Welcome to June 5, 2026

The Singularity has started writing its own sequel. The Anthropic Institute published “When AI builds itself,” showing with public and internal data that AI is already accelerating AI development, and arguing the world should keep the option to verifiably pause before recursive self-improvement, which it says “could come sooner than most institutions are prepared for.” The numbers are giddy. Anthropic engineers now ship 8x as much code per quarter as they did in 2021-2025, and Claude’s success on open-ended problems has jumped 50 points to 76% in six months at quality already on par with human and projected to pass it within the year. On a standing test to speed up model-training code, Mythos Preview hit ~52x where a skilled human reaches 4x and 2024’s Opus 4 managed 3x. It is even learning to steer, improving on researchers who hit a dead end 64% of the time, up from 22% in 2024. Anthropic admits none of it guarantees runaway recursion yet, since picking the right problems is still unproven, but if the curves hold, systems designing their own successors could revolutionize medicine, technology, and the economy for the better. That same momentum is why it asked rival labs to weigh slowing down, a brake it concedes would be harder to verify than a nuclear site, and one almost nobody is reaching for. Everyone else has the throttle pinned. Cognition will foot your bill, up to $10M, if Devin underdelivers, Mythos reportedly now zero-shots medieval Minecraft villages, and an ICML paper shows transformers can share their query-key-value projections to shrink the KV cache up to 96.9% for on-device inference. OpenAI’s Dan Roberts expects the coming months to turn AI on itself until studying it feels like physics.

Intelligence is flooding the consumer layer. Apple’s App Store ecosystem moved $1.4 trillion in 2025, nearly triple 2019 and over 90% commission-free, with AI-powered apps growing billings 4x faster than the rest. The glasses are watching back, as WIRED found Meta has quietly shipped a dormant facial-recognition pipeline, “NameTag,” to tens of millions of smart-glasses phones. OpenAI’s new “dreaming” memory keeps ChatGPT current on your life in the background, upgrading “you’re going to Singapore” to “you went,” with a 5x cheaper version bound for Free users, while Anthropic embeds engineers in the NSA to point Mythos at offensive cyber.

The grid is being reforged to feed it. Kevin O’Leary halved his 40,000-acre Utah data center after a backlash over the Locomotive Springs refuge, trimming 19,430 acres, but the clean firehose is opening. Helion raised $465M at a $15.5B valuation after its prototype, the first private machine to burn deuterium-tritium fuel, blew past 150 million °C, and Antares hit the first private non-light-water reactor criticality in the US in 40 years, with electrons flowing from 2027. Waymo is even resurrecting retired robotaxi batteries as hundreds of megawatts of storage on the very California and Texas grids its fleet charges from. Even the White House’s emergency $700M for “clean, beautiful coal” is, at bottom, a bet on powering the superintelligence boom.

Intelligence is looking less like a summit to scale than a tide rising everywhere at once. Starlink now connects 12M customers across 160+ countries, while a Science study found bumble bees can spontaneously solve a novel puzzle even when the goal was hidden in transit, proof that flexible cognition is no big-brain monopoly. Biology is sprinting to keep up with its own imagination. Caltech’s Sidewinder DNA synthesis misfires once per 10 million joins, stitching a 12,500-letter E. coli genome error-free in days, finally fast enough to build what models like Evo 2 dream up faster than anyone can assemble. And Cambridge trialled the first AI-designed vaccine in people, aimed at every coronavirus and now flu and Ebola.

The market is racing to price the upside. Anthropic’s rift with the White House is thawing ahead of its IPO, reportedly moving toward shedding its “supply-chain risk” label. Washington is quietly weighing equity stakes in AI labs, an idea Altman pitched to the President in early 2025 as a public dividend and one Anthropic says it has stayed out of. Canada is betting a C$500M fund through its “AI for All” plan to mint 250,000 jobs and lift GDP nearly 3%, almost C$200B. Fresh blood keeps pouring in. Airbnb’s Brian Chesky is funding an AI lab for interaction and design rather than text chatbots, pitting him against his former mentee Altman, Founders Fund is filming Altman and Palmer Luckey playing Mafia to escape boring VC content, and Pump.fun launched GO, escrowing $5 to pay anyone for any task. Meanwhile, Argentina’s Milei went furthest, giving AI agents a legal home, a “non-human corporation” whose independent judgment earns it limited liability.

Don’t cry for me, Argentina, the truth is AI never left you.