COMMITS
March 13, 2026
B
experiment 18: "check invariants" heuristic — F1=0.486 (22 PRs, 10 scored)
bhaktatejas922 committed
B
experiment 17: "second bug in same scope" — F1=0.338, DISCARD (quantity target in disguise)
bhaktatejas922 committed
B
experiment 16: "trace backward from side effects" — F1=0.444 (19 PRs, 10 scored, best yet)
bhaktatejas922 committed
B
experiment 15: WarpGrep v1→v2 upgrade (6 turns, Qwen3 format, 540K context)
bhaktatejas922 committed
B
experiment 14: coverage nudge at round 12 — F1=0.367 (12 PRs, keep/structural)
bhaktatejas922 committed
B
experiment 13: "trace with edge cases" heuristic — F1=0.341 (15 PRs, keep)
bhaktatejas922 committed
B
experiment 12: "compare both sides" heuristic — F1=0.373 (15 PRs, keep)
bhaktatejas922 committed
B
checkpoint before experiment 12
bhaktatejas922 committed
B
experiment 10: hypothesis-driven search — F1=0.379 (15 PRs, noise/keep)
bhaktatejas922 committed
B
checkpoint before experiment 10
bhaktatejas922 committed
B
checkpoint before experiment 9
bhaktatejas922 committed
B
checkpoint before experiment 8
bhaktatejas922 committed
D
docs: add SDK integration prompt for WarpGrep TypeScript SDK
DhruvBhatia0 committed
B
Restructure plan.md into autoresearch-style autonomous experiment loop
bhaktatejas922 committed
March 11, 2026
B
Online eval v10: F1=0.406 on 66 real-world PRs
bhaktatejas922 committed
B
Merge PR#2 features: skip_post mode, reasoning_effort, online eval, lower thresholds
bhaktatejas922 committed
March 10, 2026
B
Online eval baseline: F1=55.9% P=57.6% R=54.3% (13 PRs)
bhaktatejas922 committed
D
fix: default provider to openai, follow redirects on callback
DhruvBhatia0 committed
March 9, 2026
B
Broaden copy-paste pattern to near-duplicate line verification
bhaktatejas922 committed
B
De-overfit review prompts and add generalizable patterns
bhaktatejas922 committed
D
docs: add universal WarpGrep Python integration prompt
DhruvBhatia0 committed
D
B
Boost review recall with import verification, normalization checks, and error return detection
bhaktatejas922 committed
B
Single-stage XML review: eliminate extraction call, inline <issue> tags
bhaktatejas922 committed
B
F1=62.7%: Full iter 8 results, random eval, WarpGrep description cleanup
bhaktatejas922 committed
D
fix: post review summary even with 0 issues found
DhruvBhatia0 committed
D
fix: resolve Gemini $ref schemas, fix XML/JSON extraction mismatch, handle null content
DhruvBhatia0 committed
March 8, 2026
D
fix: define schema variable in _extract_issues
DhruvBhatia0 committed
D
fix: add missing sys import, keep 1 machine running to prevent autostop mid-review
DhruvBhatia0 committed