I built RightNow CLI, an open-source command-line tool that automatically optimizes CUDA kernels using AI. It analyzes your kernel, generates multiple optimized variants, benchmarks them, and gives you the fastest version—without any manual tuning.
Real-world gains on an RTX 4090: 10-50x speedups depending on the kernel.
It’s fully open source, free for personal and educational use. Commercial use requires a license.
Quick Start:
git clone https://github.com/RightNow-AI/rightnow-cli.git
cd rightnow-cli
pip install -e .
rightnow optimize your_kernel.cu
Check it out on GitHub: https://github.com/RightNow-AI/rightnow-cli
Open source, fast, and ready for serious GPU work.
Top comments (1)
These results are based on benchmarks from RightNow AI, the platform behind rightnowai.co/.