G-08Researcher + Implementer

/coding-agents/researcher-implementer

One agent reads the codebase and produces a digest; a second implements against it with a clean context.

Step-by-step walkthrough4 steps
  1. Exploring a huge repo fills a context window with mostly-irrelevant reading.
  2. The researcher’s whole job is compression: which files, which conventions, which traps.
  3. The implementer starts fresh with only the 2k tokens that matter.
  4. The digest is a contract: anything the researcher missed, the implementer can’t know.
Key insight

Splitting read-heavy from write-heavy work means the implementer never pays the exploration tax — its whole window is for the change.

Failure mode

A digest that omits one convention or trap: the implementer can’t know what the researcher didn’t write down.

Further reading

Elsewhere in the atlas