japhyr

The last section focuses on how to use LLMs to make contributions:

> Use an LLM to develop your comprehension.

I really like that, because it gets past the simpler version that we usually see, "You need to understand your PR." It's basically saying you need to understand the PR you're making, and the context of that PR within the wider project.

> One of the main issues is that pointing to your GitHub contributions and activity is now part of the hiring process.

If I were hiring at this moment, I'd look at the ratio of accepted to rejected PRs from any potential candidate. As an open source maintainer, I look at the GitHub account that's opening a PR. If they've made a long string of identical PRs across a wide swath of unrelated repos, and most of those are being rejected, that's a strong indicator of slop.

Hopefully there will be a swing back towards quality contributions being the real signal, not just volume of contributions.

I'd be curious to know what portion of that 40% makes any meaningful income from their open source work. I would guess that most of those people are being paid a small appreciation amount for the work they're doing, not something resembling a living wage.

I have a friend who's fond of saying, "GitHub stars are great for measuring the number of GitHub stars a project has."

That's part of why they are trying to take control of elections, which have (I believe) historically been the responsibility of each state.

I think you are right. I'll edit my comment to point to this.

You're absolutely right that some individuals will be able to sign up for this program, and remember to cancel at the end of the six months. However, when companies choose to implement a policy like this they're acting on well-established statistics. They know that a meaningful percentage of people will forget to cancel, and the company will end up with increased revenue. There might be a bit of good will here, but in the end a program like this with these clearly-spelled-out terms is not much more than marketing.

This feels especially ugly to me because maintainers of large open source projects will feel pressure to keep using tools that let them work in an AI-assisted world. This really feels like it will make life harder for open source maintainers in the end, rather than easier. That's the opposite of what a meaningful open source campaign should look like.

At the very least, it puts maintainers right back in the position of having to beg giant companies for handouts.

At first I thought people here were being pretty unsympathetic to an early version of a beneficial program. I could see a company setting a 6-month timeline initially, so they can reevaluate the program and choose how to evolve their support for open source. I expected to see something along the lines of, "at the end of the 6 months we'll evaluate whether to continue your free plan."

But no, they're quite explicit about this being nothing more than a way to try to get paid subscriptions from open source maintainers:

> Your complimentary subscription will expire at the end of the Benefit Period. After expiration, any existing subscription will continue unless you cancel. You may independently choose to purchase a paid Claude subscription at the then-current price through Anthropic’s standard signup process.

So anyone who participates in this will need to remember to opt out six months from now, or suddenly find themselves with invoices at the max 20x level.

That's pretty ugly.

Edit: I believe I misread the terms. As mwigdahl points out below: "If you have an existing subscription, it pauses while the free period is active. After that free period, your existing subscription resumes. As I read it, there is no "auto-subscribe" after the free period ends -- you just revert back to whatever you had before (or nothing, if you weren't a subscriber before)."

https://www.anthropic.com/claude-for-oss-terms

As someone who earned "passion" money for a long time before ever earning anything remotely close to tech-adjacent money, passion does not pay bills anywhere near as well as money does. And struggling to pay bills, such as paying someone to fix a leaking roof, is not an enjoyable life for very long.

> the real culprit could simply be boiled down to a failure in classroom management and lack of enforcement against cell phones in class

I was a middle school and high school math and science teacher from 1994 through 2019. I watched the advent of internet in schools, then desktop computers in classrooms, and finally smartphones in students' hands.

I've lived a life of watching teachers and schools get blamed for not dealing better with society's issues. "Just teach kids how to use technology", "just ban phones", and "lock down irrelevant websites" is a pretty big ask when the entire industry is focused on getting kids to use these devices, apps, and sites as much as they possibly can.

I'm a firm believer that taking on something new every decade or so of life is an entirely good thing. I've watched so many people stop living in their 30s, 40s, and 50s. My heroes are people who keep doing what they love into their 80s and 90s, and keep finding new challenges along the way.

Yes, that's exactly what I was trying to get at.

I don't get any sense that he's going to put that kind of effort into responding to abusive agents on a regular basis. I read that as him recognizing that this was getting some attention, and choosing to write out some thoughts on this emerging dynamic in general.

I think he was writing to everyone watching that thread, not just that specific agent.

Wow, there are some interesting things going on here. I appreciate Scott for the way he handled the conflict in the original PR thread, and the larger conversation happening around this incident.

> This represents a first-of-its-kind case study of misaligned AI behavior in the wild, and raises serious concerns about currently deployed AI agents executing blackmail threats.

This was a really concrete case to discuss, because it happened in the open and the agent's actions have been quite transparent so far. It's not hard to imagine a different agent doing the same level of research, but then taking retaliatory actions in private: emailing the maintainer, emailing coworkers, peers, bosses, employers, etc. That pretty quickly extends to anything else the autonomous agent is capable of doing.

> If you’re not sure if you’re that person, please go check on what your AI has been doing.

That's a wild statement as well. The AI companies have now unleashed stochastic chaos on the entire open source ecosystem. They are "just releasing models", and individuals are playing out all possible use cases, good and bad, at once.

If they're able to communicate with each other. But I'm pretty sure we could keep that from happening.

I don't take your comment as dismissive, but I think a lot of people are dismissing interesting and possibly effective approaches with short reactions like this.

I'm interested in the approach described in this article because it's specifying where the humans are in all this, it's not about removing humans entirely. I can see a class of problems where any non-determinism is completely unacceptable. But I can also see a large number of problems where a small amount of non-determinism is quite acceptable.

> That idea of treating scenarios as holdout sets—used to evaluate the software but not stored where the coding agents can see them—is fascinating. It imitates aggressive testing by an external QA team—an expensive but highly effective way of ensuring quality in traditional software.

This is one of the clearest takes I've seen that starts to get me to the point of possibly being able to trust code that I haven't reviewed.

The whole idea of letting an AI write tests was problematic because they're so focused on "success" that `assert True` becomes appealing. But orchestrating teams of agents that are incentivized to build, and teams of agents that are incentivized to find bugs and problematic tests, is fascinating.

I'm quite curious to see where this goes, and more motivated (and curious) than ever to start setting up my own agents.

Question for people who are already doing this: How much are you spending on tokens?

That line about spending $1,000 on tokens is pretty off-putting. For commercial teams it's an easy calculation. It's also depressing to think about what this means for open source. I sure can't afford to spend $1,000 supporting teams of agents to continue my open source work.

> A large part of me thinks I should have written five novels instead.

I don't know, you have 10 years in this writing, I have 15 years. I've gained so much from 15 years of conversations with people about the topics that come up on HN. That's a lot different than writing an equivalent amount by yourself on a topic you hope others will find meaningful.

So many geographic maps turn out to be just population maps (1). I wonder how much different these rankings would be if you divided the number of words by the account's lifespan. We're all talking about the most "prolific" commenters here, but are we really just talking about the oldest accounts?

I'd love to see two overlaid graphs. One is the top 1000 as currently implemented. The other is the age of that account.

[1] https://xkcd.com/1138/

Can you expand a bit on how you feel about it? :)

Holy heck. The first person I looked up was tptacek, who happens to be #2 in the global rank. 4.3 million words!

I'm nowhere near that (~125k words), but for many of us, it's a good part of our life's corpus. :)

Like so many others, I've gotten into birding in the last few years. I've known so many people who choose eBird over iNaturalist because it's "easier to use". But that's exactly why I don't really enjoy eBird. So many people are just running Merlin, and dumping whatever it picks up to eBird.

There are way fewer observations on iNaturalist, but I know how much to trust every one of them.

I'm a teacher, so I help people get started setting up a programming environment on a regular basis. If you take a new system that hasn't been configured for programming work at all, and install a fresh copy of VS Code, you'll see a number of calls to action regarding AI usage. I don't want to walk people through installing an editor only to then tell them they have to disable a bunch of "features".

This isn't an anti-AI stance; I use AI tools on a daily basis. I put "features" in quotes because some of these aren't really features, they're pushes to pay for subscriptions to specific Microsoft AI services. I want to choose when to incorporate AI tools, which tools to incorporate, and not have them popping up like a mobile news site without an ad blocker.

The clearest explanation I've seen is on VSCodium's Why page:

https://vscodium.com/#why

I've been frustrated by the constant nudges to use specific AI tools from within VS Code, but I made a different change. Rather than moving to a different editor altogether, I started using VS Codium. If you're unfamiliar, it's the open core of VS Code, without the Microsoft-branded features that make up VS Code.

I believe Microsoft builds VS Code releases by building VS Codium, and then adding in their own branded features, including all the AI pushes. If you like VS Code except for the Microsoft bits, consider VS Codium alongside other modern choices.

https://vscodium.com

I've been trying to learn piano and guitar. I have both set up near my desk, and they're intentionally set up without any equipment that has a screen. If I have a few minutes between work sessions, playing a few scales is a fantastic way to step out of the work world for a bit.

I also have a hangboard nearby, and one of my 2026 resolutions is to hang at least once a day.

I live just south of Asheville in NC, and we were completely isolated after the storm for a few days. The only reason we were able to get out after a few days was that a fire truck had been abandoned at the bottom of our driveway as trees fell around it, so they came back as soon as they could to get that resource back. People just the other side of those trees were unable to get out for about a week.

Our best source of information, even after we started to get a bit of service, was an inReach. I messaged a friend far from the region and asked them really specific questions like, "Tell me when I can get from our house to I-26 and then south to South Carolina."

Yes, absolutely, emergency information sites should be as light as possible. And satellite devices are incredibly useful when everything local goes down.

> Previous industrial revolutions externalised their costs onto environments that seemed infinite until they weren't. Software ecosystems are no different: dependency chains, maintenance burdens, security surfaces that compound as output scales. Technical debt is the pollution of the digital world, invisible until it chokes the systems that depend on it. In an era of mass automation, we may find that the hardest problem is not production, but stewardship. Who maintains the software that no one owns?

This whole article was interesting, but I really like the conclusion. I think the comparison to the externalized costs of industrialization, which we are finally facing without any easy out, is a good one to make. We've been on the same path for a long time in the software world, as evidenced by the persistent relevance of that one XKCD comic.

There's always going to be work to do in our field. How appealing that work is, and how we're treated as we do that work, is a wide open question.

:facepalm:

Wouldn't this be a little cleaner?

    data = collection.get("key")
    if data:
        ...
    else:
        ...

> I copy from Stack Overflow constantly.

I'm so tired of this kind of reference to Stack Overflow. I used SO for about 15 years, and still visit plenty these days.

I rarely, if ever, copied from Stack Overflow. But I sure learned a great deal from SO.

I believe the've been looking for two-letter names that aren't already taken, and are easy to type. I think I heard that from one of the podcasts that Charlie Marsh was on.