Latest AI news, expert analysis, bold opinions, and key trends — delivered to your inbox.
Google just dropped AlphaEvolve, a next-gen coding agent powered by Gemini and evolutionary strategies. Its mission? Solve the world’s hardest math problems and optimize Google’s own systems—all through AI-written code.
Gemini-Powered Workflow: AlphaEvolve combines Gemini Flash (for ideation) and Gemini Pro (for analysis) to generate code, which is then refined through testing and iterative evolution.
Real Discoveries: It’s already cracked historic challenges—like improving on Strassen’s algorithm, untouched since 1969.
Real Impact: AlphaEvolve is now helping Google optimize data center scheduling, speed up AI training (including its own), and even improve chip design.
Benchmark Results: When tested on over 50 open math problems:
It matched state-of-the-art (SOTA) results in 75% of cases
It discovered new, improved solutions in 20%
Yesterday, OpenAI’s Jakub Pachocki said there’s “significant evidence” that AI can generate novel insights. Today, Google proved it.
Math is the hidden language behind science, tech, logistics, and progress. And now, AI isn’t just solving equations—it’s inventing better ones.
AlphaEvolve could be the first glimpse of AI as a true scientific collaborator.