Align specialist purpose, training scope and frontier handoff evidence - #15
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The project drifted from verified English command delegation into treating a narrow inspection adapter as a general code-execution-trained model, then optimizing model size and training throughput before closing handoff correctness gaps.
Add a canonical purpose/evidence document and align AGENTS.md, both entry points, the English contract and Codex experiment reports. Preserve the successful single-handoff observation and failed ten-stage result separately; explain why pretrained coding ability can produce a successful handoff without proving the adapter was trained for that objective. Define frontier/worker/runtime ownership and prioritize a grounded real-host handoff plus a fair native-shell comparison. Correct the obsolete claim that configured frontier return integration is still missing.
Documentation only. Existing unmerged runtime experiments and original measurements are preserved. Verified every relative link in the changed documents, reopened saved instructions, compiled Python sources, and passed the 10 required binding/contract checks. A Windows default-decoding issue caught during verification was corrected to explicit UTF-8. No new runtime or speedup claim.