In partnership with
Good morning. It's Monday, October 5th.
Trump has launched a "Super Intelligence Force," a task force chaired by Jay Clayton that has 120 days to deliver a report on AI risks and opportunities, and its charter warns against overregulation that would stifle innovation.
Elon showed solidarity in the renaming of Artificial Intelligence to Super Intelligence, though some speculate that it was just another opportunity to mess with OpenAI.
-Jeff
AI Breakfast
The AI notetaker that gets the hard words right
Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.
Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.
Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.
Trump answered the slowdown crowd with a Super Intelligence Force
On Saturday Trump stood up a task force called the Super Intelligence Force and put Jay Clayton, the sitting director of national intelligence, in the chair. It has 120 days to deliver a report on the risks and opportunities of advanced AI. The mandate language is the tell: develop plans for responding to "SI-enabled threats to our society, while preventing overregulation and regulatory capture that would stifle innovation and competition."
The vice chairs are FTC chair Andrew Ferguson, OPM director Scott Kupor, and Emil Michael, the Pentagon's undersecretary for research and engineering. Condoleezza Rice and former AI czar David Sacks sit as outside advisers. The whole thing runs on the vocabulary from last month's executive order telling agencies to write "super intelligence" instead of "artificial intelligence."
Look again at the first vice chair. Ferguson's FTC is reportedly drafting civil investigative demands for OpenAI, Anthropic and METR over autonomous-agent risk. He now also sits on the body charged with preventing overregulation. That is not a contradiction so much as a signal about which of the two jobs is the priority.
Clayton has been blunt about the slowdown proposals now circulating: "I don't think any American should think that that's a good strategy." Four months of study, with the conclusion pre-announced.
OpenAI's safety report writer quit and took it to The Atlantic
David Robinson, who led the writing of the safety reports that ship alongside OpenAI's major launches and was among the company's longest-tenured employees at three and a half years, resigned and published an essay in The Atlantic saying the "culture is broken." His argument is structural, not personal: OpenAI "has thrived by trial and error (which it calls 'iterative deployment')," an approach that "guarantees periodic failures."
His fix is that frontier labs should run "like nuclear-power plants or busy airports, with layers of redundancy and careful, time-consuming planning." The line that should sting: in three and a half years he "never encountered a colleague who had experience making airplanes fly safely or nuclear reactors run without melting down." OpenAI spokesperson Drew Pusateri said the company is "making sure our models don't become more capable than we can safely manage," pointing to security work and the paused training runs.
A day later Altman sat for Politico's Decoded and drew the line himself. On Dario Amodei: "I think there's a lot of daylight [between us]." On the trade: he would not take a deal where "we'll make sure there's no major hacks... zero... bad things." On concentrated control by a single lab in San Francisco: a "completely unacceptable trade-off." He wants lighter-touch regulation with targeted safeguards reserved for catastrophic risk.
Robinson's exit follows safety head Johannes Heidecke's departure, and Jacob Coxon, who worked at both OpenAI and Anthropic before resigning, now says the companies are "gambling with our lives." Two frontier labs, two public philosophies, and the people who wrote the safety documents keep leaving.
OpenAI is putting ads inside image generation
OpenAI announced this morning a new visual ad format and said it will start testing it during image generation in ChatGPT, in the US, later this month. The ads show "product inspiration, product usage, or the experiences they make possible," labeled and kept separate from user content. The audience OpenAI cites for the surface: 1.2 billion people each week.
The rest of the post is plumbing, which is the real news. Data-platform integrations with Hightouch, Tealium and LiveRamp. Attribution partners including AppsFlyer, Adjust, Branch, Kochava and Triple Whale. Geo-based causal experiments with Haus, Measured and WorkMagic. Brand-suitability pilots with DoubleVerify and Integral Ad Science, which OpenAI says grade the ad environment without reading private conversations. One performance number made the post: WeightWatchers' attributed cost per acquisition on ChatGPT Ads came in 15.3% below its blended paid-search benchmark.
OpenAI's guardrail sentence is "advertising does not influence the answers ChatGPT provides." Image generation is a clever place to test that claim, because a product photo is native to the output rather than bolted onto an answer. No pricing, and no user opt-out, was published.
Nvidia added a 64GB DGX Spark at $4,999, shipping October 23, with the same GB10 superchip as the original and half the unified memory. Two units cluster to 128GB over ConnectX-7. The catch is the other half of the announcement: the 128GB model moved up to $6,950 from $3,999 at launch, which is memory pricing showing up in a desktop box rather than a data center invoice.
Cloudflare released Clef and Clef-flash on Workers AI, open-weight "decision models" at 27B and 9B that return typed probabilities instead of prose. They read text, JSON, images and video across a 65,536-token context at $0.24 and $0.09 per million input tokens. The pitch is the routing and gating layer in an agent loop, where you want a calibrated number and not a paragraph explaining one.
Claude Code shipped 2.1.288 and 2.1.289, following last week's Claude Mods launch. New for mods: $.ui.selection() returns the selected text in fullscreen mode, agent.spawn works for teammates, and a failing mod client now raises ui.fault instead of taking the surrounding UI with it. Elsewhere: --max-findings on /code-review, Ctrl+F session-name search, Ctrl+C prompt recovery via the up arrow, and long conversations now auto-compact instead of failing with "Prompt is too long."
xAI retired grok-voice-transcribe-1.0 on October 2. Requests route to 2.0 automatically at identical pricing, which xAI says is more accurate. If you pinned the version string, nothing breaks, but you are now on a different model than your evals measured.
Thank you for reading today's edition.
Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.
Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!
Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months.
