genuine understanding. That is not a trivial achievement. It reflects actual depth.

That is what we are trying to measure, reward, and help you build. One correct answer at a time.

---

*For technical questions about the XP system, or to report anomalies in your XP display, contact [email protected].*

*Jogg — Your Professional AI/ML App.*

### How Friendly Competition Accelerates Your AI/ML Learning

*by the Jogg Team | MokingBird Oy*

---

Learning AI/ML in isolation is productive. Learning in a community where you can measure your progress against others — and feel the motivating pull of healthy competition — is something else entirely.

The Jogg Leaderboard brings that dimension to your learning journey. It is where the most dedicated AI/ML learners in the Jogg community come to measure their commitment, compare their depth, and compete for the top spots.

Here is everything you need to know about how the leaderboard works, what it measures, and how to use it to accelerate your own progress.

---

## What Is the Jogg Leaderboard?

The Jogg Leaderboard is a **voluntary, opt-in ranking system** that shows the top AI/ML learners on the platform sorted by XP, level, and other achievement metrics.

When you join the leaderboard, your username and learning stats become visible to other Jogg users. You can see where you stand relative to the community, track your position over time, and set goals around climbing the rankings.

**Key word: voluntary.** The leaderboard is entirely opt-in. You are never added to it without your explicit consent.

---

## Privacy First — The Opt-In Model

We take privacy seriously at Jogg, and we designed the leaderboard with that principle in mind.

**Before your name ever appears on any public ranking, you must actively choose to join the leaderboard.** This is not buried in settings or hidden in a terms checkbox — it is an explicit opt-in prompt with clear explanation of what information will be visible.

**What is visible on the leaderboard when you opt in:**
- Your **username** (the one you chose during registration)
- Your **current level and tier** (e.g., Level 24 · Steel)
- Your **total XP**
- Optionally, your **current streak** (you can choose whether to display this)

**What is NEVER visible on the leaderboard:**
- Your real name (unless you chose that as your username)
- Your email address
- Any authentication details
- Your country or location
- Your device type or any other personal information

**Withdrawing from the leaderboard:**
You can leave the leaderboard at any time from your profile settings. When you withdraw, your entry is removed immediately — you will not appear in any ranking from that moment forward. No data is retained in the public leaderboard.

This approach reflects Jogg's broader philosophy: your learning data is yours, not ours. Sharing it publicly is always your choice.

---

## What the Leaderboard Measures

The leaderboard ranks users primarily by **total XP**. This is intentional — total XP is Jogg's most comprehensive measure of learning effort and depth combined.

Because XP reflects both the difficulty of questions answered and the speed of answers, total XP is a genuine proxy for sustained, deep engagement with the AI/ML curriculum. Someone at the top of the leaderboard is not just someone who clicked through a lot of easy questions — they have answered hard questions, answered them with fluency, and done it consistently over time.

**Leaderboard rankings show:**

| Metric | What It Tells You |
|--------|------------------|
| **Rank** | Your position among all leaderboard participants |
| **Level & Tier** | Your mastery milestone (Bronze → Mythril) |
| **Total XP** | Your cumulative learning score |
| **Streak** | (Optional) Your current daily streak |

**Planned additions:**
- Weekly XP leaderboard (most active this week)
- Lane-specific leaderboards (top learners in each of the 9 lanes)
- Paper Quest completion rankings
- Tier-restricted leaderboards (compete only within your tier)

---

## Who's at the Top? Understanding the Competition

If you are new to the leaderboard, it helps to understand what the top positions actually represent.

The highest-ranked Jogg users have:
- Worked through multiple lanes systematically and reached high mastery percentages
- Answered both Beginner and Expert-level questions across the full AI/ML curriculum
- Maintained consistent daily practice over extended periods
- Engaged with Papers Quest, which offers higher XP density for learners who are ready for it

Reaching the top of the leaderboard is not a matter of finding a shortcut. There is no hack — just sustained, quality engagement with genuinely hard AI/ML content.

That said, you do not need to be at the top to benefit from the leaderboard. Seeing where you stand in context — knowing there are learners above you who have gone deeper and others below you who are earlier in their journey — gives your progress real-world meaning that solo metrics cannot provide.

---

## Strategies for Climbing the Leaderboard

If you are competitive by nature, here is how to advance your ranking efficiently:

### 1. Prioritize Advanced and Expert Questions
XP scales significantly with difficulty. A single Expert question answered quickly is worth 75 XP — the equivalent of 7+ Beginner questions. As soon as you have solid foundations in a lane, push into the harder difficulty questions.

### 2. Build Your Fluency with RUSH Mode
RUSH Mode carries a 1.35× XP multiplier. If you have already learned a topic through Organized Lanes, drilling it in RUSH Mode earns more XP per session while also building the speed that unlocks the time bonus.

### 3. Never Break Your Streak
Consistent daily engagement compounds over time. A learner who does 20 minutes every day will outpace a learner who does 2 hours on weekends, both in retention and in cumulative XP. Protect your streak.

### 4. Go Deep on Papers Quest
Papers Quest questions skew toward Advanced and Expert difficulty, with a 1.30× mode multiplier. Learners who engage seriously with the research literature earn XP significantly faster than those who stick to standard questions.

### 5. Complete Lanes for Certificates and Milestones
Lane completion is not just ceremonially satisfying — it is also the gateway to the next lane, and the next lane's harder questions and higher XP potential. Efficient lane progression is the fastest path to high cumulative XP.

### 6. Use Mixed Practice for Gap-Filling
Mixed Practice adapts to your weaknesses. It surfaces the questions where you are performing below your potential. Closing those gaps improves both your accuracy rate and your XP efficiency.

---

## The Community Dimension — More Than Competition

The leaderboard is about more than individual competition. It is also a signal of community vitality.

When you see a busy, competitive leaderboard — dozens of learners clustered within a few thousand XP of each other — you are seeing a community of people who are all working seriously on the same challenging material. That community creates a form of collective accountability and motivation that is hard to replicate in solo learning.

**Leaderboard social features (coming):**
- Follow individual learners and track their progress
- Challenge a specific user to a head-to-head XP race over a week
- Study group leaderboards — create a private leaderboard just for your team, class, or friend group
- Lane challenge boards — see who in your network is furthest along in each of the 9 lanes

---

## Leaderboard Integrity — Fair Competition

The Jogg leaderboard is meaningful only if the rankings reflect genuine learning effort. We take leaderboard integrity seriously.

**What we monitor:**
- Abnormally high XP accumulation rates that suggest automated use
- Answer patterns inconsistent with genuine quiz behavior
- Account sharing or credential selling

**Consequences:**
Users found to be artificially inflating their scores — through bots, scripts, or other means — will be removed from the leaderboard and may have their accounts suspended.

The goal is a fair competition where the top of the leaderboard represents the most genuinely knowledgeable AI/ML learners in the Jogg community. That value is worth protecting.

---

## Leaderboard and Your Privacy — Detailed Breakdown

Since we know privacy matters to many of our users, here is the complete breakdown of what happens to your data when you join the leaderboard:

**When you opt in:**
- Your username is added to the public leaderboard index
- Your XP, level, and tier become publicly queryable
- Your streak may be visible if you choose to display it

**When you opt out:**
- Your entry is removed from the leaderboard index immediately
- No leaderboard-specific data is retained
- Your personal account data (XP, progress, etc.) is not affected by opting out — it remains in your private account as normal

**What changes and what doesn't:**
- Opting out of the leaderboard does NOT delete your account data
- Your XP and progress continue to accumulate normally — you just don't appear in public rankings
- You can rejoin the leaderboard at any time with a fresh opt-in

**The leaderboard stores:**
Only the data you have explicitly consented to share: username, XP, level, tier, and optionally streak. Nothing else.

For the full privacy policy including leaderboard data handling, see our [Privacy Policy](jogg_privacy_policy.md).

---

## Getting Started on the Leaderboard

Ready to join? Here is how:

1. Open Jogg and navigate to your **Profile** tab
2. Tap **Leaderboard** in the profile menu
3. Read the opt-in information explaining what will be visible
4. Tap **Join Leaderboard** — your entry appears immediately
5. View the leaderboard to see your starting position
6. Set a goal: pick someone a few ranks above you and chase them

That's it. You are now part of the Jogg competitive learning community.

---

## A Note on the Point of Competition

We want to be clear about what the leaderboard is and what it is not.

It is a tool to motivate consistent, quality learning through the social dynamics of friendly competition. It works because seeing your rank and seeing others' progress creates a productive external reference point for your own effort.

It is **not** a statement about who is a better person, who is smarter, or who will be more successful in AI/ML. The leaderboard measures one thing: sustained engagement with the Jogg learning system over time. It does not measure raw intelligence, professional seniority, years of experience, or any other dimension of your overall AI/ML ability.

A researcher with 20 years of experience who joined Jogg last week will be at the bottom of the leaderboard. That researcher knows incomparably more about AI than the Level 30 learner above them on the leaderboard. The leaderboard is a measure of Jogg engagement, not of intelligence or expertise in the world at large.

Use it for what it is: a motivating tool. A way to make your daily learning feel connected to a community. A friendly challenge.

And let's see how high you can climb.

---

*To join or leave the leaderboard, visit your Profile tab in the Jogg app.*

*Jogg — Your Professional AI/ML App.*

### How Jogg's Research Paper Request Feature Brings the Literature to Life

*by the Jogg Team | MokingBird Oy*

---

There is a gap in how most people learn AI and machine learning.

Courses teach you to use the tools. Blog posts summarize the highlights. Tutorials walk you through the code. But somewhere along the way, the actual research papers — the documents that contain the original ideas, the hard-won experimental results, the precise formulations that define the field — get left behind.

This is a problem. Because if you want to truly understand why transformers work the way they do, why certain training tricks matter, or why one optimization algorithm outperforms another in specific contexts, you need to go back to the source. The papers.

Jogg was built with this belief at its core. And our **Paper Request** feature is one of the most direct expressions of it.

---

## What Is the Paper Request Feature?

The Paper Request feature lets you do something simple but powerful: **ask Jogg to generate a curated quiz set from a specific research paper**.

Instead of waiting for us to add a paper to our standard library, you can submit a paper of your choosing — by arXiv ID, title, or DOI — and Jogg will generate a set of high-quality quiz questions derived directly from that paper's content, methodology, and contributions.

This means your Jogg learning experience can be tailored to exactly the papers that matter most to your specific goals:
- The paper behind the model architecture you are implementing at work
- The foundational work referenced in your PhD thesis
- The latest transformer variant from NeurIPS that everyone on your team is discussing
- A classic paper you have always meant to read properly but never got around to

---

## Why Research Papers?

Let us be honest about something: research papers are hard to read. They are dense, notation-heavy, often assume familiarity with dozens of prior works, and sometimes bury their most important contribution in a technical appendix between two figures.

But the difficulty is worth it. Here is why.

### Papers Contain the Real Reasoning

A blog post about attention mechanisms will tell you that transformers use self-attention. The paper *Attention Is All You Need* tells you *why* the authors made every specific architectural decision — the positional encoding choices, the multi-head design, the specific dimension sizes, and what they tried that did not work. That reasoning is what separates deep understanding from surface familiarity.

### Papers Are the Ground Truth

When you are in a technical interview or a research discussion and someone asks "but why does multi-head attention project into lower-dimensional subspaces?", the answer that comes from actually having read the paper is qualitatively different from the answer synthesized from a summary.

### Papers Build Research Literacy

The ability to read, evaluate, and critically discuss research papers is one of the most valuable skills in AI/ML — and one of the least taught. Using Jogg's Papers Quest and Paper Request features trains this skill actively, not passively.

---

## The Curated Paper Library — What We Already Have

Before you request a custom paper, it is worth knowing what is already in our curated library. Jogg ships with **20 landmark papers** spanning the full AI/ML stack, covering decades of foundational and cutting-edge research:

### Foundation and Architecture Papers
**Attention Is All You Need** (Vaswani et al., 2017)
The paper that introduced the Transformer architecture and changed NLP forever. Every modern LLM descends from this work. Jogg includes questions covering self-attention, multi-head attention, positional encoding, and the encoder-decoder design.

**Deep Residual Learning for Image Recognition** (He et al., 2015)
ResNet introduced skip connections that enabled training of 152-layer networks. Jogg's questions explore why residual connections work, their mathematical interpretation, and their influence on subsequent architectures.

**BERT: Pre-training of Deep Bidirectional Transformers** (Devlin et al., 2018)
Bidirectional pre-training with masked language modeling. Questions cover the difference between BERT and GPT-style models, MLM vs CLM pre-training, and fine-tuning approaches.

**Language Models are Few-Shot Learners** / GPT-3 (Brown et al., 2020)
The paper that demonstrated in-context learning at scale. Questions probe few-shot prompting, the relationship between model scale and capability, and what in-context learning actually is.

**ImageNet Classification with Deep CNNs** / AlexNet (Krizhevsky et al., 2012)
The paper that started the deep learning revolution. Fundamental questions about ReLU, dropout regularization, data augmentation, and GPU-accelerated training.

### Efficiency and Optimization Papers
**LoRA: Low-Rank Adaptation of Large Language Models** (Hu et al., 2021)
The dominant approach to parameter-efficient fine-tuning. Questions cover the mathematical intuition behind low-rank decomposition, when to use LoRA versus full fine-tuning, and rank selection.

**FlashAttention** (Dao et al., 2022)
IO-aware exact attention that achieves 2-4x speedup. Questions cover the memory bottleneck in standard attention, tiling strategies, and why IO-bound rather than compute-bound operations are the real constraint.

**Adam: A Method for Stochastic Optimization** (Kingma & Ba, 2014)
The optimizer behind virtually every modern neural network training run. Questions cover adaptive learning rates, momentum, bias correction, and when Adam underperforms SGD.

**Batch Normalization** (Ioffe & Szegedy, 2015)
Internal covariate shift and how normalization addresses it. Questions cover the mechanics, why it enables higher learning rates, and the relationship between batch size and BN behavior.

### Generative Models
**Denoising Diffusion Probabilistic Models** (Ho et al., 2020)
The foundation of modern image generation (DALL-E 2, Stable Diffusion, Midjourney). Questions cover the forward/reverse diffusion process, the denoising objective, and the relationship to score matching.

**Generative Adversarial Networks** (Goodfellow et al., 2014)
The original GAN paper. Questions cover the minimax game, mode collapse, training instability, and the theoretical guarantees.

### Retrieval and Knowledge
**RAG: Retrieval-Augmented Generation** (Lewis et al., 2020)
The foundational paper for RAG systems. Questions cover dense passage retrieval, the combination of parametric and non-parametric memory, and practical implications for reducing hallucinations.

**Word2Vec / Efficient Estimation of Word Representations** (Mikolov et al., 2013)
CBOW and Skip-gram word embedding models. Questions cover the intuition behind semantic relationships in embedding space and the training objectives.

### Scaling and Alignment
**Scaling Laws for Neural Language Models** (Kaplan et al., 2020)
Empirical relationships between model size, data size, compute, and performance. Questions cover power law relationships, compute-optimal training (Chinchilla implications), and practical guidance for resource allocation.

**Training language models to follow instructions** / InstructGPT (Ouyang et al., 2022)
RLHF methodology and the foundation for ChatGPT. Questions cover the three-stage RLHF process, reward model training, and PPO in the context of language model alignment.

**Constitutional AI** (Bai et al., 2022)
AI feedback (RLAIF) and the use of constitutional principles for harmlessness. Questions cover scalable oversight, reducing dependence on human labelers, and the distinction between supervised and RL phases.

### Multimodal and Other Architectures
**CLIP** (Radford et al., 2021)
Vision-language alignment through contrastive learning on 400M image-text pairs. Questions cover the contrastive objective, zero-shot classification, and the implications for multimodal AI.

**MoE: Outrageously Large Neural Networks** (Shazeer et al., 2017)
Sparse mixture-of-experts layers. Questions cover gating mechanisms, sparse activation, expert routing, and the computational efficiency argument for MoE.

**PPO: Proximal Policy Optimization** (Schulman et al., 2017)
The policy gradient algorithm behind RLHF. Questions cover the clipped surrogate objective, the importance of the trust region constraint, and how PPO fits into the broader RLHF pipeline.

---

## How Paper Requests Work

When you submit a paper request, here is what happens behind the scenes:

### 1. Paper Ingestion
The paper is fetched (from arXiv or uploaded directly) and processed through our content extraction pipeline. This extracts the full text, mathematical notation, figures, tables, and reference structure.

### 2. Content Processing
Our system processes the paper in multiple passes:
- Identifying key contributions and main claims
- Extracting experimental results and their context
- Mapping the paper to our 9-lane topic taxonomy
- Identifying prerequisite knowledge needed to understand the paper

### 3. Question Generation
MokingbirdDataGen generates questions across multiple question types:
- **Conceptual questions** — testing understanding of the core ideas
- **Technical questions** — testing understanding of methodology and implementation
- **Comparative questions** — relating this paper's contribution to prior or subsequent work
- **Application questions** — testing ability to apply the paper's insights to new scenarios

Questions are generated at multiple difficulty levels, from accessible entry points for readers new to the topic to expert-level questions about subtle experimental choices and limitations.

### 4. Quality Review
Generated questions are run through our quality pipeline — filtering for accuracy, clarity, and educational value. Questions that do not meet our bar are rejected or revised before they are added to your request set.

### 5. Your Question Set
The resulting question set is added to your Papers Quest library, tagged with the paper's metadata, and becomes available immediately. You can quiz yourself on it, include it in custom quizzes, or share it with study partners.

---

## Who Uses Paper Requests?

### PhD Students and Researchers
You need to truly master the papers in your literature review — not just cite them, but deeply understand their methods, assumptions, and limitations. Paper Request lets you build exam-quality quiz sets for every paper you need to internalize.

**Scenario:** You are writing a thesis chapter on efficient transformers. You submit FlashAttention, FlashAttention-2, Ring Attention, and Longformer as paper requests and build a custom study set covering the full landscape of efficient attention mechanisms.

### ML Engineers Staying Current
You want to understand the papers behind the tools you use every day — not just how to use them, but why they were designed the way they were.

**Scenario:** Your team is evaluating whether to use LoRA or QLoRA for fine-tuning a production model. You request both papers, quiz yourself on the technical tradeoffs, and come to the decision-making meeting with actual depth.

### Interview Candidates
Technical interviews at top AI companies frequently go deep into research papers. Companies like Google DeepMind, Meta AI, and OpenAI expect candidates to be conversant with recent research.

**Scenario:** You are interviewing for a research engineer role. Your recruiter mentioned the team works heavily on RLHF alignment. You request InstructGPT, Constitutional AI, and Direct Preference Optimization (DPO) and build a study set covering the alignment literature.

### Students in AI Courses
Many university AI/ML courses assign specific papers as required reading. Paper Request turns passive reading into active learning.

**Scenario:** Your NLP course assigns five papers for the week including BERT, GPT-2, and T5. Instead of just reading them, you quiz yourself after each one and confirm your understanding before the lecture.

---

## The Broader Vision: AI/ML Knowledge from First Principles

The research paper request feature reflects a philosophy that is central to everything Jogg does: **AI/ML should be learned from first principles, not just from tutorials**.

The papers are the first principles. They contain the actual ideas, the actual experiments, the actual tradeoffs that define how modern AI systems work. Everything else — the blog posts, the library documentation, the tutorials — is a derivative of the papers.

When you can discuss the content of landmark papers fluently, when you understand not just what RLHF is but why PPO was chosen as the RL algorithm and what the reward model training involves, you are operating at a qualitatively different level than someone who learned AI from YouTube.

That is the level Jogg is built to help you reach.

---

*Submit a paper request from the Papers Quest section of the app.*
*Questions? Contact us at [email protected].*

*Jogg — Your Professional AI/ML App.*

### The Science and Craft Behind Every Question You Answer

*by the Jogg Team | MokingBird Oy*

---

Every quiz app has questions. But not every quiz app's questions are worth your time.

If you have ever opened a "machine learning quiz" on some random platform and been asked *"what does ML stand for?"* or given a multiple choice question where three of the four answers are so obviously wrong they barely count as distractors — you know the problem. Shallow questions teach you to recognize the shape of correct answers, not to actually understand the material.

At Jogg, we think differently. Here is the full story of how we create, validate, and continuously improve the questions you encounter on your AI/ML learning journey.

---

## Starting Point: What Makes a Good AI/ML Question?

Before we talk about systems and pipelines, let us talk about principles. A good question in the context of AI/ML learning should do at least one of the following:

1. **Test genuine understanding**, not just memorization of facts
2. **Have a single, clearly defensible correct answer** — no ambiguity, no "well it depends on the framework" traps
3. **Have plausible distractors** — wrong answers should reflect real misconceptions, not obvious nonsense
4. **Be appropriately difficult for its level** — not too easy to feel trivial, not so obscure that it stops being educational
5. **Connect to something real** — a real algorithm, a real paper, a real architectural choice that matters in practice

These five criteria drive everything about how questions get created and reviewed at Jogg.

---

## Phase 1: The MVP Foundation — Expert-Curated Questions

The first questions in Jogg were not generated by any AI system. They were **hand-crafted by human experts** with deep knowledge of AI/ML.

Our curation team worked through the full 9-lane curriculum — from mathematical foundations through data preprocessing, model training, RAG systems, inference optimization, deployment, multimodal AI, and AI safety — writing questions that meet our quality bar.

This was deliberate. Before we could train machine learning models to generate good questions, we needed a **gold standard dataset** of what "good" actually looks like.

### What Expert Curation Looks Like

When a domain expert writes a question for Jogg, they go through a structured process:

**Step 1: Pick a key concept**
Not just a fact, but a concept that has genuine depth. For example: not "what is a transformer?" but "why does the self-attention mechanism in transformers scale quadratically with sequence length, and what architectural innovations address this?"

**Step 2: Write the correct answer**
The answer should be provably correct, not a matter of opinion. It should reference the actual technical reason, not a simplified rule of thumb.

**Step 3: Write three distractors**
Each distractor should:
- Sound plausible to someone with partial knowledge
- Represent a real misconception in the field
- Not be trivially eliminable by process of elimination

A good distractor is almost as hard to write as the correct answer.

**Step 4: Tag difficulty level**
Every question gets classified:
- **Beginner:** Requires understanding of core terminology and basic concepts
- **Intermediate:** Requires understanding of how components interact
- **Advanced:** Requires understanding of tradeoffs, limitations, and deeper principles
- **Expert:** Requires understanding of research-level nuances, edge cases, and cross-cutting concerns

**Step 5: Peer review**
Every question is reviewed by at least one other domain expert before it enters the question bank.

This process is slow. It is expensive. But it produces questions that actually teach you something.

---

## Phase 2: Research Paper-Based Questions — A Different Challenge

One of Jogg's most distinctive features is our **Papers Quest** mode — quiz questions derived directly from landmark AI/ML research papers.

Writing questions about research papers is harder than writing general curriculum questions, for a few reasons:

- Papers often contain subtle technical claims that are easy to misrepresent
- The "correct" interpretation of a result can evolve as the field's understanding matures
- The most important contribution of a paper is sometimes not its headline result, but a methodological choice or negative finding buried in an appendix

Our approach to paper-based questions:

**Read the full paper, not just the abstract.** Our question writers read every paper in its entirety before writing a single question.

**Focus on key contributions over incidental details.** A question about FlashAttention should test your understanding of *why IO-awareness matters for attention computation* — not whether you memorized a specific benchmark number from Table 3.

**Test for conceptual transfer.** The best questions aren't "what did this paper propose?" but "given this paper's contribution, what would you expect to happen if you applied it to this scenario?"

Our current curated papers include some of the most important works in modern AI/ML:

- *Attention Is All You Need* — the original Transformer
- *BERT* — bidirectional pre-training for language understanding
- *GPT-3* — few-shot learning at massive scale
- *LoRA* — parameter-efficient fine-tuning
- *RAG* — retrieval-augmented generation
- *FlashAttention* — IO-aware fast attention
- *Scaling Laws for Neural Language Models*
- *Constitutional AI* — harmlessness from AI feedback (RLHF/RLAIF)
- *Denoising Diffusion Probabilistic Models*
- *CLIP* — vision-language models through contrastive learning
- ...and a total of 20 landmark papers

---

## Phase 3: MokingbirdDataGen — AI-Assisted Question Generation

As Jogg grows, manual question curation cannot scale to meet the depth and breadth the platform needs. This is where our proprietary **MokingbirdDataGen** system comes in.

MokingbirdDataGen is a custom, two-model AI content generation system developed by MokingBird Oy. It is designed specifically for generating high-quality, educationally rigorous quiz content.

### The Architecture

MokingbirdDataGen is built on two fine-tuned language models working in tandem:

#### The Generator Model
The Generator LLM (based on Mistral-7B, fine-tuned with LoRA adapters) is responsible for actually creating questions. Given a source document, topic taxonomy tag, and difficulty target, it generates:
- The question stem
- The correct answer
- Three plausible distractors
- A detailed explanation

The Generator does not generate questions blindly. It has been trained on our expert-curated question bank, learning the patterns and quality attributes of questions we consider excellent.

#### The Classifier Model
The Classifier LLM (also Mistral-7B + LoRA) works in parallel with the Generator. Its job is to:
- Label questions with rich metadata (difficulty, topic, subtopic, cognitive level)
- Flag questions that may be ambiguous, incorrect, or poorly constructed
- Assign quality scores based on criteria derived from our expert curation standards

#### Reinforcement Learning (GPRO)
Both models are trained using a custom reinforcement learning approach we call **Mokingbird-GPRO-Hybrid** — a novel combination of field-level process supervision and outcome-level reward. This ensures both models optimize for the right goals:

- Generator reward: question quality, factual correctness, appropriate difficulty calibration, distractor plausibility
- Classifier reward: accuracy of metadata labels compared to expert-labeled gold data

The two models train together in a feedback loop, continuously improving question quality.

### Why Fine-Tuned Models Instead of Prompting GPT-4?

This is a fair question. Using a general-purpose LLM API would be faster to set up. We chose to build custom fine-tuned models because:

1. **Quality consistency:** A fine-tuned model trained on our exact quality standards produces more consistently good questions than prompt engineering a general-purpose model
2. **Domain depth:** A model fine-tuned on AI/ML content deeply understands the domain-specific terminology, paper citations, and technical nuances that general models often get wrong
3. **Control:** We control the full pipeline — we can improve quality, fix systematic errors, and retrain without dependency on external API changes
4. **Data privacy:** All generation happens on our infrastructure — research paper content is not sent to external APIs

---

## Phase 4: Difficulty Calibration — Making "Hard" Actually Mean Hard

Writing a question that is labeled "advanced" is one thing. Ensuring that it actually discriminates between intermediate and advanced learners is another.

Jogg uses **Item Response Theory (IRT)** to calibrate question difficulty dynamically after launch.

IRT is a psychometric framework used in standardized testing (it's behind exams like the SAT and GRE). The idea is that a question's "true" difficulty can be estimated statistically based on how many people answer it correctly, accounting for the ability level of the people who answered it.

Here is how it works in Jogg:

1. A new question enters the system with an **estimated difficulty** based on expert judgment
2. Over time, as more users answer the question, we collect anonymous response data
3. Nightly, a batch process fits an IRT model (2-parameter logistic) to the response data
4. The difficulty estimate is updated to reflect observed performance
5. Questions that are performing outside their difficulty band (e.g., an "advanced" question that 90% of beginners answer correctly) get flagged for review

This means the difficulty labels in Jogg are **empirically validated**, not just subjectively assigned.

---

## Phase 5: Adaptive Difficulty — Questions That Learn From You

Beyond calibrating questions, Jogg adapts which questions you see based on your personal performance profile.

The **Mixed Practice** mode uses your question history to determine:
- Which topics have gaps in your understanding
- Which concepts you've mastered and need less reinforcement
- Which difficulty level is appropriate for your current level in each topic area

This is combined with **FSRS spaced repetition** for Daily Jogg, which determines the optimal timing for reviewing each concept based on your forgetting curve.

The result is that no two users' Jogg experiences are exactly alike — the system continuously adjusts to give each individual learner the questions that will help them grow most efficiently.

---

## Continuous Quality Improvement

Question quality is not a one-time concern. We have ongoing processes to ensure quality stays high:

### User Reporting
Every question has an in-app reporting mechanism. If you believe a question contains an error, is ambiguous, or has become outdated, you can flag it. Our content team reviews all flags.

### Analytics-Driven Review
Questions with anomalous response patterns are automatically flagged for review:
- Questions with very high or very low correctness rates relative to their labeled difficulty
- Questions where users spend unusually long or short times deliberating
- Questions with unusual skip or flag rates

### Regular Content Audits
The AI/ML field moves fast. What was an emerging topic two years ago may now be foundational — or obsolete. We conduct regular curriculum audits to ensure content stays current with the state of the field.

---

## What This Means for You as a Learner

When you answer a question in Jogg, you can trust:

- The correct answer is actually correct, not just "probably right"
- The difficulty label reflects real calibration data, not just gut feeling
- The wrong answers represent real misconceptions, not random filler
- The question tests something worth knowing
- If the field has evolved and a question has become outdated, it will be updated

This is what it means to take question quality seriously. It is a lot more work than scraping the internet for quiz questions or asking a chatbot to generate 500 questions in five minutes. But it is the only approach that produces learning that actually sticks.

---

## The Future: Personalized Question Generation

Looking ahead, our ambition is to enable fully personalized question generation — where you can specify your background, goals, and areas of interest, and Jogg generates questions tailored specifically to your learning path.

Already, you can request questions derived from specific research papers. In the future, Jogg will be able to generate questions calibrated to your specific skill level, focused on the specific topics most relevant to your goals — whether that's preparing for a specific type of interview, mastering a particular model architecture, or staying current with a specific research area.

This is the promise of AI-assisted personalized learning at scale. And it is what MokingbirdDataGen is being built to deliver.

---

*Have feedback on question quality? Use the in-app flag feature or reach out to [email protected].*

*Jogg — Built for serious learners, with questions that prove it.*