More thoughts on the Resonant Computing Manifesto on my blog

Let's build hyper-personalized AI-powered software that avoids the attention hijacking anti-patterns that defined so much of the last decade of software design - here's our manifesto with principles on how we can do that

Some notes on the two features in the newly released Django 6 that most caught my eye: background workers and template partials

Gemini 3 Deep Think is now available for Ultra users, making available our IMO & ICPC Gold Medal-winning technology. Deep Think shows improved generalization on difficult benchmarks like ARC-AGI-2, and outperforms Gemini 3 Pro on HLE & GPQA Diamond. We hope this serves as a

This was a really fun conversation - I posted some of my highlights (which Opus 4.5 helped extract from the transcript) here

Four new models from Mistral today - all Apache 2 licensed, all vision-capable, and one of them is a 3GB model that can run in a web browser and answer questions about things it can see through the webcam!

Five new Mistral models today, but the one I'm most excited about is this tiny 3B model that has vision support and can run entirely in a browser!

This is so wild... the leaked Opus soul document has now been confirmed! I wrote some initial notes about it on my blog https://t.co/tkOSAU663a - I like how it opens with this section about Anthropic themselves:

Notes and pelicans for today's pair of DeepSeek v3.2 model releases

A tiny TIL: if you get "Error 153: Video player configuration error" on YouTube videos embedded on your site a likely culprit is sending the "Referrer-Policy: same-origin" HTTP header Switching to "strict-origin-when-cross-origin" fixed it for me: https://t.co/4rSGiI1OVC

Pulled other some fun quotes from the early months of ChatGPT on my blog, OpenAI really did not expect it to be as popular as it was!

Happy 3rd birthday, ChatGPT. I was at Google then, leading two products: AI Test Kitchen and NotebookLM (there was no Bard/Gemini yet). We had launched AITK much earlier than ChatGPT, but it was a heavily constrained experience in comparison. It was a cool demo of LaMDA but

"language interfaces are going to be a big deal, i think"

Three years ago today

I knock out small tools on my phone several times a week now, but my most ambitious phone coding project is still this one

One of the underrated things about LLMs for coding is that they make writing code on a mobile phone more feasible I can build small but useful pieces of software entirely on my phone now - without even needing to pay full attention, so it's compatible with being out in the world

The Netflix number is five years old and the ChatGPT number pre-dates GPT-5 by a couple of months so there are plenty of caveats here, but I expect the order of magnitude remains about correct

On my blog:

Out of curiosity I decided to try and run the numbers on how much Netflix you can watch for the energy cost of a ChatGPT prompt As far as I can tell it's between 5.1 and 10.2 seconds, depending on which end of the 2019 IEA Netflix energy usage estimate you use

Here's that thread from the demo video displayed using the tool I've been building