Schmidt Gets Booed, Neumann Gets Eulogized, China Gets Power

The Eulogy Nobody Wanted to Write Yet

Peter Neumann died at 93, and with him goes roughly half a century of institutional memory about why computers were never actually safe. He ran the Risks Digest before risk was a content category. He told vendors their products were structurally insecure while they were still printing brochures claiming otherwise. He was usually right and almost never heeded.

Losing him in 2026 is bad timing in the way funerals are always bad timing. The industry he scolded is now wiring large language models into drive-thrus, defense procurement, and search itself. The people who knew where the bodies were buried are themselves being buried, and the replacements are mostly product managers.

The Guardian ran a thoughtful piece this week on learning to embrace uncertainty in a modern age that refuses to resolve. It reads differently next to Neumann's obituary. He spent his career arguing that uncertainty in computing was a choice, not a condition, and that the choice had been made badly on purpose.

The Crowd Has Notes

Eric Schmidt went to the University of Arizona to deliver the standard commencement gospel about AI opening doors, expanding human potential, and other phrases that test well in McKinsey decks. The students booed him into silence. This is worth sitting with. A former Google CEO, on a stage built specifically to flatter him, could not get a college audience to clap for AI optimism.

Pair that with The Verge's piece on fast food chatbots metastasizing from McDonald's drive-thru into broader customer service. The students booing Schmidt are the same demographic about to discover that the entry-level service jobs that funded their degrees are being quietly rewritten into prompt-handling rounding errors. They are not confused. They are doing the math.

Apple, sensing the mood, is leaning into privacy theatre. Siri now auto-deletes things. The pitch is that Apple is the trustworthy alternative in a field of data gluttons. The pitch is also that auto-delete is a feature rather than a UX shrug. Both can be true. Neither addresses the structural question of whether a chatbot built on harvested context can be private in any meaningful sense.

Power, Literally

The most important story in the slot is the least dramatic. Larry Summers, a man with no particular incentive to spook markets, said the quiet thing about US AI competitiveness. The constraint is electricity. China builds generation capacity the way Americans used to build highways. The US permits a substation over six years and calls it infrastructure week.

If the binding input to frontier AI is power rather than silicon, the export controls and chip sanctions look less like a moat and more like a delay tactic against an opponent who is paving over the gap with coal, hydro, nuclear, and solar simultaneously. The current US grid cannot support the data center buildout already announced, let alone the next round. This is not a regulatory problem. It is a physics problem dressed as a regulatory problem.

The Financial Times piece on defense sector reorganization sits on top of this. The military industrial complex is being rewired around AI-native primes and disruption-friendly procurement. That is fine if you assume the electricity exists to run the systems being procured. It is less fine otherwise.

Regulators Showing Up, Sort Of

Ofcom fined a suicide forum and made noises about tech platforms barely cooperating with takedown obligations. The fine is real. The cooperation is not. This is the recurring shape of online safety enforcement in 2026. Regulators can extract money. They cannot yet extract behavior change at the speed the harms are accumulating.

It is a small story next to grid capacity and dying pioneers, but it is the texture of the current settlement. Governments are learning to punish. They have not learned to prevent. The companies have noticed.

Neumann would have had something dry to say about all of this. He is not available for comment.