pron

29 ms·

391.
by pron 7mo ago
1. My whole point was that it is not about anyone in the affected group taking offence. The question is whether other people can come to devalue people in the affected group. In this context "offensive" doesn't mean takin
392.
by pron 7mo ago
Somewhat on a tangent, but when people talk about offensive language in the context of cultural criticism they don't mean terms that cause the people who hear them to be offended but things that may diminish the value of some people in
393.
by pron 7mo ago
> The thing is for most places the kind of code they write is good enough. The kind of code they write is the kind of code that will be unsalvageable after 10-50 changes. That's throwaway code, although it looks good. I don't
394.
by pron 7mo ago
My problem with the code the agents produce has nothing to do with style or art. The clearest example of how bad it is was shown by Anthropic's experiements where agents failed to write a C compiler , which is not a very hard programm
395.
by pron 7mo ago
As someone who's both an IC and leads other developers I disagree with the explanation. As a technical lead, with people I can much better predict the quality of the outcome than with LLMs, and the "failure modes" are much mo
396.
by pron 7mo ago
Except there's a much simpler definition of flying than of intelligence.
397.
by pron 7mo ago
> We're not claiming to solve agentic coding. We're just testing people doing testing things, and we think that good testing tools are extra important in an agentic world. Yeah, I know. Just an opportunity to talk about some of
398.
by pron 7mo ago
Right, I said that property based tests are easier to read, and that's good. But people still have to actually read them. Also, because they still work best at the "unit" level, to understand them, the people reading them nee
399.
by pron 7mo ago
> property-based testing is going to be a huge part of how we make AI-agent-based software development not go terribly. There's no doubt, I think, testing will remain important and possibly become more important with more AI use, an
400.
by pron 7mo ago
I think Lattner was too generous and missed a couple of crucial points in the CCC experiment. He wrote: > CCC shows that AI systems can internalize the textbook knowledge of a field and apply it coherently at scale. Except that's no
401.
by pron 7mo ago
You mean trying and failing to build a C compiler. This isn't a very hard task to begin with (assuming you know compilers, and the models do), but it was made unrealistically easy by giving the agents thousands of tests written by hu
402.
by pron 7mo ago
Is that the goal, though? "Subduing the Iranian regime"? If so, shouldn't they explain that so that the war gets support and so better chances of achieving its goals? How is that goal to be achieved? Because so far it seems l
403.
by pron 7mo ago
If they could they would, and if they can they will. Maybe it will appear next month, and maybe 5 years from now, and we don't know which of these is more likely. But I think that if agents could actually produce good, reliable softwar
404.
by pron 7mo ago
You assume that the agents can produce okayish software that isn't very resilient, but that's not quite what we're seeing. Remember Anthropic's attempt to have agents write a C compiler (which isn't a super-complica
405.
by pron 7mo ago
Except I find that agents are much better at debugging than writing code. When the code gets to the state where the agents can no longer debug it and keep it running (as happened in Anthropic's failed C compiler attempt) humans will li
406.
by pron 7mo ago
Here's what I find difficult to reconcile with my own experience. I've been using Codex in anger for the past 2 months or so (with gpt5.3-codex and then gpt5.4) on projects of different complexity. It is quite good at debugging, b
407.
by pron 7mo ago
No, just a 20+ year C++ and Java developer, while you clearly haven't used modern Java. Now, I admit that because I have a lot of experience in low-level programming, I am often able to beat Java's performance in C++, but not with
408.
by pron 7mo ago
Well, JS is fast and Go is faster, but Java is C++-fast.
409.
by pron 7mo ago
You mostly need a recent JDK. Leyden has already cut down warmup by a lot and is expected to continue driving it down. https://foojay.io/today/how-is-leyden-improving-java-perform... https://quarkus.io&#x2
410.
by pron 7mo ago
From your description it sounds like you're comparing an old Java program running on an old JDK with a new Go program running on a new Go runtime. These days Java has virtual threads (similar to Go's goroutines) and doesn't n
411.
by pron 7mo ago
Use javac directly, in a recent JDK, to compile lots of files at once (i.e. one invocation of javac to compile multiple files) and you'll see. The speed is over 100,000 lines/sec: https://mill-build.org/blog/1
412.
by pron 7mo ago
And Java has some advantages that are hard to beat. It performs better than Go and builds just as fast (at least when not using some popular build tools) while offering unmatched deep, low-overhead observability. People might think they may
413.
by pron 7mo ago
> The code must pass property-based tests Who writes the tests? It can be ok to trust code that passes tests if you can trust the tests. There are, however, other problems. I frequently see agents write code that's functionally co
414.
by pron 8mo ago
> But the problem is that Church–Rosser Property[0] (proofs ~= programs) and Brouwer–Heyting–Kolmogorov Interpretation[1] (Propositions at types) are NOT binary SAT, and you have concepts like mere propositions[3] that are very different
415.
by pron 8mo ago
> For most cases we don't need nearly that comprehensive verification. This is expecting more off AI written code than we ever bother to subject most human written code to. True. > There's a vast chasm there we only need to
416.
by pron 8mo ago
First, it's not a question of decidability but of tractability. Verifying programs in a language that has nothing but boolean variables, no subroutines, and loops at depth of at most 2 - far, far, from Turing-completeness - is already
417.
by pron 8mo ago
Validating programs against a formal spec is very, very hard for foundational computational complexity reasons. There's a reason why the largest programs whose code was fully verified against a formal spec, and at an enormous cost, wer
418.
by pron 8mo ago
But the code produced from the formal spec would still be nondeterministic. And I believe CodeSpeak doesn't wish to regenerate the entire program with each spec change, but apply code changes based on the changes to the spec. Maybe the
419.
by pron 8mo ago
> copyright governs copies, it doesn't govern ideas or algorithms The second part is true. The first is a little trickier. The copyright applies to some fixed media (text in this case) rather than the idea expressed, but the prot
420.
by pron 8mo ago
If what you're after is determinism, then your solution doesn't offer it. Both the formal specification and the code generated from it would be different each time. Formal specifications are useful when they're succinct, whic

More ›