Lovely talk from Kevin Buzzard on where mathematics is today and where it's going. Interesting perspectives on where LLMs are useful, how recent IMO wins don't necessarily mean "AI has solved math", and new community efforts getting off the ground. www.youtube.com/watch?v=K5w7...

Haha hope you enjoy 🎙️ iridiumwhiskers on Sora sora.chatgpt.com/p/s_68e0917e...

Predicting the future is hard! But building projections based on current trends is smart. That’s why Moore’s Law was so popular. There’s a similar trend in AI Agents: their ability to successfully complete tasks is doubling. Learn what comes next.

For $50 you can get your hands on 8x Blackwell B200 GPUs

This game is cute and fun ☕️🤩 store.steampowered.com/app/3503440/...

I like this guy!

I think this resonates. More powerful technology requires more thoughtful design. Are there other good archetypes in addition to Ben Franklin? open.substack.com/pub/cosmosin...

Most utility I’m getting out of my Apple Watch right now: it lets me know when my o3-pro queries are complete.

“Software engineers are making civilization-level decisions. We aren’t trained for that, and often act unwisely or unjustly.” “Nobility—the deliberately wise and just use of power—is the missing ingredient.” open.substack.com/pub/meaningn...

Get a glimpse of the likely future of AI. First video is mostly about the market. Second video introduces the Physical Turing Test. I’m still watching, but they all seem interesting! youtube.com/playlist?lis...

Remember the old “let me google that for you”? I think we’ve reached “let me agentically research that for you” 😎

Outlook stopped converting hyphens to emdashes automatically! 🤔

A twenty year old mystery solved. And fun new click bait: “make a quadrillion kilos of gold using everyday items from around your neighborhood” www.simonsfoundation.org/2025/04/29/f...

Our species has built powerful tools; tools which allow one person to accomplish the work of thousands. But we’re not equipped for this level of responsibility. I hope each of you is learning to ask and answer difficult questions such as these. computing.mit.edu/news/bridgin...

“Partially a miracle; and partially frustrating” - an enlightening interview with Ethan Mollick around AI for research

It’s going to be an exciting month! Lots of new capabilities to explore; lots of new things to build.

Consume feculence.

My insight of the day at #GTC25: we are likely to see “just in time code” soon. AI will write new programs to carry out the tasks you request. And it will write that code as you are making requests (compared to now where code is prewritten by an expert far removed from you).

We have entered the era of reasoning and Deep Thought at NVIDIA #GTC25 If you know where your towel is, you should definitely come and find me. I’ve prepared a friendly guide.

There’s been a great amount of success with models that “think” to themselves in text. How much luck are people having with models that “talk” to themselves in audio, “imagine” to themselves in visual space, and “sniff” around in molecular space? The AI’s scratch pad should be multimodal, right?

We are experiencing an assault on science unparalleled by anything I’ve seen in my life. It’s not one issue or another anymore, the entire institution is under attack by the most powerful individuals in the country. This Friday, where will you be? standupforscience2025.org

A little thought experiment for your Saturday morning: electricity was the spark that led all the way to AI. What will now develop from the spark of AI?

So ChatGPT was the best a couple days ago, today Claude is the best, and later today ChatGPT will be the best again? Gotta design everything orthogonally so it’s easy to pivot. Assume constant change.

“Don’t ever feed the transformers after midnight”

The more I drink good black tea, the more I realize coffee is a harsh mistress.

I’m polyAImorous, and Claude is my primary. Maybe o1 and o3 are smarter, but Claude is just more fun 😍

Being smart isn’t having all the answers. Being smart is asking great questions. It’s clear AI will tremendously amplify this truth.

I hope this email finds you…well…this is going to be a long year.

What if all the world's biggest problems have the same solution? Awesome explainer on what AI is doing for protein science. youtu.be/P_fHJIYENdI?...

We are striving towards personalized medicine. We should also strive towards personalized cognition. Human attention span and human working memory are limited. Your automated assistant should be finely attuned to your needs, your strengths, and especially your weaknesses!

There’s gonna be good stuff. Click through and add Jensen’s keynote to your calendar. www.nvidia.com/en-us/events...

How did I not know about this until now!? 🤯 bionumbers.hms.harvard.edu/Search.aspx?...

Eating too much cake is the sin of gluttony. However, eating too much pie is okay because the sin of pi is always zero. 🖖

I didn't get to talk about this as much as I wanted at #SC24 😆 I've been particularly fascinated by the possibility of emulated FP64. Take a look at this paper! arxiv.org/abs/2411.12090

I wish my body would refuse to execute unsigned code #SC24

2025 todo list: Heat water to boiling point Spin wheel for electrons Electrons heat up accelerators Water cools down accelerators Data and Intelligence come out Looking forward to sorting out the rest of the details at #SC24 this week!