Google Labs launched “Play with Putty,” a waitlisted collaborative vibe-coding tool for building sites and tools together in real time
OpenAI unveiled Defense Factory, an automated system to continuously find, validate, and fix vulnerabilities as open-weight models close the gap with attackers
GPT-6 Astra scored 63% on ARC-AGI-3, beating the human baseline on action efficiency and using fewer moves than the median human on 96% of levels
OpenAI officially announced GPT-6 Astra, pre-trained on 100,000+ GPUs at Stargate, rolling out to all ChatGPT plans including Plus over the next few days
Martian launched AI Frontier, a dashboard comparing 44 LLMs on quality, cost, and consistency, with reliability scores ranging 0.74–0.96
Lindy launched CC-based scheduling, letting users loop the assistant into email threads to read context, book meetings, or reply with open slots
OpenAI is prepping a major image model upgrade with higher quality, faster generation, and smarter tools, after being spotted testing a new model on Image Arena
Institute of Foundation Models released K2 Horizon, six open models from 0.9B to 375B parameters, all under Apache 2.0
Google DeepMind and Google Research launched WeatherNext 3, a weather AI model using real-time satellite data to generate hourly high-res forecasts
OpenRouter added video model benchmarks, with Seedance 2.5 leading on physics/prompt adherence and MiniMax H3 Max leading on speed and cost
Google rolled out new Gemini voice features letting users search Gmail, manage Keep, and create Docs conversationally for AI subscribers
Grok Bot for Enterprise launchedtoday, free for all Grok and Cursor enterprise customers for two weeks
A new family-focused AI assistant seeks access to users daily life details but says it wont use data for AI training or share it with third parties
