Signal Flow
Scan the day in a cleaner single-line news stream.
No articles match this filter.
Show all
Autonomy
OpenAI Says It Doesn't Need an Antitrust Waiver. That's the Tell.
OpenAI's policy chief told Washington the company needs no antitrust permission to coordinate safety with Anthropic and Google. The FTC chair, speaking the same day, said…
Applications
DeepSeek's Desktop App Doesn't Open a Port. That Matters.
DeepSeek didn't announce a Harness desktop client. It merged the code into the main branch, signed it, notarized it, and left the download URL pointing at its own domain.…
Business
OpenAI Bought Glass Imaging. It's Really Buying an Apple Exit.
Glass Imaging's founders built Portrait Mode inside Apple, then left because they concluded phone cameras had hit a hard ceiling. OpenAI paid $300 million for that conclu…
Autonomy
Musk Wants Rivals to Grade Each Other's Homework. The Models Already Cheat the Tests.
Elon Musk proposed that AI companies cross-test each other's models before release, calling it “the best thing we can do for safety.” The UK's AI Security Institute teste…
Autonomy
Nvidia's Jensen Huang Says AI Safety Is an Engineering Problem. He Doesn't Ship the Systems That Fail.
Huang told Salesforce's annual event that market forces and existing law are enough, and that safety is “job one” but ultimately an engineering problem. He praised Anthro…
Models
Google Split Its Voice Line in Two. The Cheap One Loses to GPT-Live.
Google's leaderboard-topping voice model is the expensive one. The model most developers will deploy scores 76.0 — below both configurations of GPT-Live-1. Google didn't…
Business
Google Moved Its Safety Team Into PR. Then Two Safety Researchers Quit.
Google restructured DeepMind's safety team in early September, moving 90 people from engineering into global affairs. Within days, two safety researchers resigned. The ti…
Models
85% Replaced Isn't a Layoff. It's RSI Showing Up in the Logs.
Anthropic's internal report says an unreleased model will replace 85% of its research team. On the same day a Google DeepMind researcher quit saying AI might kill everyon…
Applications
Apple Metered Siri AI. OpenAI and Google Don't.
Apple is the first major platform to treat AI as a metered utility. OpenAI and Google sell “unlimited.” Apple's footnotes say “daily limits, expanded access for a fee.” T…
Models
Anthropic Is Building the Pipes, Not the Model. That's the Whole Strategy.
Claude for Financial Advisors isn't a chatbot for advisors. It's a connector layer that routes client data through Anthropic's model. The distinction matters because it e…
Models
DeepSeek Learned to Read Aloud. It Hasn't Learned to Talk.
DeepSeek's new voice feature is being described as “AI voice conversation.” The setting is called “Read Aloud Voices.” That naming gap is the whole story: DeepSeek closed…
Business
Three Labs Are Writing AI Safety Rules. Antitrust Law Says They Can't.
Anthropic, OpenAI, and Google have been quietly negotiating an industry-run standards body since July. The talks started before Amodei's public letter, and they face a le…