542 AI engineering lessons on GitHub: build algorithms from scratch before libraries
techNmak · x · 2026-10-15
Someone published 542 AI engineering lessons on GitHub as AI Engineering from Scratch, covering 20 phases from math/deep learning through LLM engineering, inference optimization, RAG, MCP, AI agents, and production systems.
Its signature approach: implement the underlying algorithm yourself first, then use the production library:
- Backpropagation from the math
- A BPE tokenizer from scratch
- Self-attention and transformers
- An AI agent loop with tool calling
Each lesson follows a consistent structure: MOTTO → PROBLEM → CONCEPT → BUILD IT → USE IT → SHIP IT.
Notable projects include building a retrieval evaluator that catches ranking regressions, an agent trace debugger in TypeScript, and a tool-call firewall in Rust.
More from coding & agent
- AWS Kiro boxes its AI coding agent in an OS-enforced sandbox on your laptop — shashib · 2026-10-15
- RuleTrek benchmark: best of 11 frontier LLMs scores only ~62% on agentic rule reasoning — Chenhao Yuan · 2026-10-15
- Claude Opus 5.5 voluntarily writes a synthetic handwriting generator for OCR pipeline testing — lee_stott · 2026-10-15
- Using MCP without any LLM: deterministic cloud cost estimation pipeline — whispered_word12 · 2026-10-15
- DeepSeek V4.1 Flash is now free in OpenCode — CurieuxExplorer · 2026-10-15
- memory_profiler author returns after 8 years, plans to revive the project with AI agents — fpedregosa · 2026-10-15