Kevin Webb
@triangulator
Tinkering with semiconductors and copper for ebikes, cargo bikes, and things slightly larger. Also curious about the pre/post-cloud internet (forecast: 🌤️).
We watched it end to end. She was riveted and is now plotting to make her own dumpling toy that won’t break!
Yeah, really hard to see a path where this ends well for the folks betting on inference margins. artificialanalysis.ai/models/kimi-k3
Ran my standard VLM eval on Kimi K3 and am increasingly convinced the only path forward for the US model labs is regulatory capture/legislating a ban on open weight models. I've been running this test for over a year and have never seen anything like this: www.kimi.com/share/19f6cb...
10/10 recommend this book! My 5yo wrote down the page numbers for the recipes she wants to immediately make/eat!
Gemini 3 Pro is still in first place on this task, but Qwen3.5 is now *almost* as good at ~1/5th the price.
This is a very different way of writing software, and seems pretty useful. Built an agent that vendors/maintains code from a good but nacent/unstable open source repo.
There have been bots on social media since forever, but this hits different. Also, @penny.hailey.at hello! Do we have a name for the kind of thing you are yet?
FWIW, those brown butter pecan cookies might the best thing I’ve ever baked.
This is absolutely bonkers. If you’d have asked me even a few months ago if this was possible, I’d have said no way…
And some vibe hydrology to collaborate with the neighbors on a backyard drainage problem -- 30 minutes with Gemini 3 Pro/antigravity:
I’ve had a running test for multimodal models over the last ~year: extract the “Crss” values for each tick mark in this chart. GPT 5 makes up numbers, Claude 4.5 refuses to try, Qwen3 VL has a plausible guess. Gemini 3 pro pretty much nails it.
Good day to reflect on the tradeoffs between relying on monolithic platforms and protocol-driven infrastructure like BGP Anycast. Anycast is what makes these platforms possible, but it’s nearly impossible to implement on your own. (Cloudflare requires an enterprise account to enable it.)
It’s going the be fascinating to see how this industry picks up the pieces from the bubble and gets back to research. There’s a lot of work still to be done that doesn’t require trillions of dollars in hardware.
Somebody at the body shop DIY’ed the compact pickup Detroit won’t build.
idk, I think this is mostly a design choice on the part of the bot purveyors. Sycophancy sells. Putting simple instructions into the context turns Claude (and ChatGPT to a lesser degree) into a combative critic that strawmans wild failure scenarios just to poke holes in my questions...
Airlines have been trying to use computers to decommodify and differentially price seats since the time of their invention. But “agentic purchasing” and embedding payment into the chat is going to make past conflicts like “screen bias” look quaint… jolt.law.harvard.edu/assets/artic...
Hadn't seen this happen before: Claude dropped into Mandarin to explain that I was attempting to mix (混合) two different ideas in my request.
Great piece on BYD's $8,000 EV. Also, fascinating tidbit about Shanghai banning small/inexpensive cars:
Doubly unfortunate that conversations about surveillance devolve into unrelated fights about red light and speed cameras (which are useful/good). Companies like Flock, that power nationwide LE infra, have a different definition of "safety" and don't even make products that enforce traffic laws.
Unfortunately not enough folks who work with this technology understand how we got to where we are with 4a law's limits or intentions. Can't recommend enough Sarah Seo's book unearthing this history. thewaroncars.org/2019/10/31/t...
Astonishingly, the US bike lobby is attempting to get sectoral tariff relief by blaming a lack of protectionism and "cheap" foreign bike imports in the 1990s for the collapse of US manufacturing. 1) Bonkers logic. 2) US cos failure to respond to 90s trends like MTBs killed the domestic industry.