Beyond Green
Your tests passed.
We check what they missed.
Your tests passed.
We check what they missed.
One line changes: archiving deletes instead of hiding. The tile stays green.
get_by_text("Archive").click()
assertVisibleBookmarks(["1","3"])
assertVisibleTags([...])
assertReloads(0)
get_by_text("Remove").click()
assertVisibleBookmarks(["1","3"])
assertVisibleTags([...])
assertReloads(0)
Two operations whose whole difference is whether the data survives. Identical assertions.
A Scout proposes. A Critic attacks it and picks what to read. An Investigator revises. A Verifier rules, and can veto.
Walk the real trace in 3D →Step 5: it agreed, at 0.31. Below the bar, so it fetched the contract instead of concluding.
| Two runs each | TypeSafe | DeepSeek |
|---|---|---|
| Correct | 4/6 · 4/6 | 6/6 · 6/6 |
| Wrong | 0 · 0 | 0 · 0 |
| False positives | 0 · 0 | 0 · 0 |
| Latency | 5.3–6.1 s | 78.8–92.5 s |
Every failure is a refusal, never a wrong answer.
| Policy | Correct | 250 tests |
|---|---|---|
| TypeSafe alone | 8 / 12 | $0.14 |
| Cascade | 12 / 12 | $5.36 |
| DeepSeek alone | 12 / 12 | $15.67 |
Escalate only the refusals. Two thirds of the work never pays the higher rate.
| Documenso #2485 | Defective | Fixed |
|---|---|---|
| Upstream tests | 2 passed | 2 passed |
| Rows claiming a send | 2 | 0 |
| Emails actually sent | 0 | 0 |
A user-reported bug that shipped. Its own test loads the corrupted row and asserts the wrong two columns.
| Capture | TypeSafe | DeepSeek |
|---|---|---|
| Defective | abstains | regression |
| Fixed control | abstains | regression |
Neither separates the builds. DeepSeek flagged the fixed one too, so its finding is not a detection. Only the control caught that.
Open the flat version at index.html instead.