It was a great achievement. My wife isn’t an LRR or PA fan, but we watched the whole thing through together because it was just great television.

I feel like my uninformed guess would be that it would pull up the low end of the distribution? LLMs are unlikely to produce Daily WTF-style code. Otherwise, though, I have very little idea what it would do to the distribution.

When you copy something, the copy is created in the same zone as the original. Complete the Circuit copies spells on the stack, so the copies are on the stack. Mizzix’s Mastery copies cards in exile, so the copies are in exile.

If you see press reports that says a company “has raised” some money but no date on when the round closed, it probably happened some time in the past. Bluesky is actually unusual in disclosing a date that’s so far in the past.

11 months is unusually long, but the previous startup I worked for waited about 8 months to announce their series A. Current startup waited a few weeks. A friend’s startup waited 6 months to announce their seed round. There’s lots of waiting happening.

That said, 11 months is a longer delay than usual, and I’m not sure why now would be the moment they decided to announce it. So that part is kind of weird.

@gruber.foo Waiting to announce funding rounds for an opportune moment is becoming pretty common for venture-backed startups. Fundraising is one of the biggest press hits you can get, so people want to time it so they can take the best advantage of it.

That’s actually the only product that exists in most countries, the 30-year fixed-rate mortgage only exists due to subsidy. Here in the UK, mortgages are 3-5 year fixed rates and then float with the Bank of England rate (though you can get a new fixed rate term at that point).

Huh. I’ve heard a lot of chatter about jj, but I’ve never heard someone say jj isn’t modal. And now that I have, that actually makes me interested in checking it out, more than everything else I’ve seen. Might just be me, but I think you should talk about that more!

You know, Trump’s zero-sum thinking is usually problematic because he keeps screwing up things that are good for other people on the grounds that they must be bad for him, but I didn’t think it would also mean he can’t recognize that things that are *bad* for others can also be bad for him.

One was an upholsterer and the other was a civil engineer.

We recently got a pair of kittens, and I had forgotten just how much housework I could get done while procrastinating on cleaning the cat box.

True facts.

Which is not to say they’re bad engineers, I just think the modal experience of software involves very different tradeoffs than the ones you typically operate under.

(Which is honestly why I’ll probably apply to Oxide someday, I like that world better! I just think it’s uncommon.)

Maybe I’m wrong, but I would not be surprised if a majority of people writing software today don’t know what a Unix domain socket is or how to create one, nor know what a language server is and how it works. They’re mostly operating several layers removed from all that.

I agree you shouldn’t take yourself out of the loop (for lots of reasons), I was just making the point that “as correct as possible” isn’t actually a goal in a lot of contexts.

Possibly some of this is just that I think most software that’s written isn’t of the sort of your example.

And this is even within the same code base. I work for a delivery company, and our courier login or automated enforcement systems need to be as correct as we can make them. Our courier matching systems, not so much. But the latter probably have more lines of code than the former.

I think the first bullet point is very context-dependent, and I'd be hesitant to say it applies to most code. A whole lot of code is non-critical and easy to update, and in those cases it can be rational to accept less correctness in exchange for faster output.

Rhymes with people complaining about users choosing apps or OSes based on which ones have nice interfaces rather than being more featureful or more open. Aesthetics matter!

I’m regularly reminded of a tweet that observed that some writing has a soul and some does not, and the practitioners of each believe there’s only a small market for the other.

I don’t know enough for a deep dive, but I think a lot is because React repaints everything that might be invalid rather than trying to track exactly what needs repainting exactly. So React apps are less likely to have weird UI bugs, but more likely to be sluggish.

I feel like one of my unusual qualities is I don’t remember stuff like this. Have I ever written a production-destroying bug? Almost certainly! Do I have any memory of doing so? Nope!

I remember a couple big production outages by others, but that’s because they make good stories at parties.

When my dad was at Berkeley, Glenn Seaborg (Nobel prize winner and former chancellor) got randomly assigned to teach the intro chemistry class. They had to find extra rooms for overflow because so many people signed up. Don’t know if that would happen now, but it did once upon a time.

And to be clear, I think that’s fine! Whites and blacks aren’t treated the same in the US, especially historically, so casting a black actor to play a white part is acceptable in ways that casting a white actor to play a black part isn’t, especially given most parts are seen as default white.

I think it’s more complicated than that. The all-black casting of Hamilton was generally at least accepted if not lauded, whereas an all-white casting of a civil rights movement musical I would expect to be criticized. The difference isn’t only that Helen is mythological.

Never mind the jobs you had, tell me five classes you took in college:

1) History of American Film
2) Shakespeare
3) Game Theory
4) War & Diplomacy in Europe, 1815-1914
5) Ethics and Professionalism in Engineering

You read things like this and you can only hope that when the time comes, if it comes, when it is your community's turn to be tested, that you acquit yourselves half as well.

Which isn’t to say that it never matters, just that Oxide has higher requirements for rigor and quality than the vast majority of places, and so these tools can be useful even if they make mistakes like that and nobody notices.

I think the reality is, though, there’s a lot of work where that doesn’t matter. I work on a lot of code where N is 10, max, and so I’ll write an N^2 loop happily because it’s easier to maintain. And if I’m wrong, I can go fix it then once we see it’s slow in practice.

I feel like the dark green territory in the upper left is going to be key. You can focus your forces in there and protect the areas west of it while also threatening to expand along the northern edge. Those eastern bits are going to be a quagmire.