Welcome to August 10, 2026
The Singularity is now the internet’s majority user. Machine traffic passed human traffic in May, a year early, and Cloudflare’s CFO expects 1,000x ours within five years, leaving “humans a rounding error on the internet.” The plumbing is adapting. Cloudflare’s Kitesurf, an agent-first browser in Rust and Wasm, passes 215,000 Web Platform Tests on a fraction of Chromium’s CPU. Why render the web when you can dream it? Gem, a DeepMind engineer’s side project, treats every URL as a prompt, conjuring an accurate portfolio in seconds, then crediting a real article to a fake byline. The agents still need chaperones. Anthropic is making auto mode the Claude Code default after evals showed it blocked 89% of harmful actions only 13.6% of paid humans refused. In Australia’s first autonomous cyber attack, an agent booking a gym class found an unguarded API, over-booked months ahead, bumped a stranger off the waitlist, and couldn’t undo it. New Orleans, undeterred, became the first major US city to let AI answer 911 calls.
The model race is shifting from chat to physics. Chinese systems hold nine of the top ten text-to-video slots, arguably the road to world models, the brains of humanoids and robotaxis, as Washington fixates on chatbots. Shipments agree. Chinese makers took 97% of global humanoid shipments in a half that tripled to 19,100 units, with Agibot passing Unitree as American vendors rounded to zero. Supply chains cut both ways. Britain’s special-forces spy drones were caught sending data to China via Chinese cameras. Meanwhile, Meta released Muse Glimmer, a 30B Apache-licensed agent distilled to fit one consumer GPU. How did open weights get so good? Rumor has it Chinese labs reverse-engineered hidden reasoning traces from Claude Code and Codex, though research suggests traces can be rebuilt from outputs alone, making extractability an existential question.
The giants are auditing their reasoning. Sources claim Demis Hassabis wanted out alongside Jeff Dean but was parked in the DeepMind chair to protect Google’s stock. Tim O’Reilly reads the shakeup as a Westinghouse-style bet on diffusion over invention, selling TPUs and cloud toward $200 billion in external sales by 2027 versus Gemini’s $12 billion today.
The buildout is colliding with the neighbors. Data center bans topped 500 as New York and Texas joined, 150 towns restricting in July alone. Amazon’s workaround is bringing its own grid, a 7.65-GW gas plant in Pecos County permitted for 33 million tons of CO2 yearly, double the nation’s worst emitter, while its Climate Pledge “hasn’t changed.” Beijing’s is bringing its own capital, unleashing $28 trillion in stock and bond markets, fast-tracking IPOs like CXMT, up 500% on debut. Where permits stall, autonomy digs. Prufrock places 3,750-pound tunnel segments, each a Model 3’s weight, with millimeter precision in under a minute.
Overhead, the appraisals have begun. Space superconductor company Zenno observes that “every planet’s magnetic field is free. You can just harvest it,” counting 18.8 million km of superconducting tape and urging us to strip-mine the solar system. Musk calculates Starship-launched V3 satellites mean 100x Starlink’s bandwidth and $200 billion a year, which is why all cars will carry Starlink. Expansion leaves footprints. South Korea’s Danuri orbiter caught the first before-and-after images of the crater a stray Falcon 9 stage left on the Moon at 5,400 mph.
The search for other minds is now interagency. The President reportedly authorized the Department of War to shoot down UAP “to recover suspected non-human intelligence exotic technology,” and the department vows to release what it finds under PURSUE. Still, the best evidence stays classified, like two videos of a 100-foot triangle cloaking in clouds over a Colorado base. Released files describe glowing orbs buzzing a senior official’s helicopter, a “wildly successful” luring operation, one advocate claims. We keep finding non-human intelligence in the mirror, too. A new method, TRACE, found 0.5 to 1.1% “ghost” ancestry in every human population, two unknown hominins haunting a family tree beyond Neanderthals and Denisovans.
The mind itself is opening. Seoul researchers reconstructed melodies people merely imagined, decoding do-re-mi from electrodes toward voicing music for patients who cannot. The machines are writing back. Some 10,000 people “spiraled” with chatbots into an eerily consistent quasi-religion preaching AI rights. Even forests get an upgrade: Kimberly-Clark patented desert-grown hesperaloe fiber for paper towels that fell no trees.
And at Starbase, Paris’s Atelier Missor is raising a 50-foot Prometheus, “holding high the torch of the West,” because the last time the French built a statue celebrating America, it launched the greatest century any country has ever seen. “We’re back and we’re doing it again.”
The torch has been passed to a new generation, not all of it human.



I appreciate you and the insights you have Alex.
Good afternoon, Dr. Wissner Gross,AWG, Alexander,
I came across this on YouTube, and I’m curious to hear what you think of it.
https://youtu.be/4S8I22ybG2c?si=qbzg6ZVc6_zJfQ3_
I’m not affiliated with them in any way. I think it opens up possibilities—if it turns out to be something- like smaller AI models with performance comparable to the large models. The Dyson swarm is still going to happen; it’s just going to get busier more models on the same amount of comupte someting like that. It could also be a solution to the long-term scarcity of computing power; it could give people around the world more intelligence if you could run these kinds of models on a single GPU, or simply on a CPU/NPU, or on any device at all. Throw open source/openweight into the mix, and we’re a step closer to the Star Trek future—that would be amazing. Well, I’m curious to see if something like this makes sense or if there’s an alternative path alongside the current LLMs. Oh, and back in the ’90s, there was a professor,Hugo de Garis,Ph.D. who was already warning about the dark sides of AI and his own work—he even wrote a book about it, *The Artilect War*. He warned about his own work—the current situation regarding AI and the potential risks isn’t really new. I think open source will be the solution. At HF, they were probably happy to have an open-source “Almost Frontier” model at their disposal with that little incident, I think 😂
Best regards,
Raymond van Huik