AI news for builders and product teamsUpdated Nov 17, 2026, 16:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Ramp data shows corporate AI token spending grew just 1.5% across August and September, down from 42% month-over-month growth in July, even as token volumes kept rising. The slowdown comes as Anthropic prepares for an IPO and follows API price cuts at OpenAI and Anthropic.

Databricks has published guidance for public sector organizations on governing AI at scale, framed around a mandate to adopt AI quickly while managing risk.

In a Latent Space interview, Fabian Theis argues that perturbation prediction on existing datasets is too easy to justify complex models, and that the field should instead model biology across scales, from molecules to organisms. The piece ties this to a $1.8B Biohub data effort backed by Biohub, NIH, DOE, Google and Meta.

Amazon launched a text-based Alexa shopping assistant for select U.S. customers, letting them search, add to cart, and purchase via SMS. The feature follows Amazon blocking Meta's Muse from its site in September and rolls out broadly in coming weeks.

METR audited Anthropic's internal research over eleven weeks and concluded model development pace has roughly doubled, with about 80 percent of the speedup attributed to AI-assisted research, possibly crossing Anthropic's own warning threshold in July.

Underdog released Woof, a 31-million-parameter text-to-speech model about 32MB in size that runs in a browser without an internet connection, and shipped it inside a small web game called Woof.

Physicl, founded by Alex de Vigan, has raised $10 million in seed funding led by Inovia Capital to build realistic 3D simulation tooling that helps robot and world-model developers close the sim-to-real gap. It is launching Studio, a product with 10,000 3D environments and 100,000 assets for generating synthetic training data.

Around 100 protesters from the Geef Tegengas group broke into a Microsoft data center construction site in Amsterdam, where some scaled a crane to hang a banner. The action targets plans for three 85-meter towers on the city's outskirts.

Thinking Machines released Inkling 2 Small, a 594B-total/17B-active multimodal mixture-of-experts model scoring 31 on the Artificial Analysis Intelligence Index, available now on Tinker with Apache 2.0 weights promised next week after external safety testing.

AWS's Continuum for code vulnerabilities is now generally available in US East (N. Virginia) and US West (Oregon), four months after its gated preview. It validates findings by running exploits in a sandbox and can deliver fixes as pull requests, priced per line of code plus per inference unit.

Tau Robotics, a San Francisco startup, offers humanoid-robot apartment cleaning at $60 per weekly two-robot visit or $100 one-time, with the robots currently teleoperated by humans via head- and wrist-mounted cameras while collecting training data for eventual autonomy.

OpenAI has partnered with Hacktron to hunt for vulnerabilities in its systems and to test how its products connect with third-party services.

Palo Alto Networks is in talks to buy Clutch Security, a startup securing AI agents' access to enterprise systems, for $500 million in cash plus milestone payouts—roughly 50 times Clutch's near-$10 million ARR. Neither company commented.

Databricks outlines a real-time fraud detection approach built on Feature Views, splitting signal into three feature types: what a customer did, and related behavioral and contextual data.

Perplexity has launched always-on "Teammates," work-focused AI agents that can take on multiple roles in a workspace and pull projects, skills and memory from existing Perplexity tools. They build on Perplexity Computer, which deploys agents on a user's device.

Google clarified that K–12 schools using its apps can switch off Gemini while leaving Search on for student accounts, but any opt-in to Search now routes through AI features—there is no AI-free web search option.

James Gates of North Carolina sued OpenAI and Sam Altman in San Francisco County Superior Court, alleging ChatGPT helped his wife Cheryl plan to poison him over months in 2025; she was indicted in October on attempted first-degree murder and food contamination charges.

AWS has made Web Answer generally available on Amazon Bedrock AgentCore, a managed AgentCore Gateway tool that takes a natural-language question and returns a cited answer in a single call, priced at $10 per 1,000 tool calls in us-east-1.

OpenAI has released GPT-6 Cyber, a model trained to find software vulnerabilities, build working exploits and propose patches. OpenAI says it found 81 previously unknown Firefox flaws, all now fixed by Mozilla, and produced 948 working exploits in the Linux kernel, more than five times its predecessor.

LangChain has added async subagents to Deep Agents for Python and JavaScript, letting an orchestrator agent keep talking to the user while background subagents run long tasks and report their results back. Turning a subagent async in Managed Deep Agents takes a single async field.