Google DeepMind's AlphaEvolve advances computer science

0.7% sounds small. It isn't. Google spends so much on compute that 0.7% is probably more than our entire agency will bill in the next decade. That's not a feature — that's a structural advantage compounding every quarter.

Here's what actually matters: this isn't a chatbot finding efficiencies. AlphaEvolve discovered novel mathematical structures. It's not optimizing known solutions — it's finding paths humans hadn't mapped yet. That's a different category of thing.

Look, we've been watching AI write mediocre CRUD apps and summarize PDFs for two years. This is the first time I've seen a credible demo of AI doing science. Actual science. Not "AI-assisted" — autonomous discovery that stuck.

The cynical read: Google keeps this internally, compounds the advantage, and the rest of us never see the real implementation. They'll publish the paper, we'll train on it, and their fleet will already be three generations ahead.

Still — if the optimization techniques make it into open tooling (some will), the infrastructure cost calculus changes. Especially for anyone running intensive training loops or high-volume inference.