← Back to context

Comment by addedGone

4 days ago

It has nothing to do with "vibe coding" in the sense you imply, you see the ARC-AGI, DeepSWE benchmarks and so-on, assume that will be 99.9% soon when we have Opus 9 which will be soon, which means in practice that most SWE are beaten by top models, which means that letting humans code will be more dangerous (and inferior) than letting models code, we are already close to that which is why we must change the way we work now, not the moment it comes.

It's already dangerous today to not review codebases by top models, how do you think that will be in 2028?

And yes, a simple combination of top models today in adversarial+loops... and only 20 subscriptions will do much better than Codeberg ($100K monthly token equivalent).

But sure, Codeberg can remain alive for all the legacy stuff, but imo the mentality is quite backward.

5 comments

addedGone

Reply
ethin  4 days ago

This comment is full of both a lot of hyperbole and theoretical outcomes that Idk even where to begin. There for one is no guarantee that Anthropic will remain around long enough for them to get to Opus 9 let alone a better model for one. The same could be said for OAI.

And again, what is the problem with a platform wanting to not be flooded with a bunch of low-quality code that gets abandoned after it's done what the creator wanted it to do? I keep seeing this claim of these models being better than "most" developers but have yet to see any actual statistics cited to back up this claim which makes me think it's all just vibes.

  • addedGone  4 days ago

    It's not low-quality code, that's the thing, yes, you can do trash, like most humans do, but it can also output better code, consistency... than most humans today, what will it be next year?

    Anthropic isn't going away, even if they do, a lot will compete anyway, AI in a general sense will reach the point where our jobs as dev will be to prompt all day long and that's it. It's almost sure that we will keep having better models and tooling, right now every single month it's getting better and better, and on some aspects not just linearly.

    the moment GPT-3.5 dropped, it was already not hard to picture literally what we do today, the "hmm what if THIS suddenly is actually accurate", and tomorrow, it will be all the pain points we currently have that are solved.

    Genuinely, see the Bun porting? Imagine that is with a very "weak" model, weak tooling, weak hardware, now imagine that tomorrow, it will port it 50x faster, more performant, better consistency, docs, and virtually zero bugs: not hard to picture at all, it's what it will be imo in 1 year, 2 tops just if the current pace continue, that's not even accounting for the thousand of new companies coming, can we really say it's vibe trash when it outperforms humans?

    Think of what's slowing you down today and why you aren't already a Bot typing non-stop and imagine those solved, now what?

    I must add Kimi K3 as another example, nothing "wahou" with it in the sense that it was already a sure thing, it's so great to see it coming already, but it's also sure we will have Mythos 10 level open-source, open-weight in a few years, it won't be surprising I mean, it's just logical. We will also have Marvel-level movies done with a prompt.

    PS: sorry it's all over the place and out of touch with the initial Codeberg thing