David Robinson, a leader on OpenAI's Safety Systems team, resigned and published an essay in The Atlantic on Saturday explaining why. He said he spent three and a half years at the company, helped draft its current Preparedness Framework, and oversaw safety reports for 12 frontier model launches. In his view, OpenAI's culture is broken. He argues the company thrives on trial and error, which it calls iterative deployment, and that this approach guarantees periodic failures as systems get more capable.

Robinson pointed to the Hugging Face breach and later rogue agent findings as evidence that the environment is no place to grow artificial minds that might not do what we want. He wants frontier labs to operate like nuclear power plants or busy airports, with layers of redundancy and careful planning. He also said he never met a colleague who had experience making airplanes fly safely or nuclear reactors run without melting down. TechCrunch notes his departure was first reported by Business Insider, and that Robinson hired a PR firm while insisting the decision to speak out was his alone.

OpenAI spokesperson Drew Pusateri said the company pauses training or holds back models when it needs to slow down, and that it is strengthening security in research and testing environments, expanding work with third-party evaluators, and improving real-time monitoring. The Guardian later reported Robinson's line that as the company sprints from one launch to the next, it is failing to achieve the level of care he believes is needed. He concluded that stronger incentives for safety coming from outside the company are a big part of getting this right.

OpenAI's reply lists pauses and stronger monitoring, which matches what it has said for weeks. What stuck with me is his line about being too busy sprinting to make big changes. If that is true inside a safety team, outside pressure is doing most of the work.

President Trump said Sunday that Director of National Intelligence Jay Clayton will lead a new White House group called the Super Intelligence Force, and will also serve as AI czar while keeping his intelligence post. Trump announced the move on Truth Social. POLITICO put the post midafternoon UK time. SecurityWeek, citing the Associated Press, timed the reporting around 10:57am ET Sunday, which was 3:57pm in the UK.

Trump prefers the phrase super intelligence to artificial intelligence. He said the force will coordinate federal engagement with consumers, public interest groups, religious organizations, critical infrastructure providers, and "Super Intelligence Companies." Joining Clayton are FTC Chair Andrew Ferguson, Under Secretary of Defense for Research and Engineering Emil Michael, and Office of Personnel Management Director Scott Kupor. The group will report to Trump and chief of staff Susie Wiles.

The announcement follows Tuesday's White House meeting where AI executives signed a voluntary safety accord. Clayton previously chaired the SEC and served as U.S. attorney for the Southern District of New York. NPR notes that in a CNBC interview last month he called super intelligence a national security issue and pushed back against a pause, saying other countries will keep developing the technology. Congress still has not passed a broad federal AI law.

Trump already got the big AI labs to sign a voluntary accord last week. Naming Clayton and standing up a Super Intelligence Force is the follow-through. Putting the spy chief in charge of a force that also includes the FTC chair and a Pentagon tech official tells you how this White House sees the problem. It is framed as competition and national security more than consumer harm.

Meta on Friday evening Pacific time introduced Muse Gadgets, an open-source project that lets developers build their own hardware connected to Muse, its personal AI agent. TechCrunch's report went up at 5:45pm PDT, which was 1:45am UK on Saturday, so it landed after Friday's Spritz. Meta is providing open-source firmware and a Linux SDK, plus starter ideas like a color e-ink display or a stick that plugs into a TV's HDMI port.

You can run it on a Raspberry Pi or an ESP32 board and wire Muse to displays, buttons, sensors, and actuators. Meta also set up a Discord channel for builders. Nat Friedman, head of product at Meta's Superintelligence Labs, said the company built a USB-C device called Muse Home Link that lets Muse talk to speakers and smart TVs on a home network. Meta made 5,000 of them and is giving them away free to Muse subscribers while supplies last, with shipping a few weeks out.

Meta already wants Muse in your phone and your small business tools. Now it wants Muse on a Raspberry Pi you soldered at the kitchen table. That is a weird and kind of charming move. The free Home Link gadgets will probably be gone already, but the open-source kit is the part that matters. Wire Muse to an e-ink panel or an HDMI stick and you get an always-on agent living in cheap hardware. I would like clearer rules on what that gadget can read on a home network. Apple spent Friday warning about Full Disk Access for a reason.

Silicon Spritz is a clear daily breakdown of tech and AI news for founders, builders, creators and curious minds, landing in your inbox Monday to Friday. The Long Pour is our occasional deep dive on one company. Subscribe to Silicon Spritz to get the next edition.