CodeDocs vault
Section 4 of 6

Tours (6)

Curated reading orders, 30 to 90 minutes each. Every stop — concept, project, or insight — is annotated with a one-line "why this stop," so the path makes sense as you walk it.

Come here when you want to be told where to go next.

Start with Build a minimal agent loop in 90 minutes — the foundation everything else builds on

v1 → v2 evolution — what gets refactored when

30 min · Engineers planning a re-architecture of a working agent. 5 stops

Tool-format taxonomy — tool_use vs XML vs OpenAI fn

45 min · Engineers picking a tool-calling format for a new agent. 6 stops

Memory compression strategies — a deep dive

60 min · Engineers building long-running agents who need state to outlive a context window. 7 stops

Architecting an AI-Act-aware compliance agent

75 min · Engineers building agents that plug into CI/CD for an EU AI Act compliance / GRC platform. Audit, scope, oversight are all first-class. 9 stops

Designing a multi-agent review pipeline (Swisscheese)

75 min · Engineers building a meta-harness on top of coding agents — multiple reviewer agents on a single PR, with cross-agent verdict synthesis. 8 stops

Build a minimal agent loop in 90 minutes

90 min · Engineers who want to internalize the irreducible structure before reaching for a framework. 7 stops