Sublime
An inspiration engine for ideas
Today we present a new approach for automated discovery of generalizable #ReinforcementLearning algorithms that evolves a population of loss functions — represented by computational graphs — over a set of simple training environments. Learn more at https://t.co/8n5aXArgg9 https://t.co/V2GmnEWx41
Google AIx.com
Completing the sprint to deliver on the commitments we made at #CASP14 last Dec, we’ve published 2 papers in @Nature, open-sourced #AlphaFold's code and provided a colab. 4/
https://t.co/OllIncSxti
https://t.co/oiiYKBQQSh
https://t.co/Q9kvuw8fGl
https://t.co/2Vd4itBfd2... See more
A Python decorator is all you need to trace LLM apps (open-source).
Most LLM evals treat the app like an end-to-end black box.
But LLM apps need component-level evals and tracing since the issue can be anywhere inside the box, like the retriever, tool call, or the LLM... See more
Avi Chawlax.com

OpenAI’s biggest rival isn’t just Deepseek. It’s Google.
We invited 200+ hackers to build what’s possible with Gemini multimodal AI and generative video
Here’s what we saw at the @GoogleDeepMind MultiModal Hackathon with @AgentOpsAI @windsurf_ai at @agihouse_org (🧵): https://t.co/UloJO189ro

Google DeepMind’s just dropped AlphaEvolve
> a self-improving agent that combines the generative power of Gemini Pro and Flash with automated evaluators and an evolutionary algorithmic loop
> generates complete algorithms in code, runs and evaluates them, and iteratively refines them based... See more

A groundbreaking study from Oxford and King’s College London just revealed that Large Language Models aren’t just text predictor, they’re strategic thinkers.
Researchers threw OpenAI, Google, and Anthropic’s top models into a brutal evolutionary version of the Prisoner’s Dilemma. And the results were straight-up... See more