AI Learning Curriculum

Goal: Develop frontier AI engineering skills at Anthropic's level

This is not a casual reading list. It's a rigorous 6-month program designed to build the skills needed to work on frontier AI systems, with emphasis on safety, alignment, and practical engineering.

Philosophy

"You do not rise to the level of your goals. You fall to the level of your systems." — James Clear, Atomic Habits

This curriculum is built around systems, not just goals:

• Daily habits over cramming
• Implementation projects over passive reading
• Portfolio artifacts over certificates
• Spaced repetition over one-time exposure

Quick Start

# Install the CLI tracker
cd tracker && pip install -e .

# See what to read next
ai-learn next

# Log a completed reading
ai-learn log "Attention Is All You Need"

# Check your streak
ai-learn streak

# View overall progress
ai-learn progress

Structure

curriculum/           # Learning modules across 5 phases
projects/             # 6 portfolio projects (implementation required)
quiz-prep/            # Technical quiz preparation
tracker/              # CLI habit tracker with XP and achievements

The 5 Phases

Phase	Focus	Modules
1	Foundations	ML fundamentals, Transformers (implement!), Scaling laws
2	Practical Skills	Finetuning (LoRA/DPO), Evals & Benchmarks, Inference
3	Alignment & Safety	RLHF, Constitutional AI, Interpretability
4	Applications	Prompting, RAG, Agents & Tool use, Code Generation
5	Mastery	System design, Capstone project

See ROADMAP.md for the detailed 6-month path.

Why Anthropic as a Benchmark?

This curriculum uses Anthropic's engineering standards as a benchmark because:

1. They value "direct evidence of ability" — Portfolio projects > credentials
2. Safety-first thinking — Their core research areas define the frontier
3. Implementation matters — Build, not just read
4. Constitutional AI is central — Deep understanding of alignment approaches

Success Metrics

By completion, you should be able to:

• Explain transformer architecture at whiteboard level
• Implement attention mechanism from memory
• Finetune a model with LoRA and evaluate it properly
• Discuss Constitutional AI vs RLHF tradeoffs fluently
• Build and evaluate a coding agent on SWE-Bench
• Have 3+ portfolio projects on GitHub
• Articulate your perspective on AI safety clearly
• Complete ML system design problems confidently

Identity Statement

"I am someone who learns AI every day."

This is your new identity. The habit tracker, the streak counter, the XP system—they all reinforce this identity with every small action.

Getting Started

1. Read the ROADMAP.md to understand the 6-month journey
2. Start with curriculum/01-ml-fundamentals or skip to where you're comfortable
3. Use ai-learn next to always know what's next
4. Never break the chain

Resources

• Anthropic Research
• Alignment Science Blog
• Latent Space 2025 Reading List
• Constitutional AI Paper

---

This curriculum uses Anthropic's engineering standards as a benchmark. It emphasizes the skills they value: deep technical understanding, safety-first thinking, and demonstrated ability through projects.