wintercarver

From one of the authors at MIT: “ We made a striking discovery: AI agents can invent and build without talking to one another, and their technologies outlive the creators. A swarm of hundreds of initially identical agents spontaneously differentiates into explorers, builders, caretakers, and coordinators - without direct communication.”

Thought this was pretty fascinating!

https://x.com/profbuehlermit/status/2093630309585531033?s=46

Your humble repo owner here: I wanted to compile a list of prompts used in math and science research in a single area to make review of said prompts easier, if only to hope that some "methods" or styles could be reviewed and applied.

Is quite nascent, but with the jacobian conjecture news (amidst other recent AI-driven results) and odds that folks may go burn some AI credits on trying their hand at random open problems, I thought I'd share.

You can clone the repo and have your agent review some of the styles or prompt guide summary to try and refine an approach.

Can confirm - worked on Super-K as a grad student and a regular part of shift work was to go through simulated supernova alert testing. Was always kind of fun, in the lotto ticket sense (alarm goes off, review to see, ah yes, simulated test, back to shift monitoring).

Would love to read any of the scientific papers that McCarthy supposedly edited diligently, should anyone happen to know of some. Very curious to see what they read like.

First thing that came to mind when I read this was the quote attributed to Haruki Murakami "If you only read the books that everyone else is reading, you can only think what everyone else is thinking." Turns out this is a quote from a character in one of Murakami's novels, but sentiment stands. Kind of gets to the essence of what "creativity" may be.

Nice post! Enjoy your blog's overall aesthetic too. Perhaps correlation in sense of style, though, as I also used RTK to learn to write Kanji, loved it, and now, ~15 years after that escapade, am kind of in a similar bucket (can write some characters, but mostly just read). I still think RTK great overall method and would do it again!

Also, shoutout to Fabrice, creator of Kanji Koohii -- that was my first foray into SRS back in ~2007/2008, after which I found Anki (pre-mobile).

Wait WHAT?! I’ve been buying ebooks off BookWalker and it’s so annoying. Their reader app sucks and they have these faux-protective constraints like not permitting a “copy” function on any highlighted text and limiting a book to no more than 50 highlights. It’s so silly… Edit: damn, had assumed you meant maximum nerd mode and that these books were available in Japanese… looks like it’s all English translations. So it goes.

Nice and thank you! Will now go head my head in shame for not having bothered to click the “menu” icon on the main report :facepalm:

Being lazy here, but would love to know more about how testing for all of these plastics chemicals that are omnipresent is done in a way that ensures the measurement process or tools themselves do not contribute trace chemicals (e.g. lab tech wears latex gloves while handling the sample, whoops, etc).

Congrats David & the whole team! Really enjoy everything Sakana AI produces and always look forward to your research results.

If you enjoyed this article I’d recommend checking out the book _Moonwalking with Einstein_ by Joshua Foer. Great read and fascinating dive into the lives and practices of individuals that participate in memory competitions.

Data team manager here, qq: how do you handle type conversions between different systems? Nice work overall!

For someone familiar with the science can you comment on the typical uncertainty/variation in epigenetic age measurement? I love error bars, wish they were mandatory.

Plus one, scrolled through waiting for someone to mention this one.

I don’t typically reread fiction but the poetic style of writing, combined with great scenes and characters makes almost any page worth a reread from time to time.

Sometimes I wonder if I’ll ever find a book as good across so many dimensions…

Awesome, I read this years ago and have always kept it in mind ever since. Just wanted to say thanks for writing/co-authoring it!

I think this quote from a recent New Yorker article[1] sums up a subtle difference that’s easy to overlook. Topic has been on my mind a bit lately[2]:

“Chandlee, who spent more than twelve years at Mark Zuckerberg’s company before moving to TikTok, dismissed the idea [of concern over competition from Facebook]. “Facebook is a social platform. They’ve built all their algorithms based on the social graph,” he said, referring to the network of links to friends, family, and casual acquaintances that Facebook users painstakingly assemble over time. “We are an entertainment platform. The difference is significant.””

[1] https://www.newyorker.com/culture/cultural-comment/tiktok-an...

[2] self plug, short set of questions on the recent Instagram makeover flop: https://kvncnnlly.bearblog.dev/copy-to-copy-copy-to-change/

I thought it was a nice summary, concise, organized, with examples and references. Will revisit it should I need a reminder on scraping. Would not call it low quality at all.

Would recommend you ignore passing comments with no constructive criticism. The title is going to be a point of contention as it’s a big claim and probably being misinterpreted as not “everything you need to know [to get started]” but rather “everything you need to know [ever is in this one article and you’ll need not read anything else]”.