TJSomething

I think this is mostly just software benchmarking. The most science that can probably be extracted from this case study is probably a couple data points on where we are and where we can improve with AI agents. It seems like a decent task to optimize for. It might also work as a proof of concept for an economics study or theoretical political science.

It's possible that any such technology would be too complicated to ensure a clean supply chain or have weaknesses that would allow it to be tracked or jammed. Real devices have some kind of engineering tradeoff.

But I generally think that trying to solve social problems with technology is a futile endeavor. Someone will always have violence on their side. And violence is the ultimate weakness.

Is this actually how water pumping works, or are we extending the metaphor to imply that the AI is creating knowledge? Because I would dispute the latter if the effect is that the number of people with a deep understanding of a subject declines.

There's the grandmother hypothesis that holds we could just not have menopause and reproduce more, but we don't because there's greater evolutionary advantage for having the elderly provide support to their kin. Maybe this type of kin selection is bounded by consanguinity though, and that limits the advantage? Or maybe population pressure discourages family units from getting too large?

It seems like Utah could do mostly do this by intercepting all consumer traffic.

In my experience, that was pretty much true until I started doing small discussion-group-style graduate classes. But those are necessarily capped by staff-to-student ratios.

Sounds like I'm not buying a computer or a phone for 2 years. I hope my Pixel 6 lasts that long.

There could be a link there. I've known people with eating disorders who only drink spirits because of the calorie content.

I've heard that it's fairly unreliable and tends to have netsplits and desyncs pretty regularly, unlike XMPP.

Probably to similar purposes as the 2024 opioid settlement: housing assistance, schools, subsidized child care. Legislators will probably try to align it with goals that seem like they would counter the stuff Meta's doing. Smaller communities will probably fail to allocate it promptly.

I'm a little unclear on the timeline here, but Mullenweg just got back from Burning Man. I suspect that's related to both the content of the post and the board's timing.

One thing that's a occasionally annoying (because I deleted my account) but understandable is that a lot of the non-English speakers have stayed on X because it's easy for them to ignore the Nazis.

It seems to me that the law in the US is set up to only establish standards of negligence after a bunch of people die.

I used to think this but it seems to me that the most scarce and valuable resource will be having humans pretend to care about you.

Chronological feeds tend to be gameable by growth hacking. Still better than a lot of things.

One of the tricky things for me is feed aggregation for balancing subscriptions between people who post a lot and people who don't.

It also might be interesting if video recommendations were curated feeds by the creator.

It's telecom infrastructure to increase the number of people who have internet in order to increase the number of people using Google. They might have been considering having satellite internet in Android phones.

This is like half of all job listings I've read recently. And it's a decent amount of fintech that's like this.

I'm assuming that the most relevant geoengineering is worldwide stuff like blotting out the sun with microsatellites at L1 and dumping iron or lime into the oceans to increase CO2 absorption.

I haven't tried it, but I saw Leanstral, an LLM specialized in writing Lean proofs, posted on HN recently and it claims to outperform some larger general purpose models. It didn't beat Claude Opus, but it seems to do decently at one tenth the cost. It's plausible that further research could yield other models that are smaller and more effective at limited tasks, reversing the trend of ever growing models.

If open frontier models start closing up and states start more export controls on AI services and hardware, it might be good to ensure the supply chain is there to reproduce the SotA, or even a couple generations behind it.

I mean I feel obligated. I'd feel bad if I skipped voting in an election where I found out after the fact that one of the candidates kicked puppies.

I'm not sure that follows as the biggest driver. 4 PM is nearly as high as as 6 PM. You would expect a big jump at 5 PM, but the biggest jump is from noon to 2 PM. Just looking at today's temperatures on my front porch in Reno, it was 93°F at 2 PM, it peaked at 95°F at 3:30 PM, and it didn't fall back down to 93°F until 5 PM. Some of that sustained power usage probably is people getting home, but a lot of it is A/C.

While Janet pulls from a few inspirations, the syntax is pure Clojure. I always figured that it was trying fix up the bumpy parts in Fennel to enable a programming style that was more consistently Clojure-like and functional than could be done in Fennel, since Fennel ultimately has to use Lua's semantics because Fennel compiles to Lua.

I feel like this supposes a better world where most people are educated, rational, and calm investors with the time, money, and energy to actively manage a diversified portfolio.

Or I guess you could get there by abolishing the stock market as it exists, which seems more likely.

"Chicxulub impact" seems to be functioning as a bit of hyperbole to imply that this collusion was absolutely devastating, by analogy to the K-T extinction event 66 million years ago.

Not that I really can tell what this was devastating to. Maybe United States v. Apple (2012), where Hachette Book Group, Inc., HarperCollins publishers, Macmillan publishers, Penguin Group, Inc., and Simon & Schuster, Inc. conspired with Apple to raise ebook prices?

I mean, it might have been at first, but Microsoft figured out that the majority of users for lists without formulas in 1993 and they've strategized around that. IMHO, the biggest concession to this was when they added Power Query to core Excel in 2016.

I love moving all my laundry to the dryer and forgetting to start the dryer. Probably happens once or twice a month.

The short answer is that the lambda calculus computes transformations on digital values while this is for building functions that can transform continuous (complex) values.

A lot of clothes aren't particularly practical. I'd argue that artifacts are more likely to be art if their purpose is more aesthetic than practical. As such, I'd say that some articles of clothing are art (like anything you see on a runway), some aren't (e.g. basic T-shirt), and some seem ambiguous to me (some ties might be art, but some are pretty boring and they don't have much practical use).

The big win I usually hear from family who use a lot of miles is on upgrading seats for free, which is really great because they have joint issues and fly fair number of international flights. But I think they also maintain a spreadsheet with a rotating schedule of like 10 credit card companies that are cycled (or maybe shifted between based on who has the best deal for a given good or store in any given month?) for maximum points.