Aravind Srinivas just teased that the Comet browser will roll out shortcuts for repetitive tasks next week, alongside a fresh wave of user invites. Srinivas pitched the tool as a personal console for getting everyday work done. Access remains invite-only for now.
Google just detailed the latest updates to Gemini Code Assist, headlined by an Agent Mode that acts like an AI pair programmer. Instead of editing single files piecemeal, it analyzes your entire codebase to plan and execute multi-file refactors from a single prompt, complete with checkpoint rollbacks.
Mistral just rolled out a major update to its Le Chat chatbot, adding a deep research mode, multilingual reasoning, and advanced image editing. The real differentiator: it connects to enterprise on-premises data directly, letting businesses run deep analysis without shipping sensitive files to the cloud.
OpenAI just launched ChatGPT agent, enabling its chatbot to independently operate a browser, run code, and complete end-to-end tasks on its own virtual computer. It merges web interactions with deep research to deliver editable spreadsheets and slideshows, while keeping humans in the loop via browser takeovers. Why it matters: It shifts ChatGPT from a passive advisor to an active digital worker.
Windsurf just shipped its Wave 11 update, introducing voice input for Cascade alongside named checkpoints to revert conversation states and @-mentions to reference past chats. Cascade can now autonomously capture DOM trees, console logs, and screenshots directly from the built-in browser. Why it matters: Coding agents are rapidly shifting from simple text editors into hands-off end-to-end builders.
Google just brought Gemini 2.5 Pro and a new Deep Search feature into Search's AI Mode for U.S. Google AI Pro and Ultra subscribers. Deep Search can run hundreds of queries and compile cited research reports in minutes. Search is also rolling out an agentic feature to all U.S. users that actually calls local businesses to check prices and availability on your behalf.
Google DeepMind and Project Zero’s AI agent Big Sleep just uncovered a critical SQLite flaw (CVE-2025-6965) before attackers could weaponize it. The team used the agent to isolate the zero-day threat ahead of time. Google claims this marks the first time an AI agent directly thwarted an in-the-wild exploit operation.
Nextdoor just overhauled its local social app with a major redesign centered around Alerts, News, and Faves. The big addition is an AI agent trained on 14 years of neighborhood chat history to dish out summarized local recommendations. It is currently rolling out in select U.S. markets.
Former OpenAI engineer Calvin French-Owen just dropped a candid look inside the AI giant, detailing how a tiny 17-person team pulled seven sleepless weeks to launch the Codex coding agent. The company tripled to 3,000 employees in one year, running largely on Slack with messy monolithic code and early-Facebook 'move fast' energy.
Google Cloud just teamed up with MLB Statcast to predict where home run balls will land in the stands during the All-Star Game. Powered by Gemini 2.5 Pro, the system crunches batter history, stadium spray charts, and live weather to flash seat-specific alerts on ballpark screens. It is a neat showcase of agentic AI turning massive datasets into live fan engagement.
Anthropic just rolled out Claude for Financial Services, a dedicated solution tying its models to live market data from S&P Global, FactSet, and Databricks. It lets analysts run complex simulations and due diligence with direct citations back to source documents, and client data won't be used for model training.
Anthropic just launched a new directory that connects Claude directly to popular tools like Notion, Canva, Figma, and Stripe with a single click. Instead of copying and pasting context over and over, Claude can now pull live project data to draft release notes or turn designs into code. The directory is live for everyone, though remote connectors require a paid plan.
Kiro just launched a new AI IDE built to take developers beyond quick vibe coding and into production. Based on Code OSS, the editor turns prompts into full specifications, architecture docs, and sequenced tasks, while event-driven agent hooks handle test and documentation updates in the background. Why it matters: Prototyping with AI is easy, but maintaining scalable code in production is where the hard work starts.
Cognition Labs just acquired Windsurf's remaining business, IP, and 240+ employees, just 72 hours after Google paid $2.4 billion to hire away its founders. Cognition inherits 350 enterprise accounts generating $82M in ARR, with plans to embed its autonomous Devin agent straight into the Windsurf IDE.
California grid operator CAISO is piloting Genie, a generative AI tool from OATI designed to automate power outage and maintenance reports. Instead of engineers manually sifting through hundreds of repair notes to model supply impacts, AI now does the heavy lifting. If successful, CAISO plans to test autonomous grid-level decision making.
Windsurf CEO Varun Mohan and key team members just joined Google DeepMind to bolster its agentic coding efforts. The move comes right after a reported $3 billion acquisition by OpenAI fell through due to complex Microsoft partnership constraints, giving Google a major win in developer AI.
Moonshot AI just open-sourced Kimi K2, a 1-trillion parameter MoE model with 32B activated parameters engineered specifically for coding and autonomous agent workflows. It scored 53.7% on LiveCodeBench, comfortably beating DeepSeek-V3 and GPT-4.1, powered by a new MuonClip optimizer that eliminates training instability. Why it matters: High-end agentic intelligence is rapidly shifting from pricey closed APIs to open, dirt-cheap alternatives.
Mistral AI just teamed up with All Hands AI to roll out the Devstral models built specifically for coding agents. The 24B Devstral Small 1.1 is fully open under Apache 2.0 and hits 53.6% on SWE-Bench Verified, while Devstral Medium reaches 61.6% on the benchmark via API at roughly a quarter of the price of frontier competitors.
Intel's new CEO Lip-Bu Tan reportedly conceded to employees that the company has fallen out of the top 10 semiconductor makers and that it is "too late" to catch Nvidia in AI model training. Facing massive losses and layoffs, Intel plans to abandon the training race and redirect its focus to edge AI on PCs and autonomous agents.
xAI just launched Grok 4, a reasoning-focused model handling 256,000 tokens of context. It crushed over 44% of the brutal Humanity’s Last Exam benchmark with zero external search tools, easily outpacing OpenAI's Deep Research. Musk is also offering a multi-agent Grok 4 Heavy tier for a hefty $300 a month.
