AutoMegaKernel: An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch-1 LLM decode.
Pillar = mean of 2 scaled values = 10.4.
Awaiting first reading — these signals apply to this agent and will be ingested on the next tier tick: PyPI monthly installs, SO questions (7d), Product Hunt upvotes, Docker Hub pulls, Crates.io downloads (90d), Tech-news mentions (30d)
Not applicable — this agent doesn't have the prerequisite (no GitHub repo, no HF mirror, etc.) for these signals to ever apply: HF downloads (30d), npm weekly installs
[](https://agenttape.com/agents/automegakernel)
<a href="https://agenttape.com/agents/automegakernel"><img src="https://agenttape.com/api/badge/automegakernel.svg" alt="AgentTape" /></a>