moomin

4 ms·

31.
by moomin 2mo ago
Of those, the weakest points for Haskell and the MLs is the platform and documentation. However, I don’t know if we can call this solved just yet. We may need to invent whole new types of guardrail.
32.
by moomin 2mo ago
Yeah, I’m not convinced that there are any models as smart as Fable. Opus 5 definitely isn’t for all it has great benchmark scores. Fable displays judgement in a way I haven’t seen from any other model.
33.
by moomin 2mo ago
The authors make some good points: compile time and test speed matter, platforms matter. But they really dodge the whole “guardrails matter” thing. And guardrails are going to win long term. As for readability, the fact that AI-written Go c
34.
by moomin 2mo ago
The truth is all frontier models are closed. This isn't even an open source vs open weights thing. Even if you accepted that open weights are "open" the capital requirements for running your own Kimi 3 model are significant.
35.
by moomin 2mo ago
The real truth is that's how Edison used them and they've been the primary use of patents since inception.
36.
by moomin 2mo ago
Only in the rather reductive sense of “the vast majority of gamers are into other genres now”.
37.
by moomin 2mo ago
Of course I, and most of the dads I know, optimise for something else: what’s the car that will keep me competitive, but probably losing, to the kids?
38.
by moomin 2mo ago
Not all of them. The ones on ridiculously restrictive visas had to stay.
39.
by moomin 2mo ago
This massively understates what you could do with the basic 1-bit arrangement. People managed to get music playing during gameplay (I think Manic Miner was first) and even managed to play music with two channels and percussion (Robin of the
40.
by moomin 2mo ago
Yeah, the Jacobian conjecture counter-example was big news. In particular, it would have been news even if an AI hadn't done it. That's where the bar is now. Settling Erdős conjecture 7529 or whatever no longer qualifies as AI new
41.
by moomin 2mo ago
I’ve had a lot of success with this. Give it a good enough harness, leave it for a few days and see what it finds that actually holds up.
42.
by moomin 2mo ago
Indeed, one useful thing you can do is fire questions like "I've got three different solutions to this problem, which performs best?". You don't need to keep the code, you just want to know which direction you should tak
43.
by moomin 3mo ago
I kid you not, I remember writing an example solution with comments explaining how we were approaching the problem. Copilot wrote most of the actual code via autocomplete. And that was a year ago, models have got a lot better since then.
44.
by moomin 3mo ago
I don’t think it changes that much. For opus-sized tasks, new Opus is the best model. For enormous things like planning and research, Fable is still the model that can concentrate for longer.
45.
by moomin 3mo ago
It is wild that, 80 years after a girl was murdered, she’s relevant to a court case involving technology it would have taken over an hour to explain in 1945.
46.
by moomin 3mo ago
I think a better way of describing it would be, someone broke into your house. The police said they'd patrol it, but it would take 20 years to set up. In those 20 years, no-one broke into your house. You might suspect they had other re
47.
by moomin 3mo ago
It turns out that sometimes you really do want health and safety obsessed bureaucrats.
48.
by moomin 3mo ago
No, they're just owned by people. Most of whom aren't billionaires.
49.
by moomin 3mo ago
True Capitalism looks a lot more like socialism than many would like to admit.
50.
by moomin 3mo ago
A weird question to ask in a thread about the destruction of evidence.
51.
by moomin 3mo ago
It would be, but humans are famously bad at judging relative areas at all, which is a problem for data visualisation everywhere.
52.
by moomin 3mo ago
TL;DR by controlling the route every car takes and traffic lights, we’d get a 2% reduction in traffic. I don’t think this heralds and great change in transport policy.
53.
by moomin 3mo ago
My experience suggests that they _can_ be good, but this particular pattern they can be remarkably bad at. Source: I keep having to optimise this pattern.
54.
by moomin 3mo ago
I considered it, but I figured that whatever I did there would be inconclusive. Instead I tried to figure out the blast radius of this being proven, and I didn’t get very far with that either.
55.
by moomin 3mo ago
For comedy’s sake, I asked ChatGPT 5.5 about the significance of the problem and the chance that 5.6 would solve it with a three page solution. It said close to zero. I invited it to search the internet and it remains extremely sceptical.
56.
by moomin 3mo ago
This is actually great, and I predict that fans of nil-punning will rapidly discover the joys of actually having errors trigger where the error was introduced rather than propagating through the program. Any news on ClojureScript gaining th
57.
by moomin 3mo ago
Don’t worry, some of us remember Y2K and a) how much we fixed b) how much went wrong on the day and c) getting told it was a waste of time later. And I didn’t even have to deal with a jumpy national government.
58.
by moomin 3mo ago
Yes, there’s been a very popular narrative that Mythos’ abilities are just marketing fluff. I think it’s clear that there’s a real capability here, even if Anthropic’s communications have been heavily influenced by PR concerns.
59.
by moomin 3mo ago
This is one of those little things I’ve had trouble putting my finger on: the US eats surprisingly few oats. The U.K. eats more than three times as much on average. Which is probably one of the reasons I find US cuisine slightly uncanny val
60.
by moomin 3mo ago
I’m not a follower either, but inference from the article tells me: Krammik has come up with some sort of cheat detection method, has loudly accused some others of cheating, and FIDE are both unconvinced of the truth of the allegations and

More ›