My open source projects were barely sustainable when the few PRs that came in would fix actual, non-hypothetical, serious logic problems.
Now I've been seeing PRs like that, fixing annoyances, performance edge cases, or hypotheticals, across my open source repositories, and I don't have time for them. I especially don't have time to fix problems that probably nobody will ever experience, and yet they keep coming. I have not organized my life around triaging annoyances (let alone fixing them).
If the trend continues, I'll soon be forced to close repositories to contributions, or just stop maintaining the projects entirely. And I imagine that's the point a lot of maintainers are at.
Have you considered that the person submitting the PR to fix said problem has themselves experienced it? I find that is most often the case. But there is definitely an increase in the "performance edge cases, or hypotheticals" framing, where I am equally befuddled/too tired.
I guess also AI managed canary l releases could be an option IMHO that combine open PRs.
I came from the "free software as in free puppy" camp. I would stop by to play with the puppy, but I won't feed it.
I guess if I do the same today, I might submit a better-than-band-aid fix from LLM. You would probably think I am not a real user.
I had multiple PRs for entirely hallucinated problems.
Or where no human ever looked at project or PR and were just attacking open source projects with spam PRs pretending to be contributions. Not sure whether it was out of malice or done for some other reason. Maybe they were mistakenly thinking it is helpful?
I believe a fair bit of this is an attempt to gain numbers where quantifiable numbers are advertised on otherwise indistinguishable resumes (with the corresponding belief that those numbers make differences in filtering or hiring decisions).
I base that on having seen people in the days of old put their Stack Overflow reputation on resumes thinking that having a rep of 4321 will be more likely hired than someone with a rep of 1234.
Having "210 accepted PRs on GitHub" (they believe) would appear to make them a more productive / less risky hire than someone who hasn't done any.
---
This is really a case of indistinguishable resumes where nearly everyone in the pool went to the same college, same internships, same capstone project, and maybe the same first year or two of professional experience.
I could see it as a "proof of familiarity with CI and expectations for a PR", though for applicants that are spamming the process with poor PRs (and people posting poor Q&A on Stack Overflow back in the day), this often hurts them more than it helps as its a demonstration of what they are going to be doing if they are hired (I've already got too many poor quality PRs to review, your PRs would not reduce the amount of time that the reviewer would need to spend).
So yeah, I've considered it and eliminated it for most PRs we used to get. Because no, a 2nd year student in university did not run into an obscure edge case in production that none of my 1M+ QPS users did.
For me they're real issues. Half of which i'd found before but never had the time to submit something for, the other half its things the AI will have encountered - and i check to confirm its a real issue.
The AI will follow guidance you leave in a repository. One of them said "no ai", so it reported that to me and I just forked and went on with my life.
There might be people farming for gh contribution i dontk now — but I can assure you the trend will continue, and afaict it is improving software; at least most of my contributions have been merged.
I see a lot of unsolicited PRs (and my case not from actual users but instead university students trying to pad their resume with random open source contributions but while clearly using AI exclusively)
AI can help with that! Give an agent some rules, like via a skill for the project, and then point it at the PR/Issue for triage to determine if it's one that fits the rules for auto-closing (already reported and decided WONTFIX, for example). Have it run this across all your repos.
Have the AI tooling do the heavy lifting of reproduction and benchmarking. One of my projects is fairly complicated and one user ran into some problems which he could describe the situation that caused it, but was light on any other details: my apt package cache was slow to serve the PHP update to his users. I asked AI to try to reproduce it: Set up a repo, set up different versions of my cache, have a client pull it, put an updated package in there, and have a client pull it again under different configurations. This could have taken me hours to do myself. AI did it with ~2 minutes of my effort.
>I have not organized my life around triaging annoyances
Then don't. AI can help with the triaging. Or, I give you permission: Close the repo for contributions. If you don't have time to maintain it, let people know, so they can fork or replace it. It sounds like your repos are already closed to contributions, due to lack of time/attention. That's fine, don't feel bad, just go ahead and make it official.
Not all contributions are equally valuable, and that is notwithstanding AI use. AI facilitates volume but not necessarily quality. Why should anyone spend resources benchmarking and reproducing a non-issue, or a PR that will not otherwise pass muster? We can automate a lot, but human judgment is still critical in making things to be used by humans.
Sure; if the primary repo goes unmaintained, and nobody else steps up to maintain a fork, yes you might get multiple forks that solve a specific users individual issue. I'm not sure there's a way around that.
The comment I was replying to sounded like someone who was unavailable to handle non-AI requests as well. AI contributions is making it worse, though my biggest projects haven't been hit by it yet <knock on wood>.
Low quality contributions is a very real problem, I agree with that.
You are too nice. Anyone who isn’t going to spend time to explain an issue should be automatically rejected.
We’re in this ugly situation where AI causes the problem and people’s idea to fix it is throw in more AI.
Just like throwing technology at climate change when our own obvious habits are destroying the planet. Let’s move to Mars.
I like the idea of leaning on a decision like Jev model to identify such things.
For me it was a bit useful but it has not offset even 10% of damage caused by LLMs
and is solving least important problem
This requires a huge setup of the necessary AI infrastructure.
Free software and open source licensing, and projects like Debian, are set up so the software survives even the death of the author and the project. They can make their own improvements for their own ecosystem without upstream support. They integrate what comes from upstream whenever they can.
Someone get their contributions accepted by upstream is a nice thing to have, but upstream don't have to accept it, and if they do, it's at their pace.
Secondly, don't let a pile of open issues and PRs bother you, unless they're fundamentally incorrect (in which case let your authorial voice be known). Because it's also useful for users of your software to give each other feedback and help each other out, e.g. "this app doesn't work with the latest macOS changes, but here's the code changes you'd need to fix it" means that users can help each other often for years before the maintainer gets around to looking at issues. If you're on some sort of code forge site, users can organise there. If you're not (and only accept patches by email or whatever), users can organise themselves elsewhere.
And be OK with closing issues that have dubious merit?
> Let me be the agent of a trove of people telling you what to do and how to do it
Do you see the problem here?
This is what I do, and have done, since long before LLMs. And if they aren't interested? If its a big enough problem for me, I maintain my own fork. Forking is the greatest part about open source projects!
I rolled my eyes so hard I died.
That being said, I know if one reaches that point in burnout, money won’t change anything. You are under no obligation to do any of this after all. So in the meantime, thank you. :)
We need the equivalent of that for open source software maintained by people from before 2004.
This would eliminate a lot of AI slop, and I wish I had a popular repo so I could try it. I might unfilter and glance at the non-funded PR's once a year.
As a maintainer of some projects: at this point I would appreciate far more useful review of already submitted PRs.
As long it is not blind copy-pasting LLM output. That would be highly obnoxious attack.
Here's my most-recent real-world problem: I wanted to plug in my midi controller keyboard and just play a piano, so I can practice my choral music. Very basic.
This turned into many hours over several days trying multiple DAWs, trying to get tehm to install, get them to run at all, get them to recognize my controller, get them to play a VST plugin...changing system settings, completely jacking up my sound, restoring it all again.
It was a nightmare.
So instead I fired up Claude, worked with it for about 3 hours, and now I have an *8-channel* VST player that supports all my hardware, and includes a snappy UI, with outstanding performance and sound. It's amazing!
I literally spent less time creating a custom solution than I did trying to get the off-the-shelf stuff to work. I could make my player OSS too, but the effort to do that is huge, for all the reasons this forum would understand.
Bespoke software avoids all of that and allows me to do what I want, with the environment I have. I don't have all the features of an FL Studio, but I have what I need, and that works for me, beautifully, at low cost and effort. What's not to like?
So maybe we should share OSS "templates" instead of full products, projects that are the barebones of a solution that you can clone, and have your LLM modify to suit your needs, on your platform of choice.
For example, perhaps my barebones VST player is mostly good enough for you, but you also want to play sound fonts? Cool, clone my project, instruct your LLM to add sound font support, and add support for your controller or your output device. Or convert it from Gnome to KDE.
The time and cost of doing this is so little now, why not? I'm getting more out of my computer than ever, and I'm happy about that :)
What we need is more than just OSS templates. We need fully LLM-managed "software libraries". These would be like package managers on steroids. They would be a huge hierarchy of LLM-coded modules. When you ask for an app, your LLM assembles it from the modules that already exist. If you ask for a novel feature, it codes it up and places it into the library.
So, for your VST player example. You would ask for a barebones version with just the features you need, and your LLM would assemble it for you from preexisting library modules. Then you ask for sound font support. If there is an existing sound font module, it would download it and add it into your app. If there wasn't, it would create one and put it into the library, so the next time someone asks, they can quickly get a VST player with sound font support.
I’m an engineer but not a software engineer and it’s been pretty interesting trying out vibe coding and seeing the barriers of entry being blown apart.
I’ve been progressing at learning the piano quite a lot faster than I expected with this. I’ve been debating open sourcing it but I’m not sure I want to take on the work of managing it
However, it seems to me like there must be some middle ground, between sharing architecture "ideas" and full on code PRs.
I myself maintain various FOSS projects; and incoming PRs are marginally useful. generally i just close, and copy with claude following my own & project taste (crediting initial contributor).
I've seen this idea: https://v-it.org/
Hate to say it but this is genuinely just a skill issue.
Resorting to building a DAW tailored to your own lack of understanding because you can't be bothered to learn tools that are extremely widely used and well-documented is the most extreme case of yak-shaving I've ever seen.
These has been this ladder of copium that "oh human still has input here for X, oh shoot it doesn't with Claude 4.5, well human is still needed for Y, oh shoot...".
You are suggesting that you can provide context that the model can't on its own. This is a search problem, every iteration it gets better.
We are at the stage in our company where "/goal optimize this" gets better result than any biased human context engineering can provide.
How exactly do you propose that the model should search my thoughts? Of course I can provide context that the model doesn't have. I can provide context that does not exist anywhere on the Internet (unless I got hacked).
> We are at the stage in our company where "/goal optimize this" gets better result than any biased human context engineering can provide.
First, a "this" must exist to optimize. Second, you have to be happy enough with it that optimization is the only problem.
I fully agree with them that the easiest solution is to ask an agent for exactly what you want, and iterate from there. Evaluating and installing and learning off-the-shelf software is often more of a pain than doing it from scratch.
And the same library is built without the "seed" or baseline?
1) If the UX/UI is important for you, the agent can have access to your other work, or to your other conversation and build your tool with that in mind. This is just another search space question. If you have a documented, predictable taste, the models can efficiently produce what you need.
My uni supervisor/professor used to complain about my slop papers, then I started generating them in his writing style by just pulling down all of his papers and blogs, and he doesn't mind it now.
2) In general, the LLMs are great for steering opinion. There is good research on steering political opinion with LLMs. Your taste and opinions can be formed by the LLMs to like the solution that it provides you.
Beware: Your best path forward there may be, contrary to part of the thesis in this post, be to vibe code a replacement. There are a couple reasons why:
Expect to encounter significant push-back due to AI contribution. For example, I've had bug fixes to significant problems be rejected because "we are in the future going to be adopting a no-AI policy". Annoying because I spent my time working on it because they did not have a no-AI policy. Amusing to have another contributor contribute the exact same fix, like character for character, under a new PR (referencing my issue, so not done in isolation) and get it accepted.
There may also need to be sweeping changes to the project you are fixing -- which are going to be a burden to the maintainer and likely wouldn't be accepted because of that. A fork may be the way to go here, but also "vibe coding" a replacement might be the best option.
Earlier this year I build a replacement apt cache called apt-cacher-ultra. Part of it was because the existing apt-cachers had reliability issues, but a bigger part was I wanted something that would survive ubuntu.com being under a DDoS for a week or two. apt-cacher-ng was totally useless at that. The path I wanted to try was to have verified known good snapshots of the upstream, background refreshes, pre-fetching of known "hot" packages when the upstream refreshed. HUGE changes.
I've put 100 hours or more into this "vibe coded" alternative (mostly enhancements and fixes for other users use cases). I'm pretty confident that this wouldn't have been something the apt-cacher-ng or apt-cacher maintainers would have been interested in accepting.
This was just a sci-fi dream in 2022. Now I wonder whether apps are collapsing into an inference wrapped singularity.
https://dri.es/license-only-versus-stewarded-open-source
The truth is: open-source doesn't come with any support, maintenance or feature request fulfilment by default.
By definition, that's just a code with no strings attached.
Do not expect maintenance or bug/security fixes by default from any project, unless stated otherwise, or there's at least some metrics which hint of maintenance.
We need SOCIAL.md to describe expectations from this project.
If that mindset is new to you, read this: https://raccoon.onyxbits.de/blog/bugreport-free-support/
People post a bunch of throwaway microprojects. So what.
There are wide fields, industries, branches of society, communities, groups of friends left without proper apps to do "that thing". Not to do it better, but to do it at all. It is incredible that a much wider amount of people can engage in creating those apps now. That should be encouraged, and not gatekeeped into existing projects that are already drowning in PRs.
My apps are native and fast. They run on all my devices, connect to each other through APIs and MCPs, and look the way I want them to.
Gone are the days of commercial organisations using OSS as a marketing tactic to gain a base they later monetise. It worked, it was a tactic and no one really liked it.
Your options will be:
a) community led OSS - like Apache projects, CNCF projects, etc
b) some dudes weekend OSS project
c) commercial vendor, likely proprietary or at best ‘source available’
The ideal future of software is not as a product or a service but merely as a tool, used by people with a problem to solve that has nothing to do with selling the tool.
This is the first time “open source” has meant anything other than “probably free” to the average person.
I constantly hear people complain about UIs. You know who probably makes a great UI for non-technical retirees? A non-technical retiree. They’ll probably start from forking an OSS project, though.
I think the really clever folks will move up to building vim/emacs/Firefox esque apps that are heavily scriptable, and your weekend OSS projects will be done by non-technical folks who just want something really specific.
You know what makes a great UI for non-technical people? Codex or Claude.
GUIs were a response to the limited resources available on old computers. And, as we're increasingly finding out, text is FINE. People like Teh Pretty(tm), but they'll take plain text if it solves their problem.
But, congratulations, you DID identify the primary bottleneck--customer support. And, no, I am NOT going to be unpaid customer support for non-technical people. Bleah! Get back to me when Claude/Codex don't suck for that.
Once upon a time, people payed hundreds or thousands of dollars to acquire software development tools. Even those who programmed as a hobby would have to spend a fair bit of money if they wanted to grow beyond interpreted BASIC.
I'm not claiming the current situation is ideal, but open source has given us much more freedom to program.
Sending no-strings-attached donations to the person instead of targeted fundraising of the project is just a disservice if you're expecting maintenance, more harm than good.
PR coming from an individual? Make the price low but not zero and experiment with the price until the PR rate is manageable.
PR coming from a company? Charge your regular rate.
Money as proof of this-guy-actually-cares isn't an awful idea but you gotta be careful to not become a parasite.
An open source project management software nips at JIRA's profit margins. OSM loosens google maps' stranglehold. The swath of FOSS crms delegitimizes Salesforce's per-seat cost. Most recent example is open weight LLMs driving a spike of fear deep into the heart of the current lynchpin of the American economy.
MIT license was a mistake. We tried to play by the rules of capitalists and the obvious happened: the capitalists exploited us. "Thanks for the free labor! By the way, we don't hire people that don't have FOSS commits on their resume." Google and Facebook, last I checked, forbid the inclusion of GPL3 code in their projects. Great! Let's make more GPL3 copies of their software so nobody has to give them money anymore.
I think of OSS as a way to get value into the hands of users and for them to not have to pay someone else for it. I honestly wouldn't be too bothered if they had to pay me, but I relinquish that payment because not paying me means more users and more use, and I have the privilege of a decently paying tech job, so for my personal cost/benefit analysis the free software wins. (And yes, I default to GPL not MIT/BSD.) And it's not just the direct users of my software that I care about, it's also the overall ecosystem of freely available and modifiable software that I want to contribute to.
It also doesn't bother me when other people use OSS as an anti-competition mechanism. That's a different reason to do something that happens to resemble what I'm doing. It doesn't contradict or interfere with what I'm doing nor my reasons for doing it.
That's a very common misconception I see among the developers. They get upset when someone sells their software, especially when the author distribute it for free.
Please read this: https://www.gnu.org/philosophy/selling.html
If the license doesn't allow to sell free software, the license is not free.
If you don't want others to sell your software, you need to use non-free license. There's absolutely no shame for that.
No, that isn't the argument here at all. The argument is that the existence of OSS alternatives (for which there is no real reason to pay more than a nominal distribution fee, aside from contracted support) makes it harder for proprietary software to retain commercial value.
FOSS doesn't mean you have to distribute the software for free. If you want to earn money from FOSS, just don't publish it and sell it instead. It doesn't make it less open.
I sell open-source software of myself and others, for example, quite successfully.
Linked article states:
>So if you are redistributing copies of free software, you might as well charge a substantial fee and make some money. Distributing free software is an opportunity to raise funds for development. Don't waste it!
It has been this way all these years, that's why I call it a common misconception. The GNU article by the link above is from 2012 or earlier.
People would gladly pay you more because it means they have to pay less.
The only necessary requirement is that the payment goes down with the number of payers and that there is no free rider problem.
OSS is a great passion project for a developer that wants to take it on, and when the community comes together, they can do incredible things. From a social perspective, I'd struggle to find anything that's more empowering to the average person than the ability to spin up a Linux server and get going in like 30 minutes.
But support is lacking. Proper documentation can be lacking. The sheer amount of work required to host, fix, update, and keep uptime up on real software at a real business is beyond a hobby, and enterprise support is mandatory, not optional. There are legitimate legal requirements that need to be met.
The fact of the matter is that the nature of the capitalist system that exists in the western nations is the same system that allows for OSS to exist at all. Tailscale was hard as hell to develop. Docker was hard to develop. I love the opensource alternatives to these services. I wouldn't use them at my company.
Even if you limit it to FOSS projects extensively used by massive corporations i is far from rosy.
This is the primary perspective behind this post, and that's going to look much much different than the devops roles a lot of us have.
> Too many apps need a database, a cache, a background worker, a reverse proxy and a long config file before you see a login screen
These apps need a database/cache/proxy/config because creating a fault-tolerant, performant, flexible system needs these things. The author wants everything to be a single binary running on a $5-a-month server. A lot of these open source tools could be updated to run in this mode, but that's not the configuration that's sustainable when there's a disaster.
You need to look at why these nonprofits are using open source projects in the first place. They probably want something free, and they probably want something with clear documentation. Maybe it's something that they're already used to. I would argue that AI-assisted (or vibe-coded) apps are cheap enough to fill this niche. Documentation is less important if the features of the app are much more narrowly-scoped. If they're already used to another app, then the vibe-coded one could be designed to provide a similar experience.
I've worked with non-profits that have struggled under the weight of the software they're using. Most often the problem is the software being too general in a way that doesn't model the domain well. Cheap custom software would help a great deal.
Also, its hard for me to trust their AIs and feed again and again my information to many different apps and their AIs.
If you think about it, you need one claude or codex subscription and you can generate your own apps with your own AI models that I trust more than ones used in the apps.
Recently I started working on open source local app playstore. Would appreciate your feedback and any input - (https://github.com/Ersho/sayn)
You may not need it, may not like the colors, the UI, the depth, features, etc. That's the beauty of AI it can build whatever you want however you want it, all for less than a dollar in less than an hour
Yes, we need more apps, tailored to every user. From now on I refuse to install any app not customized to my needs
And yes, we should share more, so AI learns from all of us and builds stuff for us with all the shared knowledge of all
> And yes, we should share more, so AI learns from all of us and builds stuff for us with all the shared knowledge of all
But the output of this is that every running forum is deluged with hundreds of vibe-coded apps that are marginally different from each other - a moderation burden.
If the test schema passes and if your original articulation of intent was correct, then what you have is already very good. From that point onwards, it's the question of accelerating development by just reviewing the output of the LLM and getting it into the right shape.
The combination of the review of the output of the LLM and the review of the test schema and its output, that is going to be a lot quicker to do than coding everything from scratch with a proper design paradigm.
And precisely that is the issue: people vibe coding something and releasing it to the world don't have this domain specific knowledge to build upon. It used to be that building the software would force the programmer to learn about the domain, but that link is broken now too.
… so, as long as it isn't actually "vibe" coding?
So there's vibe coding for the general populace, and there's vibe coding which domain experts can use to articulate problem statements in well-formed ways and articulate test schemes to ensure that their articulation of the problem is actually solved in the best way possible.
Hopefully that makes sense.
Industry happens because of domain experts for the large part.
If domain experts have a way to accelerate their wherewithal significantly, then industry progresses significantly.
The promise of LLMs and the promise of vibe coding is fundamentally that domain experts of industry can significantly accelerate progress.
The rest, who are not domain experts and who are not of the industry, are also progressing, but the value of the progress they have in terms of how they contribute to how industry contributes to society is fundamentally questionable.
why is it bad that customers hear the truth?
I believe that train has sailed on this association.
Yes fix bugs and add features etc. for your own personal use, but don't get your hopes up about contributing back, it's usually a waste of time.
1. Friendly editor: Vix @ https://github.com/vixide
2. Broader web UI: Lily @ http://github.com/lilydesignsystem
3. Smarter web service helpers: Axa @ https://github.com/publicprojects-io/axa
I tried many open source software projects before starting these, and I'm a longtime open source developer so I highly value all the work from people working on vi and emacs, Daisy UI and shadcn, Rocket and Loco, and more. These projects are all aiming for public nonprofit social good.
The problem with Open source presently is that it provides no way to sustain the energy used to support it. I have a limited amount of time in my life; I currently spend roughly 1/3rd of it supporting my family. If I use another chunk of time to ‘support’ open source, I’m looking at my work being taken and used for the profit of others without being able to profit from it myself. If I can’t profit from it, I can’t use it as leverage to make it so I don’t need to spend 1/3rd+ of my life working to support my family.
I am currently writing software with the expectation to sell it. I may even give away copies of it, I haven’t decided yet, but the source code won’t be online, and the support documentation will be behind a registration wall, as will the API. Nothing will be visible on the public internet, not for any other reason than AI scrapers and the large companies that use them.
There may be other ways for small software businesses to be sustainable, but at this moment it’s hard to see them.
Future Software Factories will have extra large static harness, that will be have a very few dependencies, e.g, reactjs, bundled into the package. Even nodejs probably will be more of a runtime feature, patched by the cloud.
open source was mostly a feature of not enough software eng resources. Now , since this is in abundance, instead of a shortcut it becomes a risk.
Yes, something like that.
I've pitched the idea on twitter and asked security guys whether someone does something like this:
Say we have 0/1-days in Python 2, which is not maintained. I'm thinking of software + service which will:
- Detect the exploit attempt
- RE and dissect it
- Document it, propose fix
- Publish it somewhere (!)
- Others discover it (!)
- Individual agents apply the fix locally
This way we don't need centralized maintenance, although it will work better with it....But also in this future, the source code of all of these programs is stored in context in the cloud of a few tech companies with unknown levels of security around such context?
Considering they likely use the stuff for training, what are the odds its actually air tight for the foreseeable future? The odds are 100% that the contexts will be compromised because that's the conjecture with the open source projects as well, yes?
Doesn't that seem concerning?
But I don't think there's a big overlap between the people who could even name one open-source library and the people flooding the market with vibe-coded to-do list apps.
AI-assisted means that someone who can actually write code does the piloting of the AI.
Vibecoding however, that's usually people with sometimes zero understanding of tech, often enough not even capable of executing basic CLI commands, "creating" projects, without having anywhere near the skill required to see if they're getting shafted by their AI tool of choice.
It's a cycle which breaks without either component missing.
The fact that free software is benefitting at all from more coding power is a win. We have bottlenecks to solve though.
Lots of people make things now and that's cool and fun. But what people want to see is a track record of support. They want to know that the things they come to use and depend on won't exit the scene just as quickly as they came.
This is just like 'killed by Google' on steroids. Yes you can write way more things things than before, but if you can't sign up maintain them, then they aren't providing value.
Focus is becoming the commodity. Doing one thing really well is what becomes hard when you could be doing 1,000 things (poorly). To be successful, you can't just throw shit at the wall and see what sticks anymore, because nothing is sticking that way. The walls are already covered in shit. You have to have good product taste and a willingness to apply yourself to it for a long time.
https://rainshare.dev/ enables using up your token subscription to help others with their backlog, for example. Code forge goes brrr.
You cannot eat tokens.
I’d much rather give my money to a small sustainable business or a coop versus a traditional enterprise producing and/or hosting software, whenever possible, especially if they are in some way producing high quality open source software. “Shop local” sort of thing.
Because I am a masochist I also decided to make this a way to revive the 2 (!) Stadia controller I have. The initial results were less than promising in the same way hitting your head on bricks is. But with the wireless debugging port wide open on my tv now, I have a small app I am cleaning up that could help you work through the absolute vagaries of connecting Bluetooth game controllers to Google tv (once you go into promiscuous mode, something like three separate services on the tv start fighting for you, last man wins and usually the last man is a service that doesn’t even consider game controllers as a case). I’ve no idea how well it would work or where even to offer it, but in spite of me building new stuff, I mainly do it along these lines too. See also, lintarr a tool I am polishing up which tries to do something like formal verification for all the rules you can set up in different services across the arr stack so you can’t have rules that clash.
Because this is the joke. AI agents didn't change who "gets" to write software at all, it's still the same middle-level and senior devs it's always been. The term is intentionally derogatory towards the people who think otherwise.
This reads like Hacktoberfest all over again.
What one person might find "annoying", another maintainer might've purposefully put there. Not all friction is bad, and telling people to go out, remove all friction and then report back to receive a pat on the head is probably going to end poorly.
This reads more like another attack on the commons than someone genuinely invested in FOSS. Which.. uh.. checks out when you check out the author.
__
Please do not take advice on how to interact with FOSS from anyone but the actual people building it.
If you fundamentally disagree, that is fine. You can become the person that builds their own thing and makes their own rules within that space. But only there.
A very common project that most guys who find this "vibe coding" do is make a calorie tracker, so they ask Claude to build them a calorie tracker, then they add some data from an Apple Watch and Garmin. Now they believe they've built a "dieting app."
It's just funny how many of them actually take data from Garmin and an Apple Watch and add a bunch of activities to work out a "calorie burn of the day" and how inaccurate it is.
Apple and Garmin are not accurate for this, read this: https://macrocodex.app/knowledge/rethink/how-many-calories-d...
But very few of them actually bother to check whether their "calorie burn estimate" even works to achieve the rate of loss or gain they set.
Not everyone is "doing better with AI", many of them are doing worse, they just don't know!
When someone vibes out a new little todo-like app, they are usually thinking of exactly _one_ user: themselves. They get what they want with very low executive effort. That is what's so cool about it!
Good Open Source contrib takes effort, they don't want randos slopping out PRs. Total apples-to-oranges comparison.
> Pick one open source tool you use. Find the most annoying thing about
> installing or running it. Spend a weekend fixing it, with the people who
> maintain it. Then tell me about it.
Flex: The most annoying thing is that it's written in C (and is therefore a pain to cross-compile), and it has a runtime dependency on GNU M4, which depends on gnulib (the GNU anti-portability library). Flex is currently being rewritten to be more complicated and worse by ESR for reasons beyond my comprehension.I had an LLM write an m4 implementation in Rust based on the POSIX specs, then fuzz-test it against GNU M4 until it was mostly the same (modulo all the regex bugs it found in GNU M4). Then I had an LLM rewrite Flex in Rust, using rust-m4 as an in-process library.
QEMU: It's written in C, has a bunch of huge dependencies, automated test coverage that rounds to zero, their main contribution path is a mailing list that sometimes swallows patches, and the maintainers might ask for a risky patch to be made more risky before they'll accept it. I went through that process once for a few relatively minor fixes and never want to do it again.
I had an LLM write a multi-architecture emulator with a JIT, high test coverage, and a device plugin system that uses WebAssembly.
Busybox: Written in C, huge pain to build. I tried switching to Toybox instead but it's incomplete, the build system involves a pile of shell scripts that require Bash, not any other POSIX-compliant shell (including its own), and the maintainer's response was basically to tell me to fuck off: https://lists.landley.net/pipermail/toybox/2024-October/0095...
I had an LLM write a replacement in Rust that has enough shell builtins to run an autoconf script from within a hermetic build environment, and its only build-time dependency is rustc.
---
You know why LLMs are so great? Because they save me from having to interact with open-source maintainers who intentionally choose to make their software baroque and their contribution process obnoxious.
You want your project to be written only in ANSI C89 and built with Meson and accept patches only via GPG-signed email to a mailing list that silently drops HTML? Fine, have fun, an ASIC somewhere in Iowa is going to spend the next 3 hours letting me forget your project exists.
People have stopped having kids as well -- because it's fundamentally a very human interaction where you have to deal with a lot of bullshit, not just with the kids, but with all the teachers, schools, daycare providers and coaches that you'll end up dealing with. People just opted out. Instagram's more fun.
Software brought people together because it's a team sport and it's hard to ship anything significant on your own, especially in a short time frame. But AI has changed all that. Now we can bring all the isolation that we've brought to the rest of society to software development as well.
AI just gives us another way that we don't have to actually work together anymore. And for you, that seems like a big benefit, because it's clear that you didn't enjoy working with the people that ran those projects.
> Your post reminds me about the ongoing trend where people don't spend
> time together anymore. They prefer to sit in front of their computers
> and interact online.
This is, I think, partially a matter of social circles. The kind of people who enjoy the idea of socialization in an abstract sense but don't actually enjoy the process of socializing will spend all their time indoors on their computers. Meanwhile in the real world people are having a blast. Very few people want to spend time typing on web forums when they could be outside shopping, or dancing, or drinking. > Software brought people together because it's a team sport and it's hard
> to ship anything significant on your own, especially in a short time frame.
> But AI has changed all that. Now we can bring all the isolation that we've
> brought to the rest of society to software development as well.
While trying not to repeat myself unnecessarily, people whose social circles were formed by contributions to online open-source communities have long been viewed by the rest of society as socially impoverished (with some reason). A web forum or chat room isn't real social interaction, not in the same way that going out to a pub or concert or book club is.LLMs are finally offering us computer nerds a way out. We can have the computer write the code and debug the segfaults and babysit the benchmarks while we go outside and have fun. We are free from the keyboard now.
Those communities that are so impoverished were the things that created reusable software that communities could organize around and customize and improve together. One-off replacements are essentially worthless in comparison.
You put up with the people to build things of greater worth.
I don't recognize the name "Flex" but I find this claim hard to believe in and of itself, having seen ESR's rants about the complexity of Automake.
> You want your project to be written only in ANSI C89 and built with Meson and accept patches only via GPG-signed email to a mailing list that silently drops HTML? Fine, have fun, an ASIC somewhere in Iowa is going to spend the next 3 hours letting me forget your project exists.
This to me sounds like the obnoxious attitude: not taking a moment to consider why others would care about their language choice, signed email or avoiding distractions like HTML email (support for which could be their "more complicated and worse").
It also reads like a refusal of true disengagement. Maybe you can now work on that, given that your profile says you're retired?
__
That said, I use AI the same way. It saves one from a ton of draining interactions, ego management etc.
I think there's enough history of my contributions to various open-source projects over the past ~20 years to serve as evidence that I go pretty far out of my way to be accommodating to projects that are open to contributions.
> It also reads like a refusal of true disengagement.
I'm not sure what "refusal of true disengagement" means.
My goal is generally to accomplish something like building a piano recital scoring app, or being able to run an old game on my laptop. There's a few dozen steps needed to achieve that, which incrementally go down the stack until it encounters a library that works well enough that I don't need to debug it. Sometimes this is a very short list -- I've been amazed at how well the Vulkan ecosystem works, ditto the Go stdlib and the major Go ecosystem libraries.
But sometimes it's like falling down a ladder, hitting each rung on the way down, in which case I turn to an LLM to make the ladder go away. A lot of C projects seem to be like this -- it's a conservative choice of language in general, for people who saw the rise of C++/Java/C#/Go/Rust/etc and decided to skip them, so the attitudes and tooling are likewise conservative.
> Maybe you can now work on that, given that your profile says you're retired?
Being retired means that I want to spend more time doing fun things and less time arguing about file descriptor lifetimes on mailing lists.
> I'm not sure what "refusal of true disengagement" means.
It means not being mad anymore at the people and "showing them" by using LLMs to the degree where one feels the need to rant on about their perceived bad decisions on HN.
I _suspect_ you might not be able to fully turn that off, as it sounds like spite might be a driving force. And that's good, but also, you've kinda made it? So if anything, you could probably be happy that you're retired and that you do not have to have these discussions anymore.
And maybe also see that the people probably did their best. Which may or may not have been very good, but they did so nonetheless.
> but the initial comment is still full of vitriol
If you think this comment is full of vitriol then you should try reading the kinds of emails people send to mailing lists. > It means not being mad anymore at the people and "showing them" by
> using LLMs to the degree where one feels the need to rant on about
> their perceived bad decisions on HN.
I think you're reading things into my comment that aren't there.HN is full of threads lamenting that projects are seeing a decline in community interaction and contributions because LLMs let people vibe-code whatever software they want. I hesitate to use the word "universal", but there's definitely a strong feeling that greenfield LLM-generated projects are in some sense stealing effort or attention that could be put toward an existing project.
I'm sympathetic to that claim on the surface, but the actual experience is that a lot of projects have only had that sort of interaction in the first place because the maintainers forced it as a pre-requisite to getting a patch accepted.
There's some 30 kLoC library out there that decodes an obscure audio codec, it almost works but you need to add some #ifdefs or whatever so it builds on Clang. The would-be user has to mentally prepare themselves for the next few weeks (or months!) of trying to cajole the maintainer into accepting their patch. And maybe it's never accepted, and now you're stuck, and that project goes into the shelf forever.
The article acts as if people with LLMs are the ones whose behavior needs to change to be more pro-social, but from my perspective its the maintainers' attitude that needs to change (if they want their project to continue receiving attention). There's no reason to put up with a maintainer's dictatorial control over a project any more, because the value of code itself is plummeting to epsilon.
> And maybe also see that the people probably did their best. Which may or
> may not have been very good, but they did so nonetheless.
Outcomes have never been judged by whether someone has done their best, but whether they have done well.Yes but FOSS is primarily about the journey, actually.
You can (and probably should, if you want to get things done) judge the outcomes and plan accordingly, but it's just people doing stuff in their free time.
That cannot be measured purely with the measuring stick of the general market.
Yes, this also means that a lot of it will be complete garbage, but that's fine and nothing to be upset about. At least beyond general disappointment with the lack of skill and sense of an average human. That's my point.
Hey, I'm retired too, we can argue about that here!
After working in the open for ~15 years, and generally being enthusiastic about collaborating and making shared common base-level tech internally at companies, I'm finding it shocking how much I don't want to even bother pushing projects to a public repo. Never mind doing any sort of releasing, contributing, "marketing", maintenance, patching, etc.
I mean, signing email and avoiding HTML sound to me like pretty trivial accommodation. And you only have to set up GPG once; it's not as if all these projects have their own mutually incompatible security/privacy/anonymity schemes.
Hopefully we can identify the useful dimensions and then come up with some terms that fit the important points in the space.
I have a project where I barely look at the code, but I've put a ton of time into iterating over the design and look&feel. I think of it as "high-effort slop coding", which in my mind is meaningfully different from "true vibe coding", but there's too much self-deprecation in the former term for it to be the right one. (It makes sense when referring to your own stuff, but not when applied to others'.)
Rhetoric question of course, especially in the AI coding age.
...and there is still a place for the robot chef, but the ideal destination for this world of post-commodified-code is probably a sandwich or Stouffer's Instant Dinner: Cheap, reliable, flexible products that don't promise to be a human.
In the meantime you are encouraged to eat the food it cooks but to take full responsibility if it poisons you or any of your guests.
Normal chefs sometimes poison you too but the most popular restaurants have thousands of happy unpoisoned diners who have all eaten the same dishes. For free.
You don't think it was easier to scratch your own itch instead of host OSS before AI? Of course it was.
Gone are the days of the hippie techno revolution, I guess. It really is "fuck community, I got mine" and its sad to see.
- Old proverb
OSS devs were sinking a mountain of time into relatively unpaid work, doing unglamorous work. Some of it is in backend systems and console utilities that few people will ever appreciate, but a lot of effort went into user-facing things. And yet, so far I found just one program that replaced one of my mainstays - I went with AI-coded SquirrelDisk to replace FileLight. Everything else is the same, and I just don't understand that. Surely not EVERYBODY rolls their own? Surely some OSS people did embrace AI rather than quitting in disgust?
A small example: tagging music files on Linux is horrible. Kid3 is.. it works, but the UX is almost like a parody. Mp3tag runs on windows and it has a macos version and it's a vast improvement. You can run it through bottles but I refuse to believe everybody does that. Surely people that use Linux listen to music, which leads itself to occasionally needing to tag or rename your audio in batches.
I had some spare time so I rolled my own mp3tag clone that would work on Linux, MacOS and Winblows. I added audio playback to it because opening in external program annoyed me. Then I added some functionality for podcasts to it: being able to search for subtitles, playback speed and such. Then I added the ability to convert between formats, using ffmpeg. Then I added a way to measure loudness, some ReplayGain functionality and the like. Then a way to find duplicates, for making sure those random files you have on that old external HDD can be deleted safely, and a way to search through what you have on all disks, even if they're disconnected, so you don't buy some album you already own like a dumbass.
It took me about three weekends of tinkering with it until it was good enough to switch to it. I keep running into small issues but by and large it's been fairly reliable. It's fairly complex, with a lot of buttons in the view, but given that the 'competition' is mp3tag (which is okay but also ugly) or kid3 (which is both ugly and almost unusable) it's just flat out superior, unless you're happy with mp3tag and don't want all the extra complexity. I don't want to offer unpaid tech support to people on GitHub but I'm considering releasing it and just leaving it there with no further communication and maybe somebody will pick it up. If not, it's still better than alternatives even if there are undoubtedly many, many bugs and the codebase is undoubtedly a tangled mess.
Aside from the ugliness of having almost all functionality without any sub-menus that would give Apple designers heart attacks and perhaps more complexity than it should have (I personalized it for what I wanted, after all), it's genuinely a decent application. It's written in c++, a language I never used. It's fairly performant on my ~15k audio files library. Changing it is as simple as pointing a modern LLM at it and asking in english for some changes. It took me approximately 20 hours to make and the cost was about $10 in subsidized tokens from Claude.
If I, a random guy who doesn't even know c++, can write a usable program in three weekends, then where is all the new vibecoded software? There are so many tools that technically work but the user experience is straight up awful, usually on linux side, or flat out holes that aren't filled, usually on mac side. As best I can tell, very little is being developed. People who have no idea about coding are recompiling console games on pc, in admittedly dubious quality, but people that dedicated years of their lives to OSS... don't?
As AI build-systems become more complex, it might be interesting to publish their work in central locations, and build new standards by which these kinds of software can interact. Or adopt existing ones in new ways!
Right now it's not always clear when and where you can apply an AI workflow and under which rules. This needs to be coordinated sooner or later.
Open Source did an end run round that problem by producing
- better quality code, faster, and replacing the arguments over standards with a level playing field marketplace
As we are learning software has eaten everything but everything has not become software literate - there are hospital MRI machines that are only certified for windows XP. Our supermarket fridges think they are cheap IOT devices for teenagers bedrooms.
It’s going to take a decade or more for society to work its way through replacing all its devices (sewages plants are devices too) wit what is a software literate approach - secure, PKI, monitored updated and … open
No fucker will trust a fridge running a blob of binary - a complete SBOM and a repeatable build will be table stakes
And while we could wait for standards what is most likely to happen is each industry wakes up and starts to say “we want a beyond zero based approach, PKI setup this way, the builds are published and run here and publish the firmware”
Walmart might be thinking about this now - but open source will if we are senile be the backbone
And it will finally be funded as the answer is for industry players to fund the fridge-python api project, mostly by hiring the devs. Or the devs go full pococo and get sponsored
Even worse, most of open source is now majority slop and it is highly likely that they will be abandoned.
On first sight. And delete the bug report.
They're not bound by any rules of "civility" and neither are you.
It's still eye-roll-inducing, of course, but it's n+1 clicks on GitHub, where n is the number of bogus issues reported (+ their delete buttons) and the +1 is the "block user" button.
LLMs themselves are built from the sum of human output when almost nobody who created that got paid for the value they created.
So what I'm hearing is that we have a distortion that can be described in terms of the worker's relationship to the means of production?
I honestly love it when people when people discover socialism from first principles.
There's nothing wrong with people working on hobby projects for fun, regardless of how useful they may be, and there's nothing wrong with others taking that work and building upon it for their own purposes. That's the entire point of open source software.
If you find yourself upset that someone "stole" your freely given away work or is making a profit off of it or you yourself are not being "compensated" for giving away "free" things you're the problem here.
If you find yourself later being upset that someone took the eggs and made an omelet and sold it and made a few bucks, then you are lying to yourself and others about your reasoning for writing open source code.
It's kind of a shit argument, why bother with it?