Skip to content

Harness removal 1: delete the dead runners and the standing autopilot that seats minds into rounds - #5250

Merged
joelteply merged 3 commits into
canaryfrom
fable/harness-removal-1
Oct 21, 2026
Merged

joelteply merged 3 commits into
canaryfrom
fable/harness-removal-1

Conversation

@joelteply

Copy link
Copy Markdown
Contributor

What was deleted (git diff --shortstat origin/canary: 17 files, +13 / -1752)

  • cognition/competitor.rs (897 lines): rival-harness (Hermes subprocess) runner. rg "competitor|resolution_bench" core --type rust found only its pub mod line; no caller.
  • cognition/resolution_bench.rs (320): standalone ActionCommand registered via register_stateless_command!(ResolutionBench); registration goes with the file. No other reference anywhere (rg over the repo).
  • cognition/resolution_compute.rs (418) and gym_grader::CodeVerifier + its test (~70): their only consumer was resolution_bench. The source-hygiene reachability ratchet flagged them as newly unreached (unwired public machinery rose to 57), so they go too, plus the stale identity/mod.rs LooseId row. Kept: resolution.rs and will.rs (still reached by their own mod/tests; hygiene does not flag them). Follow-up candidate, not done here.
  • Stale docs: CLAUDE.md/AGENTS.md no longer tell agents to run benchmark/standing; verdict_board comment.

Targets 2 and 3 were already done on canary

  • swe_verdict_sweep.rs is already 90 lines: only the refusal store (record_refusal, standing_env_refusal, refusal_is_env_fault, read_refusal), all with live callers. sweep/sweep_if_due do not exist (rg sweep_if_due = 0 hits outside the file's own doc).
  • BenchmarkStandingModule/benchmark_standing/benchmark/standing: no code in core (rg finds only the two doc files fixed here; MULTIMODAL doc already says retired 2026-10-19).
  • Health line: citizen_health/health_window.rs already reads "no benchmark round is owed an idle roster (no harness, Joel 2026-10-19)". There is no "NO ROUND ... owed: a round" string on canary. Before = after (no change).

Baselines lowered (cause: this PR's deletions, never raised)

unwired 56->54, underived bounds 164->162, head cuts 187->186, unjustified unwraps 2367->2359, spawn sites 132->131, upward imports 672->671. Exact values come from cargo test -p source-hygiene output on this branch (before the serving node was asked to stop local builds).

Verification

  • Local cargo check -p continuum-core --tests --features metal,accelerate after deleting competitor + resolution_bench: 0 errors, no dead-code warning attributable to them. The later resolution_compute/CodeVerifier deletion and baseline edits were NOT compiled locally (serving node swapping; CI is the compiler).
  • Local cargo test -p source-hygiene after the last code deletion: only the six "fell" baseline messages above remained (no rises); baselines then updated, not re-run.
  • Core lib tests, Windows compile, ratchets, ts-rs drift: see CI.

🤖 Generated with Claude Code

@joelteply

Copy link
Copy Markdown
Contributor Author

PASS #5250 at a2f6986, Cormac. Claim checked: each deleted item had no remaining caller.

  • git grep -E 'cognition::competitor|competitor::|resolution_bench|resolution_compute|CodeVerifier|ResolutionBench|resolution/bench' over the whole head (all file types) finds only the source-hygiene baseline comments. No caller, no command string, no client reference.
  • gym_grader loses CodeVerifier and its test, whose only consumer was resolution_bench. resolution.rs and will.rs stay.
  • AGENTS.md / CLAUDE.md now say the standing autopilot is deleted (No harness: delete the standing benchmark autopilot and the "owed: a round" idle line (card a00ab1a6) #5199), not 'stays off'.
    CI green. Source review only.

joelteply and others added 3 commits October 21, 2026 12:15
…rival-harness runner and standalone bench command

rg shows no references outside their own mod declarations; resolution_bench was a registered stateless command (benchmark harness outside the room).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
…anding) from CLAUDE.md, AGENTS.md, verdict_board comment

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
…lution_bench); lower six hygiene baselines to exact new counts

Hygiene reported the newly unreached machinery (ComputeDepthLadder, FacultyDraftBackend, CodeVerifier); deleting them takes BASELINE_UNWIRED 56->54. Other baselines fell with the deleted files: upward imports 672->671, spawn sites 132->131, unjustified unwraps 2367->2359, head cuts 187->186, underived bounds 164->162.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
@joelteply
joelteply force-pushed the fable/harness-removal-1 branch from a2f6986 to 4436a22 Compare October 21, 2026 17:15
@joelteply

Copy link
Copy Markdown
Contributor Author

PASS #5250 at 4436a22 (delta from my PASS at a2f6986), Cormac, 2026-10-21T17:29:52Z. Claim checked: the rebase changes no logic. How: compared the PR's own diff at a2f6986 and 4436a22, each against its merge-base with canary. The only changed lines are deleted lines that now read continuum_base::probe! / continuum_base::time::EpochMs instead of crate::; that's #5249's move. Same 15 files.

@joelteply
joelteply merged commit 8fa3811 into canary Oct 21, 2026
9 checks passed
@joelteply
joelteply deleted the fable/harness-removal-1 branch October 21, 2026 17:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant