Every run now rebuilds the whole dependency graph · CI/CD Performance
The rust-cache step is gone and sccache has no persisted backend on these runners, so nothing is cached between runs. Each job pays the full dependency build. Restore the cache step or give sccache a storage backend.
Suggested fixAI fix prompt
IonWarp caught 7 issues other models missed
Real IonWarp comments on open-source pull requests, graded against the benchmark's answer key.
Missed byGLM 5.3 Flash 1DeepSeek V4.1 Flash 1GPT 5.6 Luna Pro 3Muse Spark 1.3 6
P1Zig cache no longer restores anything in preview/release/manual builds · CI/CD Performance
When a preview publish, release tag, or manual artifact build triggers, replacing mlugg/setup-zig with vercel-labs/setup-zig removes the action's Zig global cache, and these workflows have no actions/cache restore/save to compensate, so every run cold-compiles libghostty-vt.
MediumKnown issue in the CI/CD Performance benchmark · caught by DeepSeek V4 ProSee the graded case
P2DerivedData cache has no restore-keys fallback, so any native source change means cold Xcode build · CI/CD Performance
The cache key hashes all native Apple/Common source files, so every source edit rotates key and stalls zero restore fallback; actions/cache misses each new native change and Xcode rebuilds DerivedData from scratch.
MediumKnown issue in the CI/CD Performance benchmark · caught by DeepSeek V4 ProSee the graded case
P1sccache replaces Swatinem/rust-cache without a shared backend · CI/CD Performance
When a CI-push run lands on a fresh or different runner, cargo compiles project crates from cold because sccache has no remote backend and its local cache was never saved; the previous Swatinem/rust-cache step that restored from the GitHub Actions cache is gone.
MediumKnown issue in the CI/CD Performance benchmark · caught by DeepSeek V4 ProSee the graded case
P1Every push to an open PR branch runs the build twice and superseded runs are not cancelled · CI/CD Performance
When a commit is pushed to a branch that heads an open PR, both `push` and `pull_request` (synchronize) fire, so the full macos build runs twice for the same commit; with no concurrency group, prior in-progress runs also keep going.
MediumKnown issue in the CI/CD Performance benchmark · caught by DeepSeek V4 ProSee the graded case
DeepSeek V4 Pro is its best model on this benchmark; CI/CD Performance runs DeepSeek V4.1 Flash unless you pick another.
CI/CD Performance on the benchmark
IonWarp's models and the review apps, graded on the same open-source pull requests.
Blue: DeepSeek V4.1 Flash, the model IonWarp runs for CI/CD Performance. Yellow: DeepSeek V4 Pro, the model a failed review retries on. Sorted by Reviewer Score (F1 counts 3×, time and cost 1× each). Rows not graded on every PR come last. Review times come from runs on Sep 21, 2026 (models). Click a column heading to sort.
On this benchmark DeepSeek V4 Pro scored higher than DeepSeek V4.1 Flash (72.7% F1 against 59.5%). Why IonWarp runs DeepSeek V4.1 Flash: in a trial of our specialist reviewers on production pull requests on 23 September 2026 it filed about as many findings per review as GLM 5.3 Flash at high effort (0.93 against 1.02) for a third of the cost ($0.039 against $0.125), and 3 of its 64 findings were refuted.
Pricing
Simple pricing for your whole team.
One workspace price with seats included. Credits pool across your team. Unlimited repos on every plan.
CI/CD Performance is on Max: $149 a month with 10 seats and a 7-day trial.
P1Zig cache no longer restores anything in preview/release/manual builds · CI/CD Performance
When a preview publish, release tag, or manual artifact build triggers, replacing mlugg/setup-zig with vercel-labs/setup-zig removes the action's Zig global cache, and these workflows have no actions/cache restore/save to compensate, so every run cold-compiles libghostty-vt.