I would claim that LLMs desperately need proprietary code in their training, before we see any big gains in quality.
There's some incredible source available code out there. Statistically, I think there's a LOT more not so great source available code out there, because the majority of output of seasoned/high skill developers is proprietary.
To me, a surprising portion of Claude 4.5 output definitely looks like student homework answers, because I think that's closer to the mean of the code population.
This is dead wrong: essentially the entirety of the huge gains in coding performance in the past year have come from RL, not from new sources of training data.
I echo the other commenters that proprietary code isn’t any better, plus it doesn’t matter because when you use LLMs to work on proprietary code, it has the code right there.
Author attributes past year's degradation of code generation by LLMs to excessive use of new source of training data, namely, users' code generation conversations.
Yeah, this is a bullshit article. There is no such degradation, and it’s absurd to say so on the basis of a single problem which the author describes as technically impossible. It is a very contrived under-specified prompt.
And their “explanation” blaming the training data is just a guess on their part, one that I suspect is wrong. There is no argument given that that’s the actual cause of the observed phenomenon. It’s a just-so story: something that sounds like it could explain it but there’s no evidence it actually does.
My evidence is that RL is more relevant is that that’s what every single researcher and frontier lab employee I’ve heard speak about LLMs in the past year has said. I have never once heard any of them mention new sources of pretraining data, except maybe synthetic data they generate and verify themselves, which contradicts the author’s story because it’s not shitty code grabbed off the internet.
> it doesn’t matter because when you use LLMs to work on proprietary code, it has the code right there
The quality of the existing code base makes a huge difference. On a recent greenfield effort, Claude emitted an MVP that matched the design semantics, but the code was not up to standards. For example, it repeatedly loaded a large file into memory in different areas where it was needed (rather than loading once and passing a reference.)
However, after an early refactor, the subsequently generated code vastly improved. It honors the testing and performance paradigms, and it's so clean there's nothing for the linter to do.
Progress with RL is very interesting, but it's still too inefficient. Current models do OK on simple boring linear code. But they output complete nonsense when presented with some compact but mildly complex code, e.g. a NumPyro model with some nesting and einsums.
For this reason, to be truly useful, model outputs need to be verifiable. Formal verification with languages like Dafny , F*, or Isabelle might offer some solutions [1]. Otherwise, a gigantic software artifact such as a compiler is going to have a critical correctness bugs with far-fetched consequences if deployed in production.
Right now, I think treating a LLM like something different than a very useful information retrieval system with excellent semantic capabilities is not something I am comfortable with.
I firmly agree with your first sentence. I can just think about the various modders that have created patches and performance enhancing mods for games with budgets of tens to hundreds of millions of dollars.
But to give other devs and myself some grace, I do believe plenty of bad code can likely be explained by bad deadlines. After all, what's the Russian idiom? "There is nothing more permanent than the temporary."
I'd bet, on average, the quality of proprietary code is worse than open-source code. There have been decades of accumulated slop generated by human agents with wildly varied skill levels, all vibe-coded by ruthless, incompetent corporate bosses.
Not to mention, a team member is (surprise!) fired or let go, and no knowledge transfer exists. Womp, womp. Codebase just gets worse as the organization or team flails.
It doesn’t matter what the average is though. If 1% of software is open source, there is significantly more closed source software out there and given normal skills distributions, that means there is at least as much high quality closed source software out there, if not significantly more. The trick is skipping the 95% of crap.
yeah, but isn't the whole point of claude code to get people to provide preference data/telemetry data to anthropic (unless you opt out?). same w/ other providers.
i'm guessing most of the gains we've seen recently are post training rather than pretraining.
Yes, but you have the problem that a good portion of that is going to be AI generated.
But, I naively assume most orgs would opt out. I know some orgs have a proxy in place that will prevent certain proprietary code from passing through!
This makes me curious if, in the allow case, Anthropic is recording generated output, to maybe down-weight it if it's seen in the training data (or something similar)?
This is cool and actually demonstrates real utility. Using AI to take something that already exists and create it for a different library / framework / platform is cool. I'm sure there's a lot of training data in there for just this case.
But I wonder how it would fare given a language specification for a non-existent non-trivial language and build a compiler for that instead?
If you come up with a realistic language spec and wait maybe six months, by then it'll probably be approach being cheap enough that you could test the scenario yourself!
I see that as the point that all this is proving - most people, most of the time, are essentially reinventing the wheel at some scope and scale or another, so we’d all benefit from being able to find and copy each others’ homework more efficiently.
..A small thing, but it won't compile the RISCV version of hello.c if the source isn't installed on the machine it's running on.
It is standing on the shoulders of giants (all of the compilers of the past, built into it's training data... and the recent learnings about getting these agents to break up tasks) to get itself going. Still fairly impressive.
On a side-quest, I wonder where Anthropic is getting there power from. The whole energy debacle in the US at the moment probably means it made some CO2 in the process. Would be hard to avoid?
I think you’re overstating the effect. The most volume is sold at supermarkets which have the best location for throughout but they also have the cheapest prices.
There’s only so much investable capital available, if it is going to hardware stocks it’s got to be coming from somewhere else. It’s just a substitution toward hardware tech stocks. Economics 101.
Depends what we mean by specialist. If it frontend vs backend then maybe. If it general dev vs some specialist scientific programmer or other field where a generalist won’t have a clue then this seems like a recipe for disaster (literal disasters included).
Social movements don’t just happen from grassroots these days. They’re seeded by foreign states. A simpler solution would be require ids for social media posting. If you don’t provide an id you get a limited number of views.
And I don’t see anything wrong with a preventative system in principle, we should be able to join up social services information with policing, because we have had cases where a mass murderer has been known to multiple services.
Edit: probably not ids but a token that verifies my nationality would be enough.
> A simpler solution would be require ids for social media posting
It’s strange times when even the comments on posts about government overreach are calling for more government overreach and limitations on speech and privacy.
Do you really want to have to verify your ID to post anything online, including HN?
And I am willing to bet that on top of the chilling effect on regular people, it will only act as an inconvenience for the bad actors as they will find ways to circumvent it. Controlling the online discourse is far too valuable, they are not going to just shrug and give up because the government puts up a barrier.
Running a git command on one branch and multiple branches being affected is really unusual for me! This really does look like it is designed for just this problem, though. Simple overview: https://blog.hot-coffee.dev/en/blog/git_update_refs/
How? I tried recreating the scenario from the article (the section "First rebase –onto") and ran the first rebase with "--update-refs":
$ git checkout feature-1
$ git rebase --update-refs main
Successfully rebased and updated refs/heads/feature-1.
Updated the following refs with --update-refs:
refs/heads/feature-2-base
But all it did was update feature-2-base. It still left feature-2 pointing to the old commits. So I guess it automates "git branch -f feature-2-base feature-1" (step 3), but it doesn't seem to automate "git rebase --onto feature-1 feature-2-base feature-2" (step 2).
Yeah, you need to rebase the tip of the feature branch stack. git will then update all the refs that point to ancestor commits that are moved. So in this case
The article suggests there’s evidence that screen time has the opposite effect. A little surprising but I guess for a lot of people it is more stimulating than watching the news or soaps all day
In the U.K. I was betting 5 minute binary options back in 2008 and parlays or accumulators as we call them (accys for short) have been popular for a while too.
Rightly or wrongly, The Gambling Act 2005 put the UK literally decades ahead of places like the US in terms of creating a legal framework for sports betting/gambling in general.
The forces that made those shops appear and made the greengrocers disappear are not natural, inevitable, and foolish to resist. They are just laws, laws that permitted some things and discouraged others, taxed some things and subsidized others.
In my experience there’s still much more to this. I’m sure it helps at population level like the article describes but it’s not foolproof. For our first we were feeding nuts early and still developed an allergy to all nuts. Our second didn’t get nuts until much later and he’s fine. There’s more to the story than timing, notably my first has eczema and asthma too so there’s that atopic march.
> There’s more to the story than timing, notably my first has eczema and asthma too so there’s that atopic march.
Eczema often comes with digestive issues, bowl inflammation, loose stool, blood in the stool etc.
Eczema essentially gives you wounds, if you allow allergens to enter the bloodstream directly without going through the digestive tract you are at an increased risk of developing allergies.
For kids/babies with these kind of issues it's probably better to delay introducing common allergens until their gut can heal or you will end up causing allergies rather than preventing them.
Allergy rate decreases with birth order. Of course, that's at the population level and probably not strong enough effect to notice if you only poll a dozen parents you know.
reply