Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/dogacel/auto-gpu-kernel/optimizegit clone --depth 1 https://github.com/Dogacel/auto-gpu-kernelWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.01448 |
| Opus 5 | $0.00000 | $0.00724 |
| Sonnet 5 | $0.00000 | $0.00290 |
| Haiku 4.5 | $0.00000 | $0.00145 |
Grade A, and why
optimize scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 71 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/optimize — autonomous optimization loop
Iteratively improve solution/triton/sparse_fused.py. Rules in CLAUDE.md are non-negotiable.
Loop
IMPORTANT: Make sure research agent is called every 5-10 experiments to ensure we are not going in circles.
-
Assess. Read
sparse_fused.py,sparse_baseline.py,experiments/summary.md,experiments/LESSONS.md. For directly relevant prior attempts, readexperiments/exp_N/result.md. If the highest-numbered folder hasplan.mdbut noresult.md, implement that plan — it's reserved (see §Folder reservation). -
Plan one change. Early progression: PyTorch → tiled Triton → fused → tile tuning → alternative tilings. Don't skip structural wins for micro-tuning. Scan
summary.mdfor similar past attempts; if close, articulate what's different this time.Study workload characteristics —
seq_lendistribution across the 128 workloads,batch_size,max_num_pages, any exploitable structure — and branch the kernel when a regime admits a cheaper path. Input-characteristic specialization can beat a one-size-fits-all kernel by orders of magnitude. Fair game as long as the win is real and not workload gaming.Minimize launches and copies. Fold prologue/epilogue ops (initialization, padding, masking, remapping) into the main kernel rather than calling them as separate launches. Avoid
.contiguous()when you can plumb strides into the kernel instead — each copy is both a launch and a memory round-trip. -
Implement. One optimization. Preserve DPS (don't allocate
output/lseinside the kernel). -
Validate.
/benchmark quick(2 workloads: smallest + largest — catches shape-assumption bugs). Fix compile/correctness before proceeding. -
Measure.
/benchmark stride 2. Before trusting the number:- Reference-latency sanity: if the ref latency is >30% off the moving median from recent
summary.mdrows, the VM is anomalous — re-run once. - Sub-5% deltas are noise on cross-VM comparison. Confirm with
modal run scripts/ab_benchmark.py::run --a experiments/exp_<prev-best>/sparse_fused.py. - Report latency split into small/large groups when both are present; aggregated means hide regime-specific regressions.
- If results are looking good or you are not certain due to noise, proceed with
/benchmark full.
- Reference-latency sanity: if the ref latency is >30% off the moving median from recent
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 71 lines · 0 tokens per session scan A 7d57b7c9b449
optimize is a command published in the GitHub repository Dogacel/auto-gpu-kernel (157 stars, last pushed 11d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,448 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
amk-autoresearch
Launch the unattended / overnight AMK autoresearch driver on $ARGUMENTS (model [gpu] [minutes|iters]).
amk-optimize
Drive an interactive AMK propose -> eval -> keep/revert megakernel schedule session on $ARGUMENTS (model [gpu]).
amk-compile
One-shot compile + verify a model into a CUDA megakernel via amk compile on $ARGUMENTS (model [gpu]).
initref
Build a reference for the implementation details of this project. Use provided summarize tool to get summary of the files. Avoid reading the content of many files yourself, as we might hit usage limits. Do read the content of important files though. Use the returned summaries to create reference files in /ref…
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.