docs(model-config): Setup A precise semantics — driver-swap escalation + advisor subagent fallback (#440 follow-up) - #446
Merged
Conversation
…n + advisor subagent fallback (#440 follow-up) Maintainer ask 2026-07-13: Setup A's semantics kept getting re-confused (effort ladder vs model escalation; what happens when advisor() fails). Clarity restatement of already-certified decisions — no new claims. - README: new 'Reading Setup A precisely' section — three rules, each with its why: effort escalation stays inside the Sonnet 5 driver (medium default per #440), model escalation SWAPS the driver to Opus 4.8 xhigh (2 failures / LOW confidence / high-stakes), advisor failure falls back to a Fable subagent, never to skipping - AI_SETUP_LANES: Setup A escalation row + 'Two escalation axes' and 'Advisor failure has a fallback' paragraphs; outage procedure Step 2 rewritten — swap the transport, not the check (Codex round-1 P1: it still offered a 'no advisor' path, contradicting the new rule) - tests: new test_setup_a_escalation_and_advisor_fallback_explicit (positive + section-scoped negative half, RED-first, mutation- verified); #439 reviewer-model pins re-shifted +4 after insertion - ROADMAP: parking lot — rename-research row added (wizard->harness, workshop later, expiry 2026-09-11); /insights row closed as fired Cross-model review: Codex xhigh (GPT-5.6 Sol), 3 rounds to CERTIFIED — 5 findings total (skip-path contradiction, outcome-guarantee and superlative overclaims, false-green test gap), all fixed. Doc-consistency suite: 72/72.
BaseInfinity
added a commit
that referenced
this pull request
Jul 15, 2026
…net 5 medium default (#448) Catch-up release: seven consumer-affecting commits were stranded on main since v1.86.0 (July 5) because nothing prompts the manual tag-to-publish step — most urgently #445, the Copilot CLI argument-hint fix from the repo's first external contributor (@thejesh23), which npm consumers were not receiving. - CHANGELOG [1.87.0]: leads with the first-external-contribution credit; covers #441 (GPT-5.6 Sol reviewer sweep), #443 (Sonnet 5 medium default, unbacked 5x quota claim removed, hook floor), #445/#447 (Copilot fix + regression test), #446 (Setup A semantics), #440/#442 (chores) - Version bumps at all sites found by the broad grep (v1.85.0 lesson) - ROADMAP #448 (Copilot CLI compat watch) + #449 (release-drift gate — this incident's post-mortem; scope: own-package npm-vs-main watcher) - skills/update: [1.87.0] example line; two oldest bullets trimmed to stay under the 20K session-load cap Cross-model release review: Codex xhigh (GPT-5.6 Sol), 3 rounds to CERTIFIED — caught a sed-falsified historical version citation and two unsupported duration claims in my own incident narrative. All suites green.
This was referenced Jul 15, 2026
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.
Why
Maintainer ask (2026-07-13): Setup A's semantics kept getting re-confused — "Sonnet 5 high or xhigh?", "escalation means what exactly?", "advisor failed, now what?". This PR makes the answers impossible to miss, each with its why. Pure clarity restatement of #440/#439-certified decisions — introduces no new factual claims.
What
mediumdefault); model escalation swaps the driver to Opus 4.8 xhigh (2 failures / LOW confidence / high-stakes); advisor failure → spawn a Fable subagent, never skip.test_setup_a_escalation_and_advisor_fallback_explicit: positive assertions + section-scoped negative half so the skip-path can't silently return (RED-first, mutation-verified). CC version drift: baseline v2.1.195, latest v2.1.207 (12-release gap) #439 reviewer-model line pins re-shifted +4, each verified against real content.Review trail
Codex xhigh (GPT-5.6 Sol), 3 rounds to CERTIFIED — 5 real findings across rounds, all fixed via the dialogue loop. Doc-consistency suite: 72/72.