I went aurora hunting last night!
Inspired by this post, I had Opus 5.5 build a visual explainer of how auroras actually work from my phone on the drive up. Our seasoned aurora-hunting hosts were blown away so I published it so they could use it for future guests.
You can play with the whole scene and simulate different solar wind conditions to see what the aurora might look like from the ground. It also supports pulling live satellite NOAA solar wind data and see what tonight's sky might look like.
Check it out! auroraexplained.com
I asked Opus 5.5 to explain camera focus by building an interactive lens lab
Here's what it came up with after 1 hour 26 minutes in one shot, $25.66 API cost
lens.lab.sael.net
Move the focus ring and you can see the glass elements shift the sharp plane through the scene
Sean Abraham retweeted
some people just get it. you throw ideas in the air, they catch them, toss them back, stretch them, remix them. they understand that thinking is a live process, not a checklist. every conversation feels like a high bandwidth exploration, open-ended, dynamic, fun. my friends are in this category.
then there are the ones who default to critique. the ones who immediately point out flaws, who force you to preface everything with disclaimers like “this is just a thought experiment, let’s see where it goes man.” they treat every idea like a proposal up for review instead of a stepping stone to something interesting.
the first group is amazing. the second is mostly careerist employees, trained to evaluate rather than create, like a quarterly performance review.
Sean Abraham retweeted
On one hand, data centers use insignificant amount of water. On the other hand, they cure cancer. So who’s to say if they’re good or bad.
Sean Abraham retweeted
AI dadacenters just don’t use that much water, there’s a panic over almost nothing.
Here is a short animation to put AI water use in context:
Sean Abraham retweeted
Anytime someone steps down to “spend more time with their family” I think of @andrewmason’s excellent letter when he got fired from Groupon
White House Press Secretary Karoline Leavitt is stepping down at the end of the month to spend more time with her family, President Trump said.
Read more: on.wsj.com/4zfOu8y
Sean Abraham retweeted
EXCLUSIVE: a startup called Squint says its new AI system beats comparable tools from Anthropic and OpenAI in industrials.
Squint fine-tuned a ~2 billion parameter open-weight model on manufacturing and compared it on complex tasks that are tough for a human to answer, CEO Devin Bhushan says.
The results could be a big unlock for manufacturing customers like PepsiCo.
And they present another vertical where specialist startups are racing to keep ahead of the AI labs. Link below 👇
New paradigm incoming. This is gonna be the big one. It’s only gonna take off when inference costs are significantly cheaper though, probably need a 10x reduction
Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.
@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)
Sean Abraham retweeted
We built an AI system that discovers health biomarkers from wearable data.
One of its first findings: "late-night doomscrolling" is a statistically validated predictor of depression severity (ρ = 0.177, p < 0.001, n = 7,497).
The AI named the feature. No human guidance.
Yo @KianSadeghi5 I’ve used @nucleusgenomics to get my whole genome sequenced already. What’s the best set of tools to throw a coding agent at it? Maybe I should make a skill for this.
I'm lucky enough to have a great doctor and access to excellent Bay Area medical care. I've taken lots of standard screening tests over the years and have tried lots of "health tech" devices and tools.
With all this said, by far the most useful preventative medical advice that I've ever received has come from unleashing coding agents on my genome, having them investigate my specific mutations, and having them recommend specific follow-on tests and treatments.
Population averages are population averages, but we ourselves are not averages. For example, it turns out that I probably have a 30x(!) higher-than-average predisposition to melanoma. Fortunately, there are both specific supplements that help counteract the particular mutations I have, and of course I can significantly dial up my screening frequency. So, this is very useful to know.
I don't know exactly how much the analysis cost, but probably less than $100. Sequencing my genome cost a few hundred dollars.
(One often sees papers and articles claiming that models aren't very good at medical reasoning. These analyses are usually based on employing several-year-old models, which is a kind of ludicrous malpractice. It is true that you still have to carefully monitor the agents' reasoning, and they do on occasion jump to conclusions or skip steps, requiring some nudging and re-steering. But, overall, they are almost literally infinitely better for this kind of work than what one can otherwise obtain today.)
There are still lots of questions about how this will diffuse and get adopted, but it seems very clear that medical practice is about to improve enormously. Exciting times!
Everyone is going nutso for file systems right now which I love but I feel like this one is overfitting for the AIs of today.
Not everything cleanly abstracts onto a filesystem, it’s too awkward when data is large and densely connected.
In context learning plus code is the way
Calling it now: BBQ is going to start exploding in Japan. I bet by this time next year there will be serious brisket somewhere in Japan
This lead me down a rabbit hole to learn that @demishassabis built the AI for the game Black & White back in the day!!
extraordinary work. and drawing on the og work on genius handcoded assembly by @demishassabis, and drawing on current work under his banner by nano banana. gloriously self referential
