[devin-ai-integration]

Summary

Adds capture history to the measured and rejected list in docs/ROADMAP.md.

Capture history (#419) ordered captures within the swap's band by which had cut before. It played +1 ±6 over 6,000 games at 10+0.1 against b0600cb (sprt [0, 5] unresolved at the cap, LLR -0.86) and was not merged. Docs only, so no bench change.

Link to Devin session: https://app.devin.ai/sessions/eecd5bd45e4d476db1e20ed292a722ec
Open in Devin Desktop: https://app.devin.ai/desktop/session/eecd5bd45e4d476db1e20ed292a722ec?variant=devin
Requested by: @aywrite

---

🟢 Devin fixed the finding · Follow Devin's work

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Devin Review found 1 finding. Devin automatically fixed it.

Fixed by Devin (1)

• Capture history explanation breaks writing rule — Split the entry into short declarative sentences per the house writing rule; folded into the commit.

View findings in Devin Review · Follow Devin's work

[github-actions]

Speed against b0600cb9

Measured on Intel(R) Xeon(R) 6975P-C.
Both sides built with rustc 1.99.1 (58cf8f9c3 2026-10-13).

Both sides built and run on this runner in this job, the way
scripts/speed.sh measures a perf commit. Each round runs
both sides on a layout of its own, the same compiled code with
its code and data shuffled and moved, so the 95% interval carries
where the code landed as well as the run. The verdict holds the
interval against a 1% threshold. The default layout, the one a
release ships, is measured after as a diagnostic. The node
counts are the search's: they move when the search does, and a
speed change leaves them alone.

round     base nps  candidate nps  change
    1      5376697        5536500   +3.0%
    2      5607222        5503588   -1.8%
    3      5399112        5469159   +1.3%
    4      5689302        5568130   -2.1%
    5      5787359        5750813   -0.6%
    6      5520042        5479476   -0.7%
    7      5622527        5639709   +0.3%
    8      5642873        5529790   -2.0%
    9      5481169        5380866   -1.8%
   10      5266805        5400036   +2.5%
   11      5485133        5457738   -0.5%
   12      5463441        5572188   +2.0%
   13      5551107        5444414   -1.9%
   14      5297502        5493886   +3.7%
   15      5381618        5219855   -3.0%
   16      5086025        5126397   +0.8%
   17      5072641        5106176   +0.7%
   18      5096094        5132121   +0.7%
   19      5081069        5222091   +2.8%
   20      5220039        5111552   -2.1%
   21      5167298        5195382   +0.5%
   22      5059410        5186551   +2.5%
   23      5288093        5322462   +0.6%
   24      5251287        5421826   +3.2%
   25      5208172        5222021   +0.3%
   26      5322785        5252131   -1.3%
   27      5278501        5280004   +0.0%
   28      5299097        5312549   +0.3%
   29      5209279        5238299   +0.6%
   30      5275129        5263667   -0.2%
   31      5237227        5326798   +1.7%
   32      5294033        5398715   +2.0%
   33      5201406        5238025   +0.7%
   34      5330029        5251113   -1.5%
   35      5293515        5240811   -1.0%
   36      5346382        5347229   +0.0%
   37      5244979        5298614   +1.0%
   38      5285173        5346320   +1.2%
   39      5298316        5100057   -3.7%
   40      5111240        5156254   +0.9%

             nodes    time  median nps  faster half
base       9312660  1.76 s     5293774      5459787
candidate  9312660  1.75 s     5317506      5469482
change                           +0.4%        +0.2%

paired change +0.3%, 95% interval -0.4% to +0.8%
diagnostic, on the default layout alone -0.4%, 95% interval -1.0% to +0.5%, 13 rounds

no change beyond ±1.0%: the whole interval is inside it

Speed: +0.3% (bench nps, 95% interval -0.4% to +0.8%, 40 interleaved rounds over shuffled layouts vs b0600cb9)

The instructions each side's bench executed, counted under
cachegrind. The count repeats to within a few hundred
instructions, so a small change here is a real one, but it
prices instructions only: cache misses, mispredicted branches
and where the code lands are the speed's to show.

instructions against b0600cb9, one cachegrind run a side, bench

             instructions    nodes  per node
base       13,159,580,737  9312660    1413.1
candidate  13,159,580,725  9312660    1413.1
change             -0.00%             -0.00%

within ±0.7%, as far as edits not made for speed have moved the count