"VIMPO: Value-Implicit Policy Optimization for LLMs" While GRPO is simple because it avoids a critic, it still gives every token in a reasoning trace the same reward signal. This paper tries to get the best of both worlds by deriving the value function from the policy itself
AI
-
AI generates precise mechanical iris design in CAD
By
–
Use Case 4: Computer Aided Design of Mechanical Iris
— Sakana AI (@SakanaAILabs) 22 juin 2026
Can an AI generate precise, functional mechanical designs?
We tasked Fugu Ultra with creating a mechanical iris in CAD, similar to a camera aperture where multiple blades must move together to cleanly open and close a central… pic.twitter.com/Y27QDdcYipUse Case 4: Computer Aided Design of Mechanical Iris Can an AI generate precise, functional mechanical designs? We tasked Fugu Ultra with creating a mechanical iris in CAD, similar to a camera aperture where multiple blades must move together to cleanly open and close a central
-
2015 prediction on AI assistants and agents
By
–
2015 https://youtu.be/K5a1uthRHf8 Everything is there AI assistants, agents, expertise in your pocket etc…
-
Pre-detonation file analysis for OT-connected AI pipelines
By
–
Analyzing files before detonation means evaluating risk without executing the file in a live environment. That matters when the file is destined for an AI training pipeline connected to OT systems.
-
AI-assisted file analysis improves speed but requires policies and validation
By
–
The key distinction from the session: AI-assisted file analysis improves speed and coverage but does not replace the need for defined policies, clear data ownership, and repeatable validation processes.
-
Free MIT guide to key concepts of computer vision
By
–
A free MIT guide to key concepts of computer vision: https://bit.ly/43Tn1vW
-
Fugu Ultra tested on 50 weeks of stock data
By
–
Use Case 2: Financial Time Series Prediction
— Sakana AI (@SakanaAILabs) 22 juin 2026
Can an AI agent navigate sequential, no-look-ahead market decisions?
Just for fun, we tested Fugu Ultra on 50 weeks of historical data for an anonymized equity (STOCK_X). Starting with $10,000, the agent processes weekly market data… pic.twitter.com/BcqCfUfw2dUse Case 2: Financial Time Series Prediction Can an AI agent navigate sequential, no-look-ahead market decisions? Just for fun, we tested Fugu Ultra on 50 weeks of historical data for an anonymized equity (STOCK_X). Starting with $10,000, the agent processes weekly market data
-
Autonomous ML Research: Fugu Ultra improves GPT model via AutoResearch
By
–
Use Case 1: Autonomous ML Research
— Sakana AI (@SakanaAILabs) 22 juin 2026
Can an AI autonomously improve another AI’s training recipe?
We tasked Fugu Ultra with improving a small GPT model using AutoResearch. Over 14 hours on a single H100 GPU, Fugu ran > 100 experiments. It iteratively edited the training code, ran… pic.twitter.com/Gp96FEQ797Use Case 1: Autonomous ML Research Can an AI autonomously improve another AI’s training recipe? We tasked Fugu Ultra with improving a small GPT model using AutoResearch. Over 14 hours on a single H100 GPU, Fugu ran > 100 experiments. It iteratively edited the training code, ran
-
That moment you realize a startup with zero employees exists
By
–
That 'oh sh*t' moment when you realize the 0-person startup is freaking real: https://t.co/6kzwtaqdrK pic.twitter.com/9MP2fApaeb
— Charly Wargnier (@DataChaz) 22 juin 2026That 'oh sh*t' moment when you realize the 0-person startup is freaking real:
-
LLMs cannot be trusted: alignment must move on
By
–
LLMs can’t be trusted to follow rules. Which means they can’t be trusted, period. If we are going to solve alignment we must move on.