In April, '17, James Somers of The New Yorker reached out & said he wanted to do a small profile of me & my longtime colleague Sanjay Ghemawat, watch us work for a few hours, maybe dinner, etc. It came out today. I think it captures our working style really well. newyorker.com/magazine/2018/…

Announcing TextualHeatmap for Project Jupyter and Google Colab. Since publishing my distill.pub/2019/memorizat… paper last year, many have asked me how to make the connectivity plot. Today I finally made it easy you: github.com/AndreasMadsen/…

towardsdatascience.com/springer-has-r…

Really proud to announce the open source release of Blender by @facebookai. The largest ever open-domain chatbot which feels more human than any other bot according to human evaluators. Looking forward to what the AI community can build on top of it! ai.facebook.com/blog/state-of-…

I've played a little bit with Knative on Packet blog.vtemian.com/post/serverles…

"Success = helping more people solve more problems" -- 👏Notion for the approach

Here's a sentence describing what Google's home page should look and here's GPT-3 generating the code for it nearly perfectly.

One of the most useful shortcuts ever 👇

Interesting study showing the best predictors for how well someone learns Python isn't numeracy but rather [natural] language aptitude and general cognition nature.com/articles/s4159…

Ten 10x Concepts: (thread)

Practice.

As an AWS user, I often want to know the cheapest options for compute resources given some project-specific criteria. So, I made a tool to show me that data in a way that is useful to me. Maybe it will be useful to you, too: compute-cost.com

Giving feedback is hard: • You’ve identified a problem • You know you need to talk about it • You’re worried they'll get defensive Relax! It’ll be okay. Here’s the framework I use to turn every feedback conversation into a win-win:

ChatGPT & BingGPT take user-supplied text prompts & return relevant language. But unlike their predecessor GPT-3, they're interactive & you can have a dialogue with them. How does this work? That & much more, explained here: jonstokes.com/p/chatgpt-expl…

In a way the Earth already is a super-computer. Map of submarine internet cables:

Even though Python’s standard library comes with “batteries included”, it’s still missing some essentials and here’s the list: betterprogramming.pub/pythons-missin… Better Programming #Python #programming

I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵

My biggest worries about coding with AI: 1. Beginners not actually learning 2. Atrophy of skills I’m seeing #1 happen and I don’t have a good answer yet. Leveling up as an engineer requires grinding and it’s not always fun. If AI can solve most of the problems for you, when

When we started Plain, we built it API-first. At that point, we didn't know what it would enable, but it felt to us that there was a big opportunity. For the first few years, it didn't really work because it was missing a use case to make it valuable. Now, it's the reason we win