josephg

28 ms·

601.
by josephg 8mo ago
Disabling codegen units = 1 speeds up the compilation further. But it’s still too many dependencies and too slow. The binary is pretty huge too.
602.
by josephg 8mo ago
Only by abusing interface {}. The result is horrible. Go doesn’t have sum types as a first class primitive.
603.
by josephg 8mo ago
What evil deeds are you worried about in particular? What are you afraid people will do now that coreutils is MIT?
604.
by josephg 8mo ago
> Go is a better Rust. Rust is an ugly version of C++ with longer compile times and a band of zealous missionaries. Eh. There's a lot I like about Go. I adore its compilation speed and the focus on language simplicity. But its got p
605.
by josephg 8mo ago
> 10/10 - would generate tens of thousands of lines of useless code again. Me too! A couple days ago I gave claude the JMAP spec and asked it to write a JMAP based webmail client in rust from scratch. And it did! It burned a mountai
606.
by josephg 8mo ago
Yes, but we have a perfectly serviceable term for local software already: "local software". To me, "local-first software" means something slightly different. The term was coined by this essay[1], which says: > Local-f
607.
by josephg 8mo ago
What does local first mean in this context? Does it just mean local? Like, the software runs locally?
608.
by josephg 8mo ago
I did some work ~15 years ago for a consulting company. The company pushes their own custom opensource cms into most projects - built on top of mongodb and written by the ceo. He’s a lovely guy, and good coder. But he’s totally self taught
609.
by josephg 8mo ago
Yep I think this is broadly true. Though its still not clear to me if vibe coding is going to make this better or worse.
610.
by josephg 8mo ago
Yep. Claude today? No way can it achieve this. It can barely write a working C compiler. I'm looking at the trend line. A few years ago it couldn't make a simple webpage. Now it can make a bad C compiler in thousands of dollars of
611.
by josephg 8mo ago
> So we need to pay attention to every detail that doesn't have a single obviously correct answer I really, really wish that were the case. But look at the modern web. Look at iOS apps. Look at how long discord takes to launch on a
612.
by josephg 8mo ago
> What is it about the tests that make you not care about how it’s implemented versus other code? Huh this is a thought provoking question. I think there's a few reasons. In a test suite: - I don't care about performance. - I d
613.
by josephg 8mo ago
I heard a stat on the economist podcast the other day talking about AI music production. They said spotify estimates 40% of songs on their platform are now AI generated. The AI generated songs make up 0.5% of total listening time. Low quali
614.
by josephg 8mo ago
> LLMs always seem to be inarguably worse than the “original”. True. But quantity has a quality of its own. I'm personally delighted at the idea of outsourcing all the boring cookie cutter programming work to an AI. Things like writ
615.
by josephg 8mo ago
> It works, sure, but is it worth your time to use? This is something I like about the LLM future. I get to spend my time with users thinking about their needs and how the product itself could be improved. The AI can write all the CSS an
616.
by josephg 8mo ago
There’s lots of times where I just don’t care how it’s implemented. I got Claude to make a test suite the other day for a couple RFCs so I could check for spec compliance. It made a test runner and about 300 tests. And an html frontend to v
617.
by josephg 8mo ago
And yet, claude is improving at programming much faster than I am. Maybe its skill will hit a ceiling at some point, but it hasn't happened yet.
618.
by josephg 8mo ago
> Token prices have literally gone up, where are you getting this information from. I said like for like. You can't compare GPT5.2 tokens with GPT3.5 tokens. They're different products. You can run local AI models today which c
619.
by josephg 8mo ago
Why do you think that? Do you think human level intelligence is special somehow? Do you have any facts? Or are you just hoping?
620.
by josephg 8mo ago
> there is no way he is a programmer, sorry Lol I've been programming for 30 years. > The complexity of these systems is crazy. Unless he meant ah HTML text area with "save" button - then sure, why not. What do you see
621.
by josephg 8mo ago
> I don't see it in practice though. Have you actually tried the latest agentic coding models? Yesterday I asked claude to implement a working web based email client from scratch in rust which can interact with a JMAP based mail ser
622.
by josephg 8mo ago
Oh? They tried rust? Lots of people seem really committed to OOP. Rust is definitely a bad fit if you can't imagine writing code without classes and objects. I don't think this makes rust is a bad language for the problem. Its jus
623.
by josephg 8mo ago
> Fil-C likely can do the job without a massive rewrite and achieving safety for C and C++. So long as you don't mind a 2-4x performance & memory usage cost.
624.
by josephg 8mo ago
I’m not claiming I know the exact timing. I’m just seeing a trend line. Gpt3 to 3.5 to 4 to 5. Codex and now Claude. The models are getting better at programming much faster than I am. Their skill at programming doesn’t seem to be levelling
625.
by josephg 8mo ago
> cost thousands to millions of dollars in tokens > Even if the price falls by a thousand fold, why would you spend thousands of dollars on tokens to develop an OS when there's already one you can use? Why do you assume token pri
626.
by josephg 8mo ago
> Both of those are fixed, unchanging, closed, full information games. The real world is very much not that. Yeah but, does that actually matter? Is that actually a reason to think LLMs won't be able to outpace humans at software de
627.
by josephg 8mo ago
I think validation is already much easier using LLMs. Arguably this is one of the best use cases for coding LLMs right now: you can get claude to throw together a working demo of whatever wild idea you have without needing to write any code
628.
by josephg 8mo ago
> Governments better pay attention and possibly act before it's too late. Before its too late for what? For OpenAI and Claude to privatise their models and restrict (or massively jack up the prices) for their APIs? The genie is alre
629.
by josephg 8mo ago
> Also, if you are a human who does taste, it's very difficult to get an AI to create exactly what you want. I wonder if it would be possible to fine train an AI model on my own code. I've probably got about 100k lines of code
630.
by josephg 8mo ago
For now. Claude is worse than we are at programming. But its improving much faster than I am. Opus 4.6 is incredible compared to previous models. How long before those lines cross? Intuitively it feels like we have about 2-3 years before cl

More ›