GPT-6 Shipped the Week a Safety Leader Quit -- AI Brief October 8

Good day %%first_name%%. GPT-6 is live in ChatGPT with a more visual interface, in the same week one of OpenAI’s safety leaders explained why he quit. Anthropic shipped a much cheaper Haiku, Grok is about to ask Claude for help, and one developer used Claude to clone seven Adobe apps. Let’s get into it.

ChatGPT Gets GPT-6 and Interactive Visuals TechCrunch

* What happened: OpenAI began rolling out GPT-6 in ChatGPT on Wednesday alongside Intelligent UI, a feature that lets replies include tappable buttons, task-specific calculators, interactive charts and editable graphs. Pro, Plus, Business and Enterprise users get it first. Free and Go users get it Thursday, October 8.

* Why it matters: Until now ChatGPT has mostly answered in paragraphs. OpenAI product manager Aarush Selvan said “ChatGPT has predominantly been a text-based interface,” and demoed it explaining how an airplane wing generates lift with diagrams, plus a personal savings calculator. If you’ve ever asked for a recipe or a hiking route and got a wall of text back, this is aimed at you.

* What everyone’s saying: Wired’s headline calls the new ChatGPT “more show than tell.” On Hacker News, an early commenter asked whether this is OpenAI catching up with Anthropic’s Artifacts, pointing to OpenAI’s line that GPT-6 was trained to compose responses from text, visuals and interactive elements rather than relying on a harness bolted on afterward.

* My read between the lines: The telling detail is that users can ask for fewer visuals. When a company ships a volume knob for its headline feature, it’s guessing some of us will want the paragraph back. Still, the bet is clear: the chat window is becoming a place where you do things, not just read answers.

📖 Further reading: OpenAI Made Its New Agent Adorable. That's the Part That Worries Me. — my paid look at how OpenAI’s friendlier, more visual interfaces change what we hand over to a chatbot.

A chatbot that answers in widgets is still a chatbot. Viktor is the other kind: it lives in your team’s Slack, connects to 3,000+ of your tools, and does the work you can describe, like the weekly report, the dashboard, the code or the campaign. Not a chatbot. A coworker. New readers get $50 off their first month. Hire Viktor →

Ex-OpenAI Safety Leader Calls the Industry Unsafe The New York Times

* What happened: David Robinson, who led safety transparency efforts at OpenAI, gave his first interview since quitting, telling Ezra Klein on Wednesday: “we keep publishing these warnings, but ultimately, we’re still training and deploying these dangerous models that we’re warning about.” He left last week in an essay that TechCrunch covered, saying the company’s “culture is broken.”

* Why it matters: Robinson isn’t the usual doomer. Klein describes him as a Rhodes scholar and Yale-trained lawyer who advised the Biden White House, and who thought the extinction worries were “kind of nuts” when he joined. Three years later he says the problem isn’t only OpenAI: he calls the whole industry’s culture unsafe.

* What everyone’s saying: Business Insider notes hundreds of OpenAI employees signed a July open letter, “Pacing the Frontier,” urging regulation to slow AI down, and that Sam Altman said in September, “We need to pace the frontier.” Inside the office, Robinson says, activity stayed “frenetic” and people were “running on fumes.” An OpenAI spokesperson said the company pauses training or holds back models “when we need to slow down.”

* My read between the lines: The most useful line isn’t the alarm, it’s the diagnosis. Robinson says he and his colleagues were “so busy sprinting” they seldom had time to consider big changes, so the push for safety has to come from outside the company. An industry too busy to think about its brakes won’t schedule the time. He also admits hiring a PR firm, the whistleblower playbook by his own telling, while insisting speaking out was his decision alone.

📖 Further reading: The US Government Just Took Anthropic's Best AI Model Offline — Here's Why — my paid piece on a rare time outside pressure actually pulled a frontier model, the kind of external brake Robinson is asking for.

The daily Brief is free and always will be. When a headline deserves more than four bullets, I write it up at length for members: the paywalled deep-dives behind the news, plus the full archive. Become a member →

Anthropic Ships Claude Haiku 5.5 Anthropic

* What happened: Anthropic launched Claude Haiku 5.5 on Wednesday, its fastest and most capable small model, aimed at high-volume wo