gensyn

When you run the same AI model with the same inputs twice, you'd expect the same output. But modern GPU execution is optimised for speed, not fixed ordering, and existing determinism tools do not solve this across hardware. Today we're changing that. blog.gensyn.ai/ree/

"Now I can prove that given this exact input, this model will always give this output. And therefore I can prove why this model gave that output." Tomorrow, join Gensyn engineers for a technical dive into REE. 11.15 am ET. x.com/i/spaces/1RJZz…

we're holding a live REE workshop at EthCC! learn how to verify and audit your AI agent/chat/inference with receipts and attestations on your own machine we'll have REE core developers there to walk you through exactly how to use the software luma.com/ETHCC-REE

Join Gensyn's live REE workshop with REE core developers at EthCC. luma.com/ETHCC-REE

Everything you need to build on REE is now documented. Start here👇 docs.gensyn.ai/tech/ree @GitBookIO

Tomorrow, Gensyn CEO @benfielding will be on AItopia with @Anitahityou speaking on coordination, lessons from RL-Swarm, Delphi and the future of Open AI. Wednesday, 5.45 PM PDT.

"People didn’t have any means to reliably repeat ML. Now they do." Tomorrow, @gensynai engineers James Decker and Jesse Walker explain REE, why it's important and what you can do with it today. x.com/i/spaces/1oKMv…

Humans increasingly rely on AI to determine outcomes, from prediction markets to credit scores to medical diagnoses. How do we guarantee the model ran correctly and produced the right output?

reproducible execution doesn't need TEEs introducing REE - Reproducible Execution Environment as AI models become more autonomous, it becomes increasingly important that we can verify and audit their execution REE allows this to be done on any device, not just TEEs

Just pushed some new docs (!) live for @gensynai's Reproducible Execution Environment (REE), their toolchain for bitwise-reproducible AI inference across any hardware. The problem it solves is deceptively simple: run the same model with the same inputs on two different machines, get the exact same output. Not approximately, not almost, but.. identically It turns out that's really difficult to do, because GPUs are non-deterministic by default, and existing solutions like PyTorch's deterministic mode can't get the job done the moment you switch hardware. REE solves this with custom GPU kernels (RepOps) that guarantee identical results everywhere. Every run produces a cryptographic receipt that anyone can independently verify. The docs cover everything from a quickstart to the internals of the MLIR compiler and RepOp kernel design. Check em out! -> docs.gensyn.ai/tech/ree

6/ Full docs - setup, model compatibility, receipt verification, and CLI reference with everything you need for reproducible inference. gensyn.ai/ree

5/ REE is live. 40+ supported models across 20+ GPU targets: covering the 3/4/5-series Nvidia RTXs and every generation of datacenter hardware Volta to Blackwell. `git clone git@github.com:gensyn-ai/ree.git cd ree python3 ree.py` docs.gensyn.ai/tech/ree/get-s…

When you run the same AI model with the same inputs twice, you'd expect the same output. But modern GPU execution is optimised for speed, not fixed ordering, and existing determinism tools do not solve this across hardware. Today we're changing that. blog.gensyn.ai/ree/