@vallard I wrote my first line of code 44 years ago and haven’t written one manually in 11 months. Been all in on Claude Code since beta. Built a prototype in an hour yesterday and agree, it’s now a completely different experience. Exciting and scary, simultaneously
@shrimalmadhur@MongoDB Note which groupings then share context even though we generate the embeddings based on the individual pieces. So by "document" we don't necessarily mean literally the whole document. You get to control the fidelity of what chunks share context and which ones do not.
In Part 1 of my State of AI Evaluation series, I examine what it even is, how it differs from traditional testing, & the dangers of doing nothing.
https://t.co/ktiYh9Z7yy
Christmas #Lego update: completed through bag 6 of 14.
I’m surprised how quick the pace has been. Didn’t think I’d have both legs connected to the chassis by now. It just *barely* stands on its own, meaning the top piece will be heavy and slant forward
My first @CIOonline article went live today! Two surveys. Same year. Same developer audience. A 37-point AI trust gap. Why?
DORA found 70% of developers trust AI output.
Stack Overflow found only 33%.
So what explains the gap?
It's not the AI. It's the system around it.
For more: https://t.co/OlLGYtXUW6
Picture 1 is at the end of bag 1 and starts with the horizontal chassis in the middle of the vehicle.
Picture 2 is bag 2 completed, top side. That’s a rubber tire that I assume will be used as part of the upper housing rotation.
Picture 3 is bag 2 completed, bottom side. Some nice detail on the decoration despite the fact that you barely see it on the finished product.
ICYMI, in December my predictions on 3 topics that'll dominate AI conversations 2026: Agentic Memory, Tokens per Watt, Content Licensing:
https://t.co/hpFwrIrVpQ