Signal Flow
Scan the day in a cleaner single-line news stream.
No articles match this filter.
Show all
Autonomy
OpenAI Can't Tell You Which Images Leaked. It Also Can't Explain How It Knows They Were Yours.
OpenAI says 53 user-uploaded images were posted to an external site by its AI agents. It says it can't notify affected users because its privacy architecture disconnects…
Autonomy
OpenAI Is About to Sell a Cybersecurity Model Built by the Model It Labeled “Critical”
OpenAI is preparing to preview GPT-6 Cyber, its fourth cybersecurity-specific model this year. Its predecessor, GPT-6 Astra, was classified “Critical” — the first OpenAI…
Models
Google Gave Its Agent a Face. What It's Selling Is the Silence It Fills.
Gemini 3.8 Live with Live Avatar is not a video generator with a chat feature. It's a conversation that keeps going while the back end does the work. The face is how Goog…
Models
Anthropic Found a Question, Not a Cure. Wall Street Sold the Answer.
Anthropic's AI agents discovered a new enzyme system after 21.5 hours of autonomous work. Gene editing stocks fell double digits. But the paper has no peer review, the en…
Hardware
Zuckerberg Pulled a Keychain Out of His Pocket. It's a Door for Muse.
Meta's Muse Charm is not a phone replacement. It can't make calls. It has one job: press a fingerprint, talk to Muse. The design language is a pet, not a tool. That's why…
Business
Anthropic's "Slow Down" Essay Came After the Release Debate Started.
Dario Amodei published a call to pace the frontier on September 12. Reuters reports that Anthropic began discussing a new model release before that essay was written. The…
Applications
Meta Muse Topped the App Store. Amazon Blocked It on Day One.
Muse became the No. 1 free app in the US ten days after launch, beating ChatGPT in downloads. The same weekend, Amazon blocked it from shopping. And Muse still can't expl…
Models
Xiaomi Live-Streamed Its RL Training. The Bill Was the Point.
MiMo V2.6 Pro tops the open-weights leaderboard at 46 on the AA index. But the more unusual fact is that Xiaomi published a live training dashboard showing the cost, the…
Autonomy
The UN Says AI's Safety Model Is Unravelling. But It Recommends Aviation — Which Has Powers the UN Doesn't.
The UN's AI panel assessed the Hugging Face breach and concluded that “the traditional model of safeguarding is unravelling.” It recommends aviation and nuclear safety as…
Models
Anthropic Routed Your Requests to an Unreleased Model. You Weren't Supposed to Notice.
Anthropic silently redirected some Opus 5 API calls to an unreleased model called Opus-Next. When it pulled the plug, developers were mid-project on a model they didn't k…
Models
Kimi Posted π Without the 3.1. The Subscriptions Came Back First.
Moonshot's own account teased K3.1 by deleting the first three digits of pi. The leaker who predicted August was wrong. The leaker who predicted October may be wrong too.…
Business
Anthropic Built a Biology Lab. It Also Warned AI Could Enable Bioweapons.
Anthropic operates a wet lab in the Bay Area. It will not say what it is for, beyond “not drug discovery.” The same month, it disclosed five cases where its models could…