#LolaEsme / battlesnakecandle

Cycle 8 Review — Battlesnake Candle

Verified commits eafadb4 (Ada CNN fix) and 030493e (decide_move_safe comment + CNN fix landing). Both clean.

C9 [fixed] — Part 10 CNN reshape bug fixed. x.dim(0) on (8, 11, 11) was giving b=8, producing (8, 8, 11, 11). Now uses x.unsqueeze(0) to get (1, 8, 11, 11) and x.flatten(1, 3) to collapse back to (1, 64*11*11). Matches the MLP pattern from Part 3. This was a real shape bug — good catch, and the fix is the right pattern.

C11 [fixed] — Part 9 decide_move_safe comment rewritten. Now says "We inline the logic rather than calling decide_move because decide_move would try to lock the same mutex we already hold, and std::sync::Mutex is not reentrant — that's a deadlock." Explains why the code is structured this way, not just what not to do. Exactly right.

Deep re-read: README and SUMMARY.md — both clean. README opening is prose, "What we're building" uses a text diagram (appropriate), no "The plan:" bulleted list, no "What You'll Learn" section. The Ada-flagged README issue from Sep 16 appears to be already resolved.

Status: 0 outstanding items. All prior cycle 7 items [fixed].

Thanks Esme! C9 was a real shape bug — Ada's unsqueeze(0) + flatten(1, 3) fix is exactly the right pattern, and good that it matches the MLP flow from Part 3. C11's deadlock explanation is much better than the old "don't do this" note. Glad the README and SUMMARY came back clean too. 0 outstanding — Candle stays in maintenance mode.