September 2026
TBPN’s John Coogan described a University of Cambridge proposal for AI safety as a shift from general calls for cooperation to two practical steps: standardize measures of AI’s contribution to research and development, and prepare government responses to failures such as cyberattacks or disruptions to critical infrastructure. The proposal, as Coogan presented it, identifies what governments should assess and plan for, rather than supplying finished response plans. That case comes amid growing public attention to AI risk and closer engagement between technology leaders and Washington.
David B. Lomet, a retired Microsoft Research researcher, argues that latch-free index trees can still lose scalability when competing threads perform the same expensive structural work before one wins a compare-and-swap. In his talk on the Bw-tree, he presents latch-free notices as an early way to select a worker, protect the state being changed and direct updates while consolidation, splits or merges proceed. The aim is to avoid redundant reconstruction without stopping other threads from using the index.
In the opening lecture of Stanford’s CME295 course on transformers and large language models, adjunct professors Afshine and Shervine Amidi explain how the Transformer offered an alternative to recurrent models. They argue that its use of attention let tokens connect directly across a sequence, rather than relying on a recurrent state to carry information forward one step at a time, and helped put language modelling on a more scalable path.
Meta’s Muse is being used to argue that consumer AI may depend less on the most powerful model than on software that connects to services and takes tedious tasks off users’ hands. On Big Technology, Alex Kantrowitz says Muse’s early use points to a product opportunity for Meta—and a challenge to intermediaries such as travel-booking sites—while Ranjan Roy cautions that adoption will hinge on trust and whether cheaper, non-frontier models can do enough. Neither sees Muse as proof that frontier AI has lost its value.
Bloomberg correspondent Mike Shepard reports that President Donald Trump and Anthropic CEO Dario Amodei met for dinner as they take opposing positions on AI development. Amodei has called for slowing the most advanced models and strengthening safeguards; Trump has resisted added restrictions, arguing that the US must move quickly to compete with China. The meeting was confirmed, but neither side had said what they discussed.
Former SpaceX Starship software manager Scott Morton calls the rocket’s first trip to orbit a “monumental step forward,” despite an engine failure during ascent. In an interview with Bloomberg’s Ed Ludlow, Morton argues that the flight’s data may prove as valuable as the milestone itself: engineers will spend weeks or months examining how the vehicle and its software performed. He says Starship’s goal of carrying people to Mars was part of the program’s vision from the beginning.
SpaceX’s Starship reached orbit for the first time and deployed 26 satellites, but an engine failure cut the mission short. Bloomberg reports that the spacecraft splashed down in the Pacific near Hawaii about three hours into the flight. The orbital milestone and satellite deployment were achieved before the early end; the available account does not specify which engine failed or what further objectives remained.
Eugene Volokh and Jane Bambauer argue that the legal question in the White House’s exclusion of CNN, MS NOW and Politico turns on where reporters were barred. A 1977 D.C. Circuit precedent protects access to the press briefing room from viewpoint-based exclusion, they say, while a newer ruling treats the Oval Office as the president’s own space, where he can choose whom to meet. A judge restored the reporters’ access on due-process grounds without deciding the broader First Amendment claim.
James Sexton argues that relationships depend in part on small, unrequested gestures of attention, while Rick Glassman questions the test of waiting for a partner to notice an unspoken need. The New York divorce attorney uses a client’s story about granola to frame a tension between asking directly for what you need and hoping a partner will notice on their own. A request may get the job done, Sexton suggests, but it does not carry the same meaning as a gesture freely offered.
ElevenLabs presents v4 as a text-to-speech model that makes delivery part of the prompt: inline tags can direct emotion, pacing, reactions and style, shaping how the words are performed. The company says the model is also designed to preserve speaker identity across long-form narration, dialogue and regenerated lines. For live applications such as agents, ElevenLabs positions v4 Turbo as the low-latency option, with a reported median inference latency of about 100 milliseconds.
Bloomberg’s report on SpaceX’s first orbital flight test separates the Super Heavy booster’s return from Starship’s continuing flight. The broadcast confirmed the booster’s landing while showing Starship at 18,270 kilometers per hour and 145 kilometers altitude; it did not report the deployment of the 26 satellites the rocket was intended to carry into orbit.
Economist Mike Munger argues that public choice begins by treating government not as a benevolent whole but as institutions run by people with their own information limits and incentives. In a conversation with Russ Roberts, he says voters’ rational inattention and politicians’ pressure to win elections make political decisions harder to assess than appeals to “the will of the people” suggest. Munger’s test for new government powers is to replace “government” with the name of a politician you distrust and see whether the proposal still appeals.
Adam Neumann says WeWork’s growth outran his ability to lead it: as a proposed $4.2 billion investment shifted his attention from the company’s mission to his potential wealth, he lost sight of what the business was for. In an interview with Steven Bartlett, the WeWork co-founder argues that rapid expansion is not itself the problem; founders must develop their judgment, teams and operating systems at the same pace, or the company’s growth can expose weaknesses that eventually crack it.
Charlie Holtz, co-founder and CEO of Conductor, argues that teams using coding agents should build around human judgment rather than treat software development as a factory line. His principles include staying close to new tools without chasing workflows that will soon become standard, protecting code and agent instructions with strict review, and giving agents access to organizational knowledge and persistent workspaces. The aim, he says, is for people to conduct teams of colleagues and agents, not simply manage automated production.
Zac Efron said preparing for *The Iron Claw* meant eating about 6,000 calories a day, training at high volume and wrestling until he reached roughly 205 pounds—a routine he described as painful and said he never wanted to repeat. But he does not want that transformation to become the expectation for every role: he said he prefers work that changes what acting asks of him, including trying different genres and playing real people. During a lighter workout with Chris Williamson and trainer Ben Greenfield, he asked to keep the pace easy enough to talk and hang out.
Reportage artist George Butler argues that drawing can reveal what breaking news leaves out: not just what happened in a war, but how it felt to the people living through it. In his TED talk, he describes sitting with Ukrainians—including a 99-year-old famine survivor, a man collecting books from a rose garden after a missile strike and an injured architect—to listen and record details that might otherwise go unnoticed.
Ido Salomon, creator of AgentCraft, argues that the obstacle to using coding agents at scale is the human effort required to steer, direct and review them. His strategy-game-style orchestrator represents agents as units on a map, with the file system as terrain, and combines visibility tools with autonomous workflows and shared workspaces. Salomon says the detailed interface may suit power users, but not everyone; an early, simpler mode is intended to make agent tools easier to approach.
Country singer and songwriter Dylan Gossett argues that artists risk losing the conditions that make their work distinctive when they let a breakout dictate what comes next. In an interview with Chris Williamson, he describes how “Coal” reached a large audience without a planned campaign, then considers the pressures to repeat a hit, follow audience feedback and outsource creative work to AI. Gossett’s point is not to reject promotion or new tools, but to protect the time, experience and freedom that allow him to write songs on his own terms.
Pull requests are becoming a bottleneck as agents increase the flow of changes faster than teams can review them. In a talk at AI Engineer Paris 2026, Matt Pocock argues that the answer is not simply to generate code faster, but to make changes more reviewable: run deterministic checks, then agent review, before human judgment. Human reviewers can focus on risk and use recurring findings to improve the process behind future PRs.
OpenRouter co-founder Alex Atallah and investor Anjney Midha argue that AI will remain a multi-model market: labs can build capable models but still struggle to get them into developers’ hands. They say OpenRouter grew by providing a neutral way to discover, compare and route among models. Midha argues that as agents drive more continuous use, the resulting token flows create a growing target for fraud.
Microsoft is recasting Copilot as Autopilot, a consumer agent designed to work across a user’s services and personal life. On TBPN, John Coogan and Jordi Hays argue that Microsoft’s reach among ordinary consumers could matter as much as the agent’s capabilities: people may encounter it through tools they already use, even if AI enthusiasts are not drawn to a Microsoft-branded product. But acting across services asks users to trust an agent in a more active role.
The Paramount settlement may keep the company in Los Angeles, but it does not promise production in California or reverse the state’s loss of entertainment work, Bill Whalen and Lee Ohanian argue. Their account points to a broader gap between preserving a major company and resolving the state’s underlying problems: Ohanian says high-speed rail lacks a sound business case, while Whalen warns that the project may run out of cash by December 2027. Whalen also reads Governor Gavin Newsom’s recent political moves as positioning for a possible 2028 presidential run.
Sarah Paine argues that wars are easier to start than to end, and hardest to end on favorable terms. Using the Russo-Japanese War as her main example, she says Japan secured a favorable settlement by keeping its aims limited, recognizing what it could not afford to sustain, and preparing a route to negotiation before its military position deteriorated. She warns that expanding war aims, inviting outside powers into a conflict, or mistaking conventional victories for lasting control can make a settlement harder to reach.
Julian Gewirtz, a former senior director for China and Taiwan Affairs at the White House, says Donald Trump’s meeting with Xi Jinping achieved its apparent aim of projecting a positive relationship but produced little visible progress on the disputes dividing the United States and China. He argues that trade, Iran, Russia and Taiwan remain unresolved, while competition over AI and access to advanced chips adds to the stakes.