Robot cage fights and snortable brain chips: Inside San Francisco's AI boom
San Francisco's AI boom is producing scenes that blur the line between real technology and science fiction. In a Fortune report published November 18, 2026, Tech editor Alexei Oreskovic describes a city where humanoid robots fight in cages, startups want to put nanoparticles in people's brains, and even the people building AI models admit they do not fully understand how those models work.
Robot cage fights in an old mattress showroom
An old mattress showroom on San Francisco's Van Ness Avenue now hosts a strange kind of entertainment. Six-foot-tall humanoid robots, some fitted with steel claws, square off in a small arena and hit each other with heavy force. The events are run by a company called REK, short for Robot Entertainment Kombat. According to Fortune, these metal-on-metal fights have built a passionate fan base.
A brain chip you take through your nose
Oreskovic also visited Subsense, a startup in Palo Alto. The company is building a brain-computer interface called NanoBCI — a system meant to connect a human brain to a computer. The field grew after Elon Musk's Neuralink showed that a wireless chip placed on the brain could turn a person's thoughts into actions. But most people do not want holes drilled in their skulls. Subsense is trying a different route: users would snort nanoparticles through the nose, and a cap that looks like a mix of a baseball cap and a bike helmet would control them through metal receptors.
Ray Kurzweil, the 78-year-old futurist known for popularizing the idea of "the Singularity" — the moment when AI becomes smarter than humans — works with Subsense as an advisor. He told Fortune that AI could help humans reach what he calls "longevity escape velocity" by 2032, a point where technology extends life faster than aging shortens it. "We can actually speed up the brain itself; that's the ultimate goal," he said. He also held up his smartphone and said the goal was to get it inside the mind, adding that he was tired of losing the device in the bathroom.
These claims about the future are Kurzweil's own views, not confirmed results. The article presents them as predictions from a well-known optimist, not as proven science.
Everyday life starts to look like a satire
Fortune describes daily details that feel futuristic. Self-driving cars weave around cyclists with no driver inside. Billboards tell employers to "stop hiring humans." One company is reportedly paying people $20 an hour to "walk" robot dogs, training them to move through city streets — Fortune labels this as a report, not a confirmed fact. Even coffee shops are part of the trend: at Corgi Cafe, customers can order a "Coretado" co-branded with CoreWeave or a "Sip It and Ship It" drink by Pendo.
Even the builders do not understand the models
Some of the world's top researchers work at OpenAI, Anthropic, and other companies building the large language models behind generative AI. But Fortune reports that even these people say they are not sure how the technology actually works. The models act as a black box: workers can adjust settings, but what happens inside remains a mystery. Oreskovic argues this is a key reason the city feels disorienting. The usual tools for judging a real breakthrough against a sales pitch no longer feel reliable.
The report also notes bigger open questions now being asked in public: Is AI conscious? Does it suffer? Can it improve itself without human control? These remain unanswered, and the article treats them as open debates rather than settled facts.
Faith over understanding
Oreskovic contrasts today's boom with past ones. The internet was new, but it did not challenge human ideas about nature or consciousness. When the dotcom bubble burst, business models failed — not people's understanding of the world. Now, he writes, the Anthropic cofounders described an early "age of miracles" in their IPO letter to investors, and the phrase may fit more deeply than they intended. Miracles rely on faith rather than understanding, and in Silicon Valley today, even the builders struggle to explain what they have made.
What is confirmed
- REK runs robot cage fights with humanoid robots in a former mattress showroom on Van Ness Avenue in San Francisco.
- Subsense, based in Palo Alto, is developing the NanoBCI brain-computer interface using snorted nanoparticles and a sensor cap.
- Ray Kurzweil, 78, serves as an advisor to Subsense.
- Neuralink has demonstrated a wireless brain chip that can turn thoughts into actions, according to Fortune.
- Fortune reports that people building large language models say the models work as a black box.
What is still unclear
- Whether nanoparticles can safely and effectively work as a brain-computer interface is unproven; the technology is still in development.
- Kurzweil's timeline of reaching "longevity escape velocity" by 2032 is a personal prediction, not a confirmed outcome.
- The $20-an-hour robot dog walking jobs are described only as reported, with no named company confirmed in the article.
- Questions about AI consciousness, suffering, and self-improvement remain open.