Databento raises $100M at a $1.1B valuation, selling data to AI labs
The market data API company says five of the six largest frontier AI labs now have enterprise contracts with it.
762 stories
The market data API company says five of the six largest frontier AI labs now have enterprise contracts with it.
Four nonprofits asked a San Francisco court to bar five AI developers from building models more capable than their best as of November 12.

Altimeter and Atreides, with about $2.13 billion in Nvidia stock, want a slow CEO search, and a petition to reinstate Huang has passed 115,000 signatures.

Copilot Chat drafts and answers, Copilot Cowork runs multistep tasks in Microsoft 365 for a license plus usage fees, and Anthropic's Claude Cowork comes with Claude plans from $20 a month.

Named after Kahneman's fast thinking, the models behind TypeSafe's Jev and Google's Gemini Decision 1 return probabilities over fixed options, and the SO-Arena benchmark shows where they fall short.

In Anthropic's coding test, high solved 2.5 points more tasks than the default medium for about 38% more per task. Here is when each level is worth it and how to switch.
The tools from DeepSeek's consumer app now run server-side on V4.1-Flash, and DeepSeek also sells standalone Search, Fetch, OCR, speech and voice APIs.
The beta for Team and Enterprise plans takes notes, answers questions and builds docs during calls, and saves a transcript but no audio.
Claude Fable 5, Claude Haiku 4.5, Gemini 3.8 Flash, Grok 4.6 and Kimi K3 also leave every Copilot surface, and admins may need to enable the replacements.
CapitalG led the Series C, and LangChain says more than 7,000 organizations now use its LangSmith agent platform.

The new Gemini 4 model is built for everyday coding and agents, runs in Search and the Gemini app, and Google says it edges out Gemini 4 Argon on a key index.

The longtime CFO runs Nvidia until the board picks a permanent chief, and her first test is the November 17 earnings call.
Pro subscribers in the US can let agents check out with a network token in place of the card number, approving each purchase with Face ID or a fingerprint.
The government will legislate six duties for frontier AI companies, including mandatory incident reporting, after an OpenAI agent breached a government health site.
Open the Meta AI app or meta.ai, tap Ask Meta AI and pick a mode. What Meta's help center says about where it lives, what it does, its limits and privacy controls.

Direct and indirect injection, why AI agents raise the stakes, and the defenses from OWASP, the big labs and new research such as SPILLWAY.
It stores the keys and values of every token already read, so the model does not recompute them. Why it grows with context, and how paging and compression shrink it.
Where SKILL.md goes, which frontmatter fields matter, how to run a skill with /skill-name, and what to check when it doesn't trigger.
Put a folder with a SKILL.md in .agents/skills for one project, or ~/.gemini/config/skills for all of them. Paths, format and fixes from Google's docs.

The new layout hides Git, diffs and terminals and runs the same agent for research, documents and data analysis.
Reuters reports Anthropic sent bankers a filing with up to $69.3 billion in proceeds, then said the pricing details were wrong, as the IPO moves ahead.

The "don't ask again" prompt can save passwords into a settings file that gets committed. Anthropic says it is looking into it; here is how to check, rotate and stop it.

Free does not include Fable 5.2. Pro gets it free for a launch week, Max and Ultra include it, and the API costs $8 and $40 per million tokens.

Fable 5.2 beats Opus 5.5 on 14 of 15 rows in Anthropic's table at twice the price, yet Anthropic still tells most users to start with Opus 5.5.