# OpenBMB/ForgeKernel

> A self-evolving GPU kernel agent that starts from a reference kernel, profiles it, and iteratively generates faster implementations via a coding agent. Aimed at researchers and engineers optimizing kernels on NVIDIA Hopper/Blackwell.

- Magnitude: 2.6 out of 10 — Cooling
- Stars: 155 total · +6 stars measured 2026-11-02, 08:35–10:56 UTC, ≈ 16 by evening
- Star trust: star growth looks organic
- Category: AI agents · Language: Python · License: Apache-2.0 · Created: 2026-10-30 · Last push: 2026-10-30
- GitHub: https://github.com/OpenBMB/ForgeKernel · Page: https://gitnova.dev/en/r/OpenBMB/ForgeKernel

## Useful for

- Run forge new tasks/rmsnorm to get an optimized CuTe DSL RMSNorm kernel replacing nine PyTorch launches.
- Run self_evolve.py to iteratively improve a kernel by promoting the best version to the next reference.
- Check the optimization plan via forge new --dry-run before starting a multi-hour run.

## Why it’s here

- Star-counter measurements on 2026-11-02 (UTC), 08:35–10:56: 149 → 155 stars (+6). This is the change over that interval.
- Estimated end-of-day forecast: about +16 stars, using observed gains and the previous day.
- The repository is 3 days old and already has 155 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #162.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 18
- Issues and pull requests: 1
- Watchers: 3
- Average over the last week: 39 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 0

## Stars per day, last 4 days (oldest → newest, today is partial)

2026-10-30 … 2026-11-02: 81, 32, 25, 6

## Spotted in now

- Top new repositories this week: #162

## Similar by description

1. **NVIDIA/cutlass** — 1.2 · Steady · Language models · C++ · +0 stars measured 2026-11-02, 00:14–11:03 UTC, ≈ 1 by evening
   CUTLASS is a collection of CUDA C++ templates and Python DSLs (CuTe DSL) for implementing high-performance linear algebra, primarily GEMM, on NVIDIA GPUs. It targets researchers and performance engineers writing optimized Tensor Core…
   Full card: https://gitnova.dev/en/r/NVIDIA/cutlass.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-11-02 11:06 UTC, updated every 30 minutes.