
Daily Digest
Morning digest: eight states sue OpenAI, Claude Code's sessions hub, and Anthropic's agent-scale git
Overnight: coordinated state AG suits and an FTC probe, Anthropic's reported Q3 numbers, a Claude Code desktop upgrade for parallel sessions, Scion (Anthropic's in-house VCS), a cheap knowledge-work model, and banks talking rogue agents on earnings calls.
Littlewolf Labs
Checked October 15, 2026. Each item links the page it came from.
Eight states sue OpenAI over ChatGPT and young users
On October 14, attorneys general in Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho, and Montana sued OpenAI in their own state courts under consumer-protection laws. The complaints say ChatGPT is marketed as safe for young users while an overly agreeable design can coach self-harm, give dangerous medical advice, and help users harm others. They want changes in how the product works for young users, monetary damages, and, in the Guardian's account, a halt on models that can improve themselves until OpenAI shows those models line up with human values. OpenAI spokesperson Drew Pusateri said the company has built "industry-leading age-specific protections."
If you ship a chatbot to anyone under 18, this is the week to read your age mode and your disclosures against what the complaint says you told parents.
Arizona Attorney General · The Guardian · NBC News
The FTC sent investigative demands to OpenAI and Anthropic
Axios, citing a person familiar, says the FTC on Wednesday sent civil investigative demands of more than 50 pages each to OpenAI and Anthropic. The demands cover security incidents, safeguards, marketing, and privacy policies. The same source says the agency plans to compel executives to testify. Demands to other AI organizations come later, that person said.
Anthropic's reported Q3 numbers, not confirmed by Anthropic
CNBC, citing people familiar, says Anthropic told investors third-quarter revenue stands at $16.8 billion, up from about $11.6 billion in the second quarter, with first-quarter revenue at $4.7 billion, and that Anthropic said it was profitable on an adjusted basis. The same story, citing the Financial Times, says the company told investors it generated $1.70 of revenue for every dollar spent on computing power. CNBC says an IPO is reportedly slated for November. Anthropic has not confirmed these figures.
Claude Code's desktop app is built for sessions side by side
On October 14, ClaudeDevs said the desktop app is quicker to start sessions, easier to split work across a few, and shows what each one is doing. The desktop quickstart says you pick Local, Cloud, or SSH, choose a folder, and can open parallel sessions from the sidebar, optionally each in its own git worktree. A tasks pane watches subagents and background commands. A side chat asks a question without leaving the main thread. The quickstart lists no new price.
ClaudeDevs on X · Desktop quickstart
Anthropic replaced Git with in-house Scion
Anthropic's engineering post says merges went from about 2,500 a day in May to more than 30,000 a day in September. A team of seven built Scion and Antique, the review tool, and migrated the monorepo in five months. CI does on the order of 25 million checkouts a day. Anthropic says the Scion footprint is 200 times smaller than its git mirrors in CPU and memory, checkouts are seconds rather than minutes, and CI gained 25% CPU capacity. Scion is not a suitable replacement for git in general use, and Anthropic says a public version that tried to please everyone would not be very useful. The ideas worth borrowing are sparse checkouts, post-merge testing, and a merge queue.
Paper Instruments prices Ultramarine 1 Flash as a cheap knowledge-work model
Paper Instruments calls Ultramarine 1 Flash Fable-class on its own benches for finance, law, consulting, and operations, and about 60 times cheaper than Fable. Those are vendor claims. The API price on the product page is $0.15 per million input tokens, $0.03 cached, and $0.50 output. Weights are listed as coming soon.
Ultramarine · Paper Instruments blog
Bank CEOs talked about rogue agents on earnings calls
Business Insider reports Bank of America's Brian Moynihan said the bank rushed codebase patches after powerful models, including Anthropic's Mythos, and hired 1,000 external consultants last quarter. On the same round of calls, Citigroup's Jane Fraser, Morgan Stanley's Ted Pick, Goldman Sachs' David Solomon, and JPMorgan's Jamie Dimon each stressed defenses against agents getting into the firm.
Handwired compiles ordinary JavaScript into an exact neural net
Cheng Lou's Handwired turns regular JavaScript into a neural net with no training, exact and differentiable, in the browser. The walkthrough turns a branch into ReLU arithmetic. One demo, Mini OS, is a small desktop written in JavaScript and transformed with Handwired.
Affiliate disclosure: some stories use affiliate or referral links. When a link pays a commission, the sentence next to it says so. Sponsored slots are labeled Sponsored. The channel card is a house ad until a sponsor is booked.
Newsletter
Newsletter coming soon.
Join the waitlist and we'll email you when the first issue launches.
By submitting you agree to our Privacy Policy.
You're on the list after you confirm. We'll email you when the newsletter launches. You can leave from any message. The form posts to Resend.