Project Watch

Learn by studying real AI projects

Curated, annotated deep dives into notable AI projects. Understand the engineering decisions, trace the source code, and connect everything back to what you’re learning in the cohort.

Project Watch is where we study real AI projects in depth. Each entry is a full annotated walkthrough — what the project does, why the engineering decisions matter, and how it connects to what you’re learning in the Roadmap.

Projects

How monkey patching makes LLM inference 2x faster by replacing HuggingFace internals at runtime. Fused GPU kernels, practical optimization, and why it matters for every model you will ever serve.

How the community turned a toy training loop into an agentic ResearchOps platform. Agents, search policy, verification, and the future of AI R&D.