feat: complete Phases 2-5 (tool descriptions, prompt sections, code evolution, continuous loop) - #162

Open

MaxFreedomPollard wants to merge 27 commits into

NousResearch:mainfrom

MaxFreedomPollard:feat/phases-2-5-complete

Commits on Jul 31, 2026

core: span-exact artifact IO and the shared gate ladder
MaxFreedomPollard
committed
890ace9
feat: complete Phases 2-5 (tool descriptions, prompts, code, continuous loop)
MaxFreedomPollard
committed
2cf6d45
core: paired statistics with power reporting
MaxFreedomPollard
committed
23bc5e6
stats: paired tests and power reporting across every phase gate
MaxFreedomPollard
committed
3f4c029
stats: fix the defects an adversarial audit found in the audit itself
MaxFreedomPollard
committed
047309a
stats: drop unmeasured pairs, correct the multi-section disjunction, pin judges to temperature 0
MaxFreedomPollard
committed
b55dc47
core: run cost accounting and the deployment-via-PR step PLAN.md requires
MaxFreedomPollard
committed
ee6d034
phases: emit a reviewable pull request and report what the run cost
MaxFreedomPollard
committed
937f933
fix the defects a completeness and accuracy review found
MaxFreedomPollard
committed
8cb722a
tests: lock in the review fixes, and correct three README claims
MaxFreedomPollard
committed
fcaefe8
close the rest of the review findings
MaxFreedomPollard
committed
279882c
correct the family-wise error figure and the test count
MaxFreedomPollard
committed
c089133

Commits on Aug 15, 2026

Commits on Aug 16, 2026

fix(stats): land the Wilson bounds exactly on 0 and 1
MaxFreedomPollard
committed
d68e7a9

Commits on Aug 29, 2026

Commits on Aug 30, 2026

fix(code): bind the interpreter's whole symlink chain into the sandbox
MaxFreedomPollard
committed
dee3c88