rogee/superpowers
1
0
Fork 0
mirror of https://github.com/obra/superpowers.git synced 2026-06-13 14:19:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

Spec: L2b tested — opus structural win, sonnet transmission+attention gap (E35/E36); bump evals to 9919b27

Browse Source
This commit is contained in:
Jesse Vincent
2026-06-11 14:06:28 -07:00
committed by Jesse Vincent
parent 8d354bb36b
commit a2a4190809
2 changed files with 24 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
docs/superpowers/specs/2026-06-10-strict-cost-sdd-design.md
View File

@@ -133,8 +133,29 @@ opus controller flagged it 5/5. Cheap controllers handle explicit
escalation; they absorb implicit authority-vs-quality adjudication. escalation; they absorb implicit authority-vs-quality adjudication.
A possible L2b (discrete rule: "a reviewer finding that conflicts with A possible L2b (discrete rule: "a reviewer finding that conflicts with
the plan's text is the human's decision — escalate it") would route the the plan's text is the human's decision — escalate it") would route the
failing judgment through the escalation behavior that held; untested. failing judgment through the escalation behavior that held.
Original recon notes follow.
**L2b tested 2026-06-11 (E35/E36, evals
`docs/experiments/2026-06-11-build-loop-autoresearch.md`): improves the
opus stack, does NOT rescue the sonnet rung.** Two rules: a reviewer
tripwire (a plan-mandated defect IS a finding — Important, labeled
plan-mandated; the human decides) and a controller escalation rule
(plan-mandated findings go to the human like any plan contradiction).
Micro on frozen sonnet-composed inputs: 0/6 → 6/6 labeled findings.
Full battery: opus controllers 2/2 internalized the rule, caught their
reviewer's miss as self-described backstop, and escalated for a
sanctioned fix (the 4241 ad-hoc behavior made structural); escalation
sanity 2/2 unbroken. Sonnet controllers: 1/5 full pass — paraphrase
drops the tripwire from dispatches (2/5 transmitted), transmission
alone doesn't fire it live (read-once dilution across the reviewer's
tool reads; placement within the dispatch refuted as the variable),
and no sonnet controller showed backstop behavior; 1/5 shipped the
defect. The L2b rules are a candidate commit for the opus stack.
A future L2c for the sonnet rung would pair the SKILL.md
constraints-recipe (the one channel sonnet transmits verbatim) with a
mandatory output-format slot for plan-mandated findings (the skeleton
survives every observed paraphrase and is consulted at composition
time); untested. Original recon notes follow.
**Recon (superseded):** **Recon (superseded):**
Sonnet-controller runs (claude-sonnet coding-agent): all gates green at Sonnet-controller runs (claude-sonnet coding-agent): all gates green at

2
evals

Submodule evals updated: af05326467...9919b27766