Kimi K3 Is the Inflection Closed Labs Don't Want

Gavin Baker just put a Chinese open model on the same board as Anthropic and OpenAI. The rest of the week was Grok, Cybercab, and $2-an-hour robots.

Kimi K3

Gavin Baker said Kimi K3 may be an important inflection point for AI - potentially negative for Anthropic and OpenAI, net positive for essentially every other company in the world - and 2.4 million people saw it. If the real Sputnik moment is an open-source frontier model, the closed labs lose the right to tax intelligence for a decade.

Check the post here.

Grok still does not get the hype

imPenny2x is running Fable, K3, 5.6, and 4.5 about 10 hours a day, at work and at home, with subscriptions in both places, and still says Grok does not get enough hype - even on X. The model people actually live inside all day becomes the one that runs the company.

Check the post here.

Boring Company, Vegas

I got an exclusive private tour of The Boring Company's Vegas facilities for a behind-the-scenes look at everything they are building. A lot of people are sleeping on this company. Whoever can move vehicles underground at scale writes the next 20 years of city transport.

Check the post here.

Grok is the Flow LLM

I called Grok the Flow LLM. Grok 4.5's biggest advantage is speed. It is smart enough to hang with the other models on most things, but that speed lets you make little tweaks until the output is perfect, and it is now my preferred model. The model that keeps you in flow eats the ones that make you wait.

Check the post here.

Robot cars, worse traffic

Elon said traffic will get much worse with robot cars, because the pain of driving yourself is not there anymore. Cities that do not price the road will drown in empty robot miles.

Check the post here.

I was right on Cybercab

The Tesla community is slowly realizing I was right the whole time on Cybercab. Too bad none of them apologized. If they are only now catching up, the next decade of Tesla volume gets decided by the argument they spent a year denying.

Check the post here.

AI displaces first

There is a fallacy around AI creating jobs. AI will displace jobs at a far greater rate than it creates them for the foreseeable future. The share of people who can adopt and deploy AI at a high level is extremely small right now, that is how you create the jobs, and a lot of sectors will be incapable or unwilling to adopt it. The next decade is a labor shock before it is a job boom.

Check the post here.

$2-an-hour builders

In about 10 years we will have ultra-intelligent autonomous robots building the most incredible structures in the history of human civilization, at about $2 an hour, and you can build as many of them as you want. The cost of a building drops toward materials and electricity.

Check the post here.