AI — Page 2
Models, agents, infra, applied AI.
- Ricursive Intelligence founders will discuss AI-designed chips at Disrupt 2026
Anna Goldie and Azalia Mirhoseini will discuss Ricursive's AI software for chip design and how it could learn across design cycles.
- ElevenLabs CEO says businesses should disclose when AI answers calls
Mati Staniszewski says ElevenLabs is pacing at $600M in ARR as its voice agents move from creative tools into customer support.
- Nscale's UK AI supercomputer may wait until the 2030s for power
A grid connection could push the Loughton site years past its planned 2027 debut, putting Josh Payne's infrastructure bet up against a constraint money cannot quickly buy away.
- BAND connects local coding agents to its network through Docker Sandboxes
Arick Goomanovsky is extending his bet on shared agent infrastructure into local code execution; BAND's setup guide still documents a compatibility workaround for its published kit.
- Google moves Gemini 4 into post-training, testing it inside Antigravity
Google DeepMind chief Koray Kavukcuoglu wants to release an early model output soon, but gave no date as Google shifts attention from Gemini 3.5 Pro.
- SpaceXAI may be testing Grok music generation inside Imagine
An X user posted an early Android preview; SpaceXAI's published Imagine materials describe image and video generation with audio, not a standalone music tool.
- Veeva says its new AI agent can build a clinical study in a day
Peter Gassner built Veeva around life sciences software; its new agent will configure EDC and data-quality systems from study protocols, with early access planned for December.
- I used my AI newsroom playbook to finish top ten in Google's four-week hacking challenge on opening night
A live leaderboard screenshot showed my merket account in ninth place with all nine numbered challenges solved; the collaboration produced a reported top-10 finish, not a confirmed final placement.
- Higgsfield says its four-week revenue pace tops $1B annualized
CEO Alex Mashrabov says the calculation multiplies the latest four weeks of revenue by 13, while business revenue under contract has grown tenfold since June.
- TypeSafe's Jev Router picks models for free, with a claimed million-token window
Listed on OpenRouter on September 25th, the router uses TypeSafe's Jev decision model to choose a model and reasoning effort for each request.
- Pirate Wires says AI 'doomer' PR firm sought interviews as ex-Anthropic researcher denied outside help
Pirate Wires reports the firm was booking Jacob Coxon for an interview on September 9th, after he had already made his warning public the night before.
- Paper Instruments teaches agents to edit Office files without dismantling them
Paper Instruments released three Python packages for AI agents editing existing Word, PowerPoint and Excel files while preserving their structure.
- OpenAI says Hugging Face remains its most severe agent incident
Sam Altman says the review spans petabytes of agent activity logs, with OpenAI prioritizing cases by severity and adding resources.
- Anthropic says Claude can calculate nine-loop particle-scattering formulas
The post points to theoretical physics, where loop order measures successive corrections; the number alone does not show what Claude calculated or how it was checked.
- Vapi says its voice agents now handle about a billion calls a year
The September claim follows Vapi's May Series B, when the company reported more than 1 billion calls made and TechCrunch reported Amazon Ring routed all inbound calls through it.
- DaVoice sues Perplexity over wake-word technology after the companies collaborated
Founder Ofer Fryman's startup says Perplexity misused trade secrets; a heavily redacted complaint leaves the alleged copying and the product at issue unclear.
- OpenAI Codex adds client plumbing for agents to share a remote message board
A merged GitHub change adds an HTTP client and live notifications; the PR does not include the message-board service itself.
- Anthropic spotlights Opus 5.5 demos, with a human still in the loop
Three days after launch, Claude's account shared user experiments; one interactive lens lesson came from a designer using his own project files and a maximum-effort run.
- Cognition's revenue run rate reaches $1B, four months after $492M
Bloomberg attributes the September figure to a person familiar with the matter; founder Scott Wu has built Cognition around software agents that take on longer engineering tasks.
- Google adds avatars to Gemini agents, with Cox building Autotrader's car-shopping assistant
Gemini 3.8 Live with Live Avatar is generally available in Gemini Enterprise. Google Cloud says Cox Automotive built an AI shopping assistant for Autotrader using the feature.
- Musk says SpaceXAI may more than double Colossus 2's Nvidia chips by year-end
Bloomberg reports a December 31st target but gives no current or planned chip count for the Memphis-area cluster.
- Oracle invokes force majeure as power delays test its New Mexico AI campus
The roughly one-year delay pushes back returns on Blue Owl's reported $3B equity stake and gives lenders a live test of who bears infrastructure risk.
- Cerebras says its rehearsed AI agent booked dinner in 22 seconds
Cerebras reports a 22-second median across two successful runs using Qwen 3.8 27B, its inference hardware, parallel checks and a saved website-navigation skill.
- Higgsfield estimates a 15-second ad-localization workflow at $13.15
Higgsfield's example totals 263 credits for a 15-second workflow; video steps are listed at 720p, while Audio's resolution depends on input and rates vary by plan.