Progress & loops — reading a re-review
A re-review returns a fresh list of findings. On its own, "4 findings became 5" tells you nothing — did the old four get fixed, or come back? Kiln answers that by comparing rounds and labelling what changed.
The delta summary​
After a re-review, the layer shows a one-line summary of what changed since the previous round:
- ✓ resolved — flagged last time, gone now;
- ↻ still open — was flagged in the previous round too;
- ✦ new — surfaced this round (often introduced by a regeneration);
- ↺ recurring — flagged in an earlier round, went away, and is back.
Resolved items are listed struck-through so you can see what your changes cleared, and each remaining finding carries its own ↻ / ✦ / ↺ marker.
The oscillation warning​
↺ recurring is the important signal. A generative Apply rebuilds the whole layer, so it often fixes the flagged items but re-breaks others it fixed a round or two ago — the layer oscillates between two sets of concerns instead of converging. When Kiln detects this, it drops the soft nudge and shows an explicit warning:
⚠This layer is going in circles: Apply regenerates the whole layer and re-introduces concerns it fixed earlier. Re-applying won't converge here — edit the flagged spots by hand, or stop.
That's your cue to stop pressing Apply and instead Fix the specific concerns (which converges) or Ignore the ones you accept.
The AI reviewer is a fresh, subjective pass each time, and Apply is generative (it rebuilds the layer rather than patching one spot). Both together mean a layer won't naturally settle to zero findings — which is expected, not a bug. The delta and the warning make that legible so you know when to stop.