# lockedapart — VALIDATION RESULTS

## Signal hypothesis

WebGPU is unavailable in most headless Chrome configurations (no GPU process,
software rasterizer limitations). 10 WebGPU traces that all fail is a strong
headless indicator.

## 2-test results (vanilla headless vs CDP-attached headless)

**0 lines differ** between conditions. Both produce identical output:
"All 10 traces failed — no WebGPU support"
"🔴 Detection Signal: All traces failed — consistent with headless/emulated GPU"

## Verdict: ✅ VALIDATED — already in production

This is the strongest 2-test signal in the entire spike suite. Zero variance
between vanilla and CDP. The WebGPU check is fully decoupled from CDP
attachment state.

## Recommendation

The existing `webgpu_lockedapart` Phase 1 check covers this signal. No
additional finding required. The lockedapart spike is the **gold standard
for "clean headless detector"** — use it as the reference pattern for future
spikes.

## Files

- `demo.html` — the spike page
- (no test.py — used `_generic_test.py` for validation)