Mind Lab's unfair advantages: 1T-scale LoRA RL. Train 10x cheaper. Everyone else is bleeding cash.
Memory Diffusion. Forgets the noise. Keeps what counts.
OpenAI, DeepMind, MIT, Tsinghua refugees. 200+ papers. 30k citations. These people actually built shit that matters. Not a
RESEARCH
-

Mind Lab’s Technical Innovations in AI Training and Memory Architectures
By
–
-
CuspAI hiring: join generative models team for materials research
By
–
π₯π₯Amazing opportunity π₯π₯ Work with my awesome and wonderful colleagues @johannbrehmer @wellingmax and @lonepair on generative models for materials here at @cusp_ai! Fun and exciting research on real-world problems that actually matter! #MachineLearning #Materials Johann Brehmer (@johannbrehmer) Come work with us at @cusp_ai in the generative model team! Excited about flow / diffusion models and chemistry? Looking for impact? jobs.ashbyhq.com/cuspai/b810β¦ Join a great team lead by @wellingmax & Aron Walsh, work in Amsterdam / Cambridge / London / Berlin. 1/2 β https://nitter.net/johannbrehmer/status/2000488276029813139#m
β View original post on X β @wellingmax, 2025-12-15 09:08 UTC
-
CuspAI hiring generative model researchers in multiple European cities
By
–
Come work with us at @cusp_ai in the generative model team! Excited about flow / diffusion models and chemistry? Looking for impact? jobs.ashbyhq.com/cuspai/b810β¦ Join a great team lead by @wellingmax & Aron Walsh, work in Amsterdam / Cambridge / London / Berlin. 1/2
β View original post on X β @wellingmax, 2025-12-15 08:49 UTC
-

2.6K Parameter RL System Matches 62B Specialized Model
By
–
Oh My! I just built a new RL-native system with 2.6k parameters that matches a specialized 62B model from a 2024 paper, specifically on GSM8k. That's over 20,000,000x smaller than the baseline β but the approach is so different the traditional comparison doesn't make sense? π
-

AI Tools Empower Early Childhood Education Progress
By
–
#TFW Small steps guide children into the big world. Our #ECCE tools: Are being built with early educators
Aim to align with daily routines Strive to help children progress at their pace #AIforSocialImpact #EarlyLearning #WadhwaniAI #AIforBharat -

Technical Comparison of Claude and ChatGPT Memory Architectures
By
–
someone just reverse-engineered claude's memory system and it's nothing like chatgpt's. turns out most people think AI memory works the same way everywhere. not really. chatgpt pre-computes summaries and injects them into every single prompt. convenient. predictable. always
-

RouteRAG: Adaptive Retrieval for Multi-Hop Reasoning Systems
By
–
Adaptive retrieval is the way to go! And this RouteRAG paper shows why. Let's talk about it: RAG systems have a retrieval problem. The default approach to multi-hop reasoning today relies on fixed retrieval pipelines. It typically involves fetching text + maybe graph data, and
-

Budget-Aware Tool-Use Enables Effective Agent Scaling
By
–
Budget-Aware Tool-Use Enables Effective Agent Scaling Liu et al.: https://
arxiv.org/abs/2511.17006 #ArtificialIntelligence #DeepLearning #MachineLearning -

Analyst Consensus Index Built with Claude AI for Bitcoin Prediction
By
–
Built an Analyst Consensus Index from 246 curated analysts using Claude AI. 5-year backtest on Bitcoin: predicted the 2022 crash, 2023 rally, and this correction.
-
The Digitalist Papers: Economic Challenges of Transformative AI
By
–
It's here! The Digitalist Papers is the definitive set of essays by some of the world's most insightful economists, AI researchers, policymakers and innovators. This volume explains the economic challenges created by transformative AI and what we must to do now.