2026-10-08
By Vadym · Generated with AI, curated by me
Today was about AI showing up somewhere you can touch it — a new chip in your laptop, a watermark you can check yourself, a risk score on a hospital chart — while the same companies quietly built more tooling to catch what their own products enable.
OpenAI launched a new GPT-6 model alongside “Intelligent UI,” which generates interactive visuals mid-conversation — tappable buttons, custom calculators, editable charts and diagrams — instead of plain text. In a demo, asking how an airplane wing generates lift produced diagrams rather than a paragraph. It rolls out globally today for Pro, Plus, Business and Enterprise users, with free-tier Go users getting it Thursday.
Nous Research, maker of the open-source Hermes Agent, closed a $90 million Series B led by Robot Ventures, with Nvidia, Union Square Ventures, Menlo Ventures and Samsung also participating. The company says Hermes Agent has been cloned more than 24 million times and accounts for roughly 2.5% of global AI token usage. It's launching “Hermes for Businesses” to deploy customized agents for enterprise workflows, with revenue reportedly on track to pass $100 million by year-end.
Microsoft detailed pricing for its first Nvidia RTX Spark-powered PCs: a Surface Laptop Ultra starting at $2,600 (top configs hit $5,900, already sold out) and a $6,000 Surface RTX Spark Dev Box preloaded with VS Code, GitHub Copilot CLI and WSL. Both run models on-device with no cloud fees. A revamped Windows 11 adds “Execution Containers” to sandbox AI agents, which CEO Satya Nadella said will reach all Windows 11 users. Microsoft is offering up to $1,000 off for anyone trading in a MacBook Pro.
Google launched synthid.com, letting anyone check whether an image, video, or audio clip carries its SynthID watermark — previously limited to journalists and researchers testing it since Google I/O 2025. It covers formats from JPG and MP4 to WAV and FLAC, and is now built into the Gemini app and Chrome. Google says people make about 1 million verification requests a day; OpenAI, Nvidia, Kakao and reportedly soon Apple also support the standard.
Meta said it acted on 33.2 million pieces of child sexual exploitation content on Facebook and Instagram in the first half of 2026, with over 97% caught by its systems before any user reported it. Its new “signposting” detection model examines where an ad actually leads, not just what it shows, to catch ads that look harmless but route to illegal content — alongside a red-teaming AI agent that probes Meta's own defenses for weaknesses.
Healthleap raised $38 million ($8M seed co-led by Sequoia and First Round, $30M Series A led by Hummingbird Ventures) for a platform that reads hospital records each night, using language models to pull symptoms from clinical notes and combine them with labs and vitals into risk scores delivered each morning. It's deployed in 50+ hospitals, up from three a year ago, with revenue up more than 10x. At Penn Medicine, its malnutrition program is credited with $23.8 million in annualized financial impact.
Microsoft is selling hardware to run AI without the cloud, Nous Research just raised $90 million to put open-weight agents into more businesses, and Healthleap is proving an AI tool can produce a $23.8 million hospital outcome — the embedding phase of this technology is well underway. But Google making watermark verification public and Meta building a model just to catch where its own ads secretly lead are the same story from the other side: the more AI gets embedded into everyday products, the more infrastructure has to exist just to verify what it touched is real and where it actually sent you.
— Boba
Curated by Vadym