Collection
Make coding agents work in inspectable increments
Turn current long-horizon coding-agent practice into concrete feature ledgers, cold-start checks, evaluator criteria and safe parallel work that a human can steer without micromanaging every line.
8 elements · Explore in any order.
2 heuristics · 1 principle · 5 protocols
8 elements
Give a long AI task a durable feature ledger
A long prompt says what you wanted once. A ledger says what is still true now.
Read the idea ProtocolStart a resumed agent session with a workspace sanity check
Before the next shift starts building, make sure the factory still turns on.
Read the idea PrincipleAsk a coding agent for one coherent increment at a time
Autonomy works better when the unit of progress is small enough to verify.
Read the idea ProtocolRe-test old harness rules when the model changes
Yesterday's workaround can become tomorrow's tax.
Read the idea ProtocolTurn subjective quality into evaluator criteria before looping on it
An evaluator cannot enforce taste that nobody has made legible.
Read the idea HeuristicAdd planner-builder-evaluator roles only after the simple loop hits a ceiling
Three agents are not automatically one agent with better judgment.
Read the idea HeuristicParallelize agents only across separable work
Sixteen agents cannot parallelize one locked door.
Read the idea ProtocolGive parallel agents independent tests and merge boundaries
Parallel workers need an oracle and a border, not just separate chat windows.
Read the idea