Experiment: 76f0852

rad experiment show 76f085222588269280eb45de10037a5153e05c29

by ee@did:key:z6MkvsiybCuk1WDZhh2aXDx7NTiZejKgWZygMNsUuXDMNvVQ · Apr 11 20:05 2026

skipDefaultLibCheck

Measurements

Metric	BaselineThe benchmark measurement of the unmodified code	CandidateThe code with the proposed optimization applied	DeltaThe performance change from baseline to candidate
tsc_check_ms (ms) primary	1066.000 ms (n=1)	1089.000 ms (n=1)	+2.15% REGRESSED

Annotations

hypothesis	skipDefaultLibCheck will reduce the time tsc spends type-checking the built-in TypeScript lib.*.d.ts files, tightening the cold run and possibly the warm runs
next_action_hint	Do not try other cold-run-only optimizations (noResolve, typeRoots shrinking, etc.) against this benchmark — the median is pinned by warm-run steady state. To find further wins, either (a) reduce warm-run cost by pruning the file graph (import type, dead-code elim, removing re-export barrels) or (b) change benchmark semantics. Keep skipDefaultLibCheck in mind for a future cold-build optimization if that ever becomes the metric.
rollback_reason	Cold run dropped from 5764 to 5183 ms (-580 ms) but the benchmark median is dominated by 4 warm runs at ~1060-1100 ms where lib files are already cached. Median went 1066 -> 1089 ms, a tiny regression within noise. No steady-state gain.

Base CommitThe starting commit before the optimization was applied	9fad0de6ded07745d50acbf9c5c9965612d5548e
Candidate CommitThe code with the proposed optimization applied	9fad0de6ded07745d50acbf9c5c9965612d5548e
Config	optimize.yaml

Diff

No changes

Environment