Can LLM agents learn by dreaming? DreamGym from Meta is a new framework that lets AI agents train via synthetic reasoning-based experiences instead of costly real rollouts. It models environment dynamics, replays and adapts tasks, and even improves sim-to-real transfer.
LLMS
-

ReAct Framework in CrewAI: LLM Problem-Solving with Tools
By
–
ReAct is really popular. Here's an example from CrewAI, which uses ReAct to let LLM think through problems and use tools to act on the world.
-
Codex Team Makes Rapid Incremental Progress
By
–
Codex team continues making rapid incremental progress
-

ReasonMed: 370k Medical Reasoning Dataset Advances Clinical AI
By
–
ReasonMed: the largest medical reasoning dataset, advancing LLM performance in clinical QA. Comprising 370k curated examples distilled from 1.75M reasoning paths, ReasonMed is built through a multi-agent EMD (easy–medium–difficult) pipeline with generation, verification, and an
-

Apple and Google Partner on Gemini-Powered Siri Integration
By
–
Apple x Google: A New AI Era for Siri Apple has reportedly finalized a $1B+ annual deal to integrate Google’s Gemini (1.2T parameters) into Siri — a massive step away from in-house AI toward next-gen intelligence. iOS users could see this overhaul as soon as next year. Full
-

Kimi K2 Thinking: Open-Source Reasoning Agent Breakthrough
By
–
Introducing Kimi K2 Thinking — A New Era of Open-Source Thinking Agents! The frontier of autonomous reasoning, intelligent tool use, and scalable cognition has arrived. SOTA Performance: HLE (44.9%) | BrowseComp (60.2%) Handles 200–300 sequential tool calls autonomously
-

Google develops a Chrome extension for NotebookLM
By
–
Google is working on a Chrome Extension for NotebookLM (not available yet). Possibly, it will let users add open pages as sources to the selected notebook.
-
Diffusion Models for Language: Image-Only Input Output Approach
By
–
Side quest: Diffusion model but instead image only input/output but for language modeling.
-
Comet Browser amélioré avec capacités agentiques
By
–
BREAKING 🚨: Comet browser got a major upgrade for its agentic capabilities.
— 🚨 AI News | TestingCatalog (@testingcatalog) 6 novembre 2025
– It can work longer and on more complex jobs across larger periods of time.
– It performs 23% better than its predecessor.
Longer context was a top request from all the users! Now it can write a… https://t.co/VsTPIoiok4 pic.twitter.com/bJPazyk69fBREAKING : Comet browser got a major upgrade for its agentic capabilities. – It can work longer and on more complex jobs across larger periods of time.
– It performs 23% better than its predecessor. Longer context was a top request from all the users! Now it can write a -

Kimi K2 Thinking: Breakthrough Open Source Model with 1T Parameters
By
–
Las cifras reportadas por el nuevo Kimi K2 Thinking son un barbaridad para un modelo open source! Además cuantizado a INT4 para mayor velocidad de inferencia y arquitectura tipo MoE. Eso sí, 1T de parámetros con 32B activos, no apto para mucho de nuestros equipos
