[ashhart]

What this changes

Require ATC 0.9.7 so contributors can check and claim issues when TensorFold has more than 300 open PRs. The fixed reader supports up to 2,500 PRs; documentation keeps overflow, timeout and board-file comparison limits explicit.

• Issue: n/a, direct maintainer-requested coordination repair. Native CUDA: bring up Spark-X2.6-4B affine-4 inference on Pascal sm_61 #3136 exposed the upstream read limit; this does not implement that engine issue.
• Stacked on: main, base 223da76. Merge after Support repositories with up to 2500 open pull requests ATC#17 is available to install.
• atc overlap --json: run 13 November 2026 against origin/main. Repository-wide diagnostic remains incomplete because its ordinary overlap budgets are smaller than discovery. Two working-tree conflicts with lfm3: serve LFM3-1.7B on CUDA: NVFP4 and FP8 weight-only, MLX 4-bit (stacked on #1513, #1492 and #1503) #1508 and lfm3 CUDA: serve the block FP8 export weight-only and the MLX 4-bit export as stored (stacked on #1508) #1634 are inherited README conflicts. The ATC inherited-conflict proof returns clean for this exact patch against b60ae9e and 3551fd8, with cacheable false. It does not grant repository-wide clearance.

Checks on this head

• Head commit: 9ac3293
• Lean checker: 341 on base and 341 on head, zero new findings. These are inherited findings, not a clean lean count.
• Tests run: python3 tools/pr_check.py --base origin/main, 4 documentation files, 0 errors and 0 warnings; git diff --check passes. ATC candidate has 455 passing tests and 174 module checks. A real candidate check 3136 returns complete, no limitations and free, where 0.9.6 returned pullRequestsLimit.
• CI: pending on submission; no CI pass claimed.
• Not run: Zig host/device/model suites and app suites, because only contributor documentation changes. No inference, build or app code changes.

Receipt

n/a, contributor documentation cannot reach inference. No model, kernel, arithmetic, setting or platform implementation changes. No runtime or app publication in this PR.

Checklist

• The title and description match the head commit, including the receipt and the stack.
• New tests fail before the change, pass after it, and skip cleanly without their dependency. n/a for these documentation changes; the companion ATC PR contains the regression tests.
• Comments and docstrings are one line. No measurements or history in the source.
• No personal data, machine names, internal hosts or local paths in the diff or its commits. No AI attribution lines.

[ashhart]

/atc exchange ed95be023aed1fde5b13706c44e1a5b9380025ea1c9a2b6a5eb2e218f87145a4