Sources
July 20, 2026 · View on GitHub
Primary stack (this author)
- Loop Engineering — design systems that prompt agents over time
- Fleet Engineering — govern populations of agents
- Harness Foundry — composable harness runtime
- Outerloop — evidence → verdict → answerability
Related public practice
- Context engineering discussions (what the model sees this turn)
- Agent memory systems (Mem0, Letta/MemGPT-style tiering) — we stay store-agnostic; markdown first
- Maker/checker and harness engineering patterns from the broader agent coding community
How this repo relates
Turn “agents need memory” from a product claim into copyable patterns + init/audit CLIs, the same way loop-engineering did for loops.