Show newer
Simon boosted

God no one can argue with numbers the way I can though. Sitting in these offices in such severe pain and my stack of research papers and my logs. I am the reason I got care, over and over again. Paying for the wrong tests as a price for getting them to give me the right ones finally. Knowing how to spot bad interpretations of results.

I do not feel that I ever sat in those offices as an actual patient receiving actual guidance and expertise. What would that be like.

Show thread
Simon boosted

@catsalad the pipegreppers union! there’s so many things to love … no idea where I got it tho

Simon boosted

Germany-based #SovereignTechFund contributes 384.000 EUR to #OpenStreetMap over two years. And the best part: it's not for fancy features big sponsors think they need, but for technical debt removal, documentation, testing infrastructure, vandalism prevention, and two additional paid roles.

@sovtechfund once again made a great choice and continues to follow their strategy of sustainable support of critical #OpenSource projects, which @openstreetmap definitely is.

Src: blog.openstreetmap.org/2024/12

Simon boosted
Simon boosted
Simon boosted
Simon boosted

so many #threat modeling workflows are uncivilized, creaky, positively antediluvian.

#threatmodeling should be modern, configured as code, a creative, collaborative romp to reify a defensive strategy that outmaneuvers attackers.

thus, this yule, my deciduous.app co-conspirator @rpetrich and I bear a gift: Deciduous-VS, a #VSCode extension to build and visualize decision trees within your IDE 🎄 (== local dev for classified/regulated envs, too)

learn more in my post: kellyshortridge.com/blog/posts

Simon boosted

I want to be like the cool kids and understand how to do fun artsy hobby things with electronics. I'm just putting that out there. 2025 wish list 🥺

Simon boosted

lead: a story of courage that likewise involves scientific brilliance as dramatic as realizing *most measures of this entire phenomenon were tainted on a global level, that's how big the problem was*, all the time I think about how Claire Patterson knew what everyone was denying

Show thread
Simon boosted

By far the best coverage of o3 is this essay by François Chollet, it's crammed with interesting insights beyond just reporting on the benchmark score: arcprize.org/blog/oai-o3-pub-b

Published my own notes on that here: simonwillison.net/2024/Dec/20/

Show thread
Simon boosted

@grimalkina @felipe Dr. Paige Harden (or kph3k as she is known sometimes) is an amazing communicator about all things behavior-genetics. I'll tell you a short story in the next post.

The short answer is that it's not surprising that variations in the recipe for building a brain (DNA) might have some impact on how those brains vary in their expression of a trait like conscientiousness. In fact, this is such a trivial mechanic that we see non-zero heritability in basically any trait (that is, in anything that varies across people). It's a consequence of the notion that there are no uncaused causes. DNA + environment are the causes of our behavior.

So a non-zero or even high (like h2=.50) heritability doesn't tell us that much. And it's such a complex causal web from DNA -> Behavior that even knowing the hundreds or thousands of genetic variations that are associated with variations in conscientiousness basically gives us no information about how biology impacts it. As a former fMRI neuroscientist, it's basically the same story with brains. We know that brains do the behavior, but knowing some brain-behavior associations actually isn't that illuminating re the behavior.

Re twin studies, they do have their problems but especially at the level of their general conclusions they are pretty solid, as far as I know. There is a lot to argue about in the details though. :D

Simon boosted

I figured out a prompting pattern for getting Claude to produce fully self-contained Python scripts that execute with "uv run" using PEP 723 inline script dependencies - and now I can one-shot useful Python utilities with it simonwillison.net/2024/Dec/19/

Simon boosted

Pull quotes are advertising. They advertise the article itself. I can see them as marginally useful as you flip pages in a magazine, scanning quickly as you're deciding whether to read the article.

The affordance of a single page web article seems different. I doubt people scroll down an article looking at the pull quotes instead of just reading the first paragraphs.
(1/2)

@simon @feoh in enterprise/ business accounts, an admin needs to enable it for the org first.

Simon boosted

GitHub added a permanent free tier for GitHub Copilot today, including access to both GPT-4o and Claude 3.5 Sonnet. My notes here: simonwillison.net/2024/Dec/18/

@simon wow, cursor has been around for longer than I thought...

@dbread datomic seems pretty cool. sqlite is neat, I don't know about cool.

@carnage4life just tried to open an article on Firefox mobile (incognito mode), there is a banner on top but it's not so bad. Maybe they tweak the behavior depending on browser etc?

Show older
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.