Project

AlphaEvolve

Google DeepMind's Gemini-powered coding agent, announced May 2025, that evolves programs by proposing, scoring and mutating them to discover new algorithms.

Known aliases

AlphaEvolve V2

Current stories

investConfirmed4 publishers

Google's $3 million for up to 200 European scientists works out to $15,000 each

Google DeepMind and Google Cloud will give up to 200 European researchers access to frontier models and $3 million in compute credits. The pool is small, so any lock-in would come from the habits labs build around five Google models.

Publishers:blog.google channelnewsasia.com cryptobriefing.com mezha.net thenextweb.com

Perspective Coverage

4 publishers

Builder: Builder 41%
Operator: Operator 20%
Investor: Investor 39%

Reality

Evidence55
Adoption: Insufficient
Hype gap+30
Incentives: Insufficient
Confidence60

buildOne report1 publisher

Google's Dream-RSI cuts Gemini calls 42% on a Lasso task by evolving only its search policy

Google researchers' Dream-RSI cut Gemini calls on a Lasso solver task from 550 to 317 by rewriting a Python search policy, with every model frozen. Teams running scored code search can test it as a call-budget saving on their own tasks.

Publishers:dev.to

Reality

Evidence45
Adoption: Insufficient
Hype gap+60
Incentives: Insufficient
Confidence40

buildConfirmed6 publishers

Anthropic's 26% figure grades Claude's R&D tasks by how much human steering they took

The lab says Claude leads 26% of its model research and collaborates on about 90%, with both tiers defined by how closely a human directs each task, and it wants rival labs publishing the same measure.

Publishers:dev.to economictimes.indiatimes.com mezha.net the-decoder.com theneuron.ai thesequence.substack.com

Perspective Coverage

6 publishers

Builder: Builder 39%
Operator: Operator 33%
Investor: Investor 28%

Reality

Evidence50
Adoption65
Hype gap+30
Incentives70
Confidence60

buildConfirmed2 publishers

Replaying a logged search tree cut a Dream-RSI task from 550 attempts to 317

Google and DeepMind's method records a live search, replays thousands of exploration policies against the stored results, and sends only the winner into the next run. No weights are retrained.

Publishers:arxiv.org the-decoder.com

Reality

Evidence55
Adoption12
Hype gap+20
Incentives68
Confidence58

buildOne report1 publisher

The judge went synthetic first, which tells you which part of your pipeline is next

Latent Space dates five flips from human-made to model-made since 2022. The order they happened in predicts more than the annual cadence the essay claims for it.

Publishers:latent.space

Reality

Evidence42
Adoption58
Hype gap+32
Incentives55
Confidence48