BUILDWAKE PILOT MEASUREMENTS — 21 OCTOBER 2026

These are single observed cold/warm pairs within Buildwake, not a promise of
future performance or a controlled comparison against another CI provider.

Application                   Workflow step        Cold   Warm   Difference
Next.js marketing app          Checks + unit tests  235s   224s   11s less (4.7%)
Next.js marketing app          Build                156s   142s   14s less (9.0%)
Next.js marketing app          Integration tests    335s   326s    9s less (2.7%)
Integration-heavy web app      Type checks          101s    93s    8s less (7.9%)
Integration-heavy web app      Tests                237s   227s   10s less (4.2%)
Integration-heavy web app      Build                265s   275s   10s more (3.8%)

METHOD
- One pair per workflow; identical source for cold and warm phases.
- Each step has 4 CPU quota and 6 GiB configured memory; capacity one workflow.
- Elapsed workflow-step time includes installation and the lane's checks.
  Clone, queue and full pipeline overhead are excluded.
- Cold begins without the package download cache. Warm retains only that
  workflow's npm/Bun download cache. Both phases reinstall, regenerate and run
  the full selected build/test checks. Test databases are fresh in each phase.
- Matching existing CI lane results were checked; pre-existing skips remain.
- No cross-PR persistent cache, concurrent load or daily-volume qualification.

IMPORTANT DETAIL
The Next.js marketing app's compiler took 96 seconds in both phases. The build
step's 14-second improvement is not a 14-second compiler speedup. A warm download
cache did not make every workflow faster: the other build took 10 seconds longer.

LIMITS
The sample size is one pair, so variation and repeatable savings are unknown.
Different hardware/cache policies prevent attributing a provider speedup.
Deployment time, monetary savings, production capacity and availability are not
measured here. No daily/monthly saving is extrapolated.

SOURCE RECORD
Accepted M5 same-source comparisons retained by the Buildwake application owner:
Next.js marketing app tree 82c7ae818c2cb554e07b00b0b4fa7c3480f48677;
integration-heavy web app tree aac14b411166ad7425e648dac563923e0f4e3921.
This is a sanitized aggregate; raw logs and private repositories are not public.

TERMINAL DEMO
The hero shows selected output from an actual local `bw demo` run on Node.js
22.23.1 on 21 October 2026. A temporary PATH entry selected the standalone Buildwake
CLI; no global install, pilot request or credential was used. The full transcript
is available in the page. Its three sample tests are not a hosted build or a
performance benchmark. Replay itself executes nothing.

The downloadable bw file is generated unchanged from apps/cli/bw.mjs in the
Buildwake source. After downloading, run chmod +x ./bw then ./bw demo, or use
node ./bw demo (including on Windows). Node.js 22 or later is required.

The read-only pilot commands are bw builds OWNER/REPO [NUMBER] and
bw logs OWNER/REPO NUMBER STEP_ID. They require privately configured authorized
BUILDWAKE_SERVER and BUILDWAKE_TOKEN environment values. No build-start or
deployment command is implemented or implied. No live pilot API request was made
for this website; CLI protocol/security checks use controlled fixtures.