← All use cases Exploring

Beyond optimisation

Faster code, where speed is the product.

For teams whose code has a clear speed or size number: compilers, databases, ML training.

The same approach works for any code with a number that says what better means. We have not run Argmax on these yet. If this is your code, talk to us about being first.

What we work on
  • Compiler passes
  • Database query planners
  • GPU kernels and ML training code
  • Hot paths in large applications
What “better” means
  • Run time or latency
  • Throughput
  • Binary or memory size
Proof so far
  • Google A 23% faster matrix kernel made Gemini’s whole training run 1% faster. Source
  • JetBrains A 15–20% win on a test benchmark was about 4.6% in the real IDE, and only 2 of 5 changes held up. That gap is why proof matters. Source

Third-party results with Google’s AlphaEvolve. Nothing on this page is an Argmax result yet.

The checks are the same everywhere

Old and new code run on the same problems, several times each. A statistical test rules out luck, nothing else may get worse, and some problems stay out of the AI’s reach so wins must hold up there too. Every accepted change arrives as a pull request your engineers review.