But the header is just "90% of Claude-linked output going to GitHub repos w <2 stars". No conclusion, just some random fact.
The problem is that this title is editorialized, and the fact is cherry-picked. Why not =0? Why not >1000? This is just a dashboard, it highlights "Interesting Observations", but stars statistics is not there.
Sounds like Claude commits are, on average, going into higher visibility repositories than humans… maybe the author would like to reconsider their approach?
There are also plenty of super high utility repos that are widely used (often indirectly), but don't have a lot of stars, or even a meagre amount.
Also there is the issue of star != star, because it's not granular.
It's similar to upvotes on general social media platforms. Everyone likes cute cats doing funny things somewhat, but only few people appreciate something that's more niche but way more impactful, useful or entertaining (or requires some effort to consume), but those who do, value it very highly. But the same person might use the same score (single upvote) for a cat video and a video that they value much higher.
It is relevant because if the vast, vast majority of repos have 2 or less stars then it's not that weird that a great deal of repos linked are, too, 2 or less stars.
Stars occasionally correlate with quality but more often it's timing and naming. I have a total of 40k stars on GitHub, and I know the code is shit in most of those repos (many written back when I was 16-18 as I was just learning to code). Jumping on hype trains before they start is how you get stars.
I guess you linking to it was a self fulfilling prophecy
If you read your own reference (not the picture, but where you took it from on Wikipedia) really really carefully, you might be able to tell why it so perfectly applies to you
The person with little knowledge overestimates they're capability, and the person which actually knows how complicated [the thing] is , usually isn't as confident they mastered it.
You’re talking about a confidence and ability gap. I have heard of the Dunning-Kruger effect. I accept all of that.
But the claim above was that having low confidence was correlated to higher skill. Ie, skill and confidence are anti correlated. The chart does not show that. The lowest data point for confidence is the point on the left of the chart. This is also the data point corresponding to people who have the least competence. Having low confidence is not evidence that you’re secretly an expert. Confidence and competence are still positively correlated according to that chart.
The Dunning-Kruger effect is not so strong that there are scores of novices convinced they are experts in a field. But in your case, I admit the data may not tell the full story.
That isn't what that shows, and the article you linked to even warns:
> In popular culture, the Dunning–Kruger effect is sometimes misunderstood as claiming that people with low intelligence are generally overconfident, instead of denoting specific overconfidence of people unskilled at particular areas.
Dunning-Kruger has also been discredited with suggestion they may have been over confident themselves:
Debunking the Dunning‑Kruger effect – the least skilled people know how much they don’t know, but everyone thinks they are better than average (2023)https://theconversation.com/debunking-the-dunning-kruger-eff... the Dunning‑Kruger effect – the least skilled people know how much they don’t know, but everyone thinks they are better than average
Self-reported studies are arguably weaker evidence, but are common in some areas for ethics reasons. In general, if errors are truly random, than they will cancel out over larger/frequent population samples.
The study conclusion inferred the skills needed to be effective at some task, are the same skills needed to correctly evaluate if you are actually proficient at the same tasks.
Or put another way, the <5% population of narcissists by their nature become evasive when their egos are perceived as threatened. Thus, often will pose a challenge in a team setting, as compulsive lying or LLM turd-polishing is orthogonal to most real world tasks.
People are not as unique as they like to believe, and spotting problems is trivial after you meet around 3000 people. Best to avoid the nonsense, and get outside to enjoy life. Have a great day =3
No idea why we all get negative karma on this thread, as I do respect a cited source opinion even if we disagree. Do have a look around for papers rather than editorialized content in the future, and note account LLM agent output is a violation of YC usage policy. Have a great day =3
Self-reported studies are arguably weaker evidence, but are common in some areas for ethics reasons. In general, if errors are truly random, than they will cancel out over larger/frequent population samples.
The study conclusion inferred the skills needed to be effective at some task, are the same skills needed to correctly evaluate if you are actually proficient at the same tasks.
If the data infers another explanation is more applicable, than I'd be interested in the primary papers/studies the editorialized opinion seems to have omitted. =3
Why would anyone settle for underpaid positions from an agency taking a 7% contract cut, and purging CVs from any external firm also contracting with their services.
Most people figure out this scam very early in life, but some cling to terrible jobs for unfathomable reasons. =3
If folks expect someone to solve problems for them, than 100% people end up unhappy. The old idea of loyalty buying a 30 year career with vertical movement died sometime in the 1990s.
Ikigai chart will help narrow down why people are unhappy:
> If folks expect someone to solve problems for them
In this type of situation, the fundamental issue is that making progress depends on many people acting in unison to increase their bargaining power, which is (a) hard to arrange even if everyone who acted this way would benefit, and (b) actually may be detrimental to some people (usually the high performers).
I agree it is nearly impossible to alter the inertia of existing firms. Most have entrenched process people that defend how things are done right up until a company enters insolvency. Fine if you sell soda or rubber tires, but a death knell for technology or media firms.
In my observations it is usually conditioned fear, personal debt-driven risk aversion, and or failure to even ask if the department above you is really necessary. These days, it is almost always easier to go to another firm if you want a promotion. =3
OP. I'm not a confident SQL user so encourage someone to double check this. From what I can tell have been 2.21x10^12 additions on GitHub in total, 6.36x10^11 of which are in repos with 2+ stars. That's about 29%. People earlier were comparing the star distribution of repos which is not really what this is about - it is about OUTPUT, as measured by additions.
Interestingly, there are 21.37b commits in GitHub, implying 104 additions per commit. Per the dashboard, Claude is linked to 20.81m commits and 50.44b additions - or 2,424 additions/commit. So additions for Claude-linked repos is higher, and it's actually higher for repos with 0-1 stars (2,568 additions/commit for Claude, 91 for all GitHub). None of this is a smoking gun but aligns with the intuition that Claude is producing enormous amounts of code. TBD whether it is 'adding value'.
shouldn't a serious heatmap (or any comparative graph for that matter) normalize the stat being displayed versus the baseline population in that bucket?
in otherwords, plot the percentage or average metric and not the absolute metric.
e.g. number of lotto winners per thousand people living in that grid, percentage of starred repos as a percentage of all repos, per capita alcohol consumption, average screen-time etc.
Edit: unless ofcourse the point of the heatmap is to show the population distribution itself. In which case the metric would be number of people per square kilometer or some such.
Still would just show where people live. If nobody lives there, you've got a null (or divide by zero) spot on the map ... so you just show where people live.
Activity isn't a good measurement for this, because AI can vibeslop thousands of lines per day of code that isn't necessarily useful for anything but increasing activity.
There is still a sampling bias if you compare blanket human written repos. I would guess people are far more likely to share their homework assignments, experiments, hackathon results, weekend toys, etc. as a public repo if they put some amount of work into it. I would guess minority of those would get any stars at all. If the whole thing was generated by AI in less then 20 minutes, I would guess they are more likely to simply throw it away when they are done with it.
Personally I think comparing github stars is always going to be a fraught metric.
One of the products my employer builds is used twice a year. People pay tens of thousands of dollars for the privilege of using it twice a year. It's tremendously valuable to be used twice a year.
Already enough comments about base rate fallacy, so instead I'll say I'm worried for the future of GitHub.
Its business is underpinned by pre-AI assumptions about usage that, based on its recent instability, I suspect is being invalidated by surges in AI-produced code and commits.
I'm worried, at some point, they'll be forced to take an unpopular stance and either restrict free usage tiers or restrict AI somehow. I'm unsure how they'll evolve.
Having managed GitHub enterprises for thousands of developers who will ping you at the first sign of instability.. I can tell you there has not been one year pre-AI where GitHub was fully "stable" for a month or maybe even a week, and except for that one time with Cocoapods that downtime has always been their own doing.
In a (possibly near) future where most new code is generated by AI bots, the code itself becomes incidental/commodotized and it's nothing more than an intermediate representation (IR) of whatever solution it was prompt-engineered to produce. The value will come from the proposals, reviews, and specifications that caused that code to be produced.
Github is still code-centric with issues and discussions being auxilliary/supporting features around the code. At some point those will become the frontline features, and the code will become secondary.
I'm definitely not an AI skeptic and I use it constantly for coding, but I don't think we are approaching this future at all without a new technological revolution.
Specifications accurate enough to describe the exact behaviors are basically equivalent to code, also in terms of length, so you basically just change language (and current LLM tech is not on course to be able to handle such big specifications)
Higher level specifications (the ones that make sense) leave some details and assumption to the implementation, so you can not safely ignore the implementation itself and you cannot recreate it easily (each LLM build could change the details and the little assumptions)
So yeah, while I agree that documentation and specifications are more and more important in the AI world, I don't see the path to the conclusions you are drawing
I think you're directionally correct, but this stuff still has to live somewhere, whether the repo is code or prompts. GitHub is actually pretty well-positioned to evolve into whatever is next.
I don't think GitHub's product is at risk, but its business model might.
I keep hearing this, and I know Azure has had some issues recently, but I rarely have an issue with Azure like I do with GitHub. I have close to 100 websites on Azure, running on .NET, mostly on Azure App Service (some on Windows 2016 VMs). These sites don't see the type of traffic or amount of features that GitHub has, but if we're talking about Azure being the issue, I'm wondering if I just don't see this because there aren't enough people dependent on these sites compared to GitHub?
Or instead, is it mistakes being made migrating to Azure, rather than Azure being the actual problem? Changing providers can be difficult, especially if you relied on any proprietary services from the old provider.
Running on Azure is not the same as migrating to Azure.
Making big changes like the tech that underpins your product while still actively developing that product means a lot of things in a complicated system changing at once which is usually a recipe for problems.
Incidentally I think that is part of the current problem with AI generated code. Its a fire hose of changes in systems that were never designed or barely holding together at their existing rate of change. AI is able to produce perfectly acceptable code at times but the churn is high and the more code the more churn.
> Its a fire hose of changes in systems that were never designed or barely holding together
Yeah... my career hasn't been that long but I've only ever worked on one system that wasn't held together by duct-tape and a lot that were way more complicated than they needed to be.
The assumption is it would be mistakes in their migration - edge cases that have to be handled differently either in the infrastructure code, config or application services.
Text is cheap to store and not a lot of people in the world write code. Compare it, for example, to email or something like iCloud.
Also I would guess there would be copy-on-write and other such optimizations at Github. It's unlikely that when you fork a repo, somewhere on a disk the entire .git is being copied (but even if it was, it's not that expensive).
My friend and I are usually pretty good at ballparking things of this nature; that is "approximately how much textual data is github storing" and i immediately put an upper bound of a petabyte, there's absolutely no way that github has a petabyte of text.
Assuming just text, deduplication,not being dumb about storage patterns, our range is 40-100TB, and that's probably too high by 10x. 100TB means that the average repo is 100KB, too.
Nearly every arcade machine and pre-2002 console is available as a software "spin" that's <20TB.
How big was "every song on spotify"? 400TB?
the eye is somewhere between a quarter and a half a petabyte.
Wikipedia is ~100GB. It may be more, now, i haven't checked. But the raw DB with everything you need to display the text contained in wikipedia is 50-100GB, and most of that is the markup - that is, not "information for us, but information for the computer"
Common Crawl, with over one billion, nine hundred and seventy thousand web pages in their archive: 345TB.
We do not believe this has anything to do with the "queries per second" or "writes per second" on the platform. Ballpark, github probably smooths out to around ten thousand queries per second, median. I'd have guessed less, but then again i worked on a photography website database one time that was handling 4000QPS all day long between two servers. 15 years ago.
P.S. just for fun i searched github for `#!/bin/bash` and it returned 15.3mm "code", assume you replace just that with 2 bytes instead of 12, you save 175MB on disk. That's compression; but how many files are duplicated? I don't mean forks with no action, but different projects? Also i don't care to discern the median bash script byte-length on github, but ballparked to 1000 chars/bytes, mean, that's 16GB on disk for just bash scripts :-)
i have ~593 .sh files that everything.exe can see, and 322 are 1KB or less, 100 are 1-2KB, 133 are 2-10KB, and the rest - 38 - are >11KB. of the 1KB ones, a random sample shows they're clustering such that the mean is ~500B.
oh, i didn't see that the 1.97 billion pages were crawled in a 11 day period earlier this month. either way, nearly 2,000,000,000 pages fit in ~third of a petabyte...
p.s. thanks for correcting me, i was using this information for something else, and now it's correct!
yeah i assume all the artifacts[0] and binaries greatly inflate that. I have no idea how git works under the hood as it is implemented at github, so i can't comment on potential reasons there.
Is there some command a git administrator can issue to see granular statistics, or is "du -sh" the best we can get?
0: i'm assuming a site-rip that only fetches the equivalent files to when you click the "zip download" button, not the releases, not the wikis, images, workers, gists, etc.
I don't think the issue at hand is a technical challenge. It's merely a sign, imo, that usage has surged due to AI. To your point, this is a solvable scaling problem.
My worry is for the business and how they structure pricing. GitHub is able to provide the free services they do because at some point they did the math on what a typical free tier does before they grow into a paid user. They even did the math on what paid users do, so they know they'll still make money when charging whatever amount.
My hunch is AI is a multiplier on usage numbers, which increases OpEx, which means it's eating into GH's assumptions on margin. They will either need to accept a smaller margin, find other ways to shrink OpEx, or restructure their SKUs. The Spotifies and YouTubes of the world hosting other media formats have it harder than them, but they are able to offset the cost of operation by running ads. Can you imagine having to watch a 20 second ad before you can push?
I think the instability is mostly due to the CEO running away at the same time as a forced Azure migration where the VP of engineering ran away. There’s only so much stability you can expect from a ship that’s missing 2 captains.
I mean the fish rots from the head, but at the end of the day that rot translates into an engineering culture that doesn't value craftsmanship and quality. Every github product I've used reeks from sloppiness and poor architecture.
That's not to say they don't have people who can build good things. They built the standard for code distribution after all. But you can't help but recognize so much of it is duct taped together to ship instead of crafted and architected with intent behind major decisions that allow the small shit to just work. If you've ever worked on a similar project that evolved that way, you know the feeling.
But also, GitHub profiles and repos were at one point a window into specific developers - like a social site for coders.
Now it's suffering from the same problem that social media sites suffer from - AI-slop and unreliable signals about developers.
Maybe that doesn't matter so much if writing code isn't as valuable anymore.
That doesn’t make sense. Commits are all text. If YouTube can easily handle 4PB of uploads a day with essentially one large data center that can handle that much daily traffic for the next 20 years, GitHub should have no problems whatsoever.
The true value prop of github isn't "hosted git + nice gui", it is the whole ecosystem of contributers, forks, and PRs. You don't get that by hosting your own forge.
Also, I wouldn't say GitHub is a corporate attempt to own git... GitHub is a huge part of why Git is as popular as it is these days, and GitHub started as a small startup.
Of course, you can absolutely say Microsoft bought GitHub in an attempt to own git, but I think you are really underselling the value of the community parts of GitHub.
The base rate argument here is the right one. I maintain a solo project with 3,800+ tests and 92% coverage — zero stars for months because I never promoted it. Stars measure marketing, not quality.
What's more interesting to me is that Claude dramatically lowers the barrier to _testing_, not just writing code. I can mass-generate edge case tests that I'd never bother writing manually. The result is higher-quality solo repos that look "abandoned" by star count.
Is anyone tracking test coverage or CI pass rates for AI-assisted repos vs traditional ones? That seems like a much more useful signal than stars.
Do people really put weight in stars? It seems completely unrelated to anything but, well, popularity. Even when I modify other peoples' code I fork to a private repo and maintain my changes separately, and I'm fairly certain I have never starred a repo.
Stars have been useless as signals for project quality for a while. They’re mostly bought, at this point. I regularly see obviously vibe-coded nonsense projects on GitHub’s Trending page with 10,000 stars. I don’t believe 10,000 people have even cloned the repo, much less gotten any personal value from it. It’s meaningless.
I'm with you on all points except for it being bought.
Programming has long succumbed to influencer dynamics and is subject to the same critiques as any other kind of pop creation. Popular restaurants, fashion, movies - these aren't carefully crafted boundary pushing masterpieces.
Pop books are hastily written and usually derivative. Pop music is the same as is pop art. Popular podcasts and YouTube channels are usually just people hopping unprepared on a hot mic and pushing record.
Nobody is reading a PhD thesis or a scholarly journal on the bus.
The markers for the popularity of pop works are fairly independent from the quality of their content. It's the same dynamics as the popular kid at school.
So pop programming follows this exact trend. I don't know why we expect humans to behave foundationally differently here.
> Nobody is reading a PhD thesis or a scholarly journal on the bus.
As someone who is involved in academia, I can attest that most of my colleagues (including myself) do in fact read quite a few papers on buses (and trams - can't forget those)
> I'm with you on all points except for it being bought.
Stars get bought all the time. I've been around startup scene and this is basically part of the playbook now for open core model. You throw your code up on GitHub, call it open source, then buy your stars early so it looks like people care. Then charge for hosted or premium features.
There's a whole market for it too. You can literally pay for stars, forks, even fake activity. Big star count makes a project look legit at a glance, especially to investors or people who don't dig too deep. It feeds itself. More people check it out, more people star it just because others already did.
I have 60-ish repos, vast majority are zero star, one or two with a star or two, one with 25-ish. It’s a signal to me of interest in and usage of that project.
Doesn’t mean stars are perfect, or can’t be gamed, or anything in a universally true generalization sense. But also not meaningless.
For example, it's used as a kind of internal bookmarking system. I don't necessarily star a repo because I think it has good code, but maybe a good idea or something related to something I'm interested in developing.
Claude's Code dev here, and I thought I would chime in on this point to clarify why I track it at all.
When I started reading commit data, it became painfully apparent that a very large number of repos are tests, demos, or tutorials.
If you have at least 1 star, that excludes most of those - unless you starred it yourself.
Having 2 stars excludes the projects that are self-starred.
Starring is also quite common with my friends and colleagues as a way to find repos again later, so there is some use to it, but I agree it's not a perfect indicator of utility or quality.
Just to clarify as OP, the point here is not that Claude is not contributing to serious work, just that the dashboard suggests a lot of usage in public GitHub repos seems to be tied to low attention, high LOC repos. This is at least something to keep in mind when considering the composition of coding agent usage, and when assessing the sustainability of current trends.
In hindsight the headline was a bit more sensational than I meant it to be!
This seems to be the same misunderstanding about agentic coding I see a lot of places.
Agentic coding is not about creating software, it's about solving the problems we used to need software to solve directly.
The only reason I put my agentic code in a repo is so that I can version control changes. I don't have any intention of sharing that code with other people because it wouldn't be useful for them. If people want to solve a similar problem to me, they're much better of making their own solution.
I'm not at all surprised that most of Claude linked output is in low star repos. The only Claude repos I even bother sharing are those that are basically used as context-stores to help other people get up to speed faster with there of CC work.
I can’t imagine that will stay the case though. I built https://github.com/andonimichael/arxitect as a first step to using agentic coding in a more production ecosystem. Agents will be able to write useful (and high quality) software over time, their training has just under-prioritized code quality thus far
Most of my side projects have 0-1 stars and I use them daily. Stars measure sharing appetite, not usefulness.
The more interesting question to me isn't "are AI-assisted repos less starred" but "are people building more stuff that's useful only to themselves." That feels like a good outcome — software that was previously only economically viable to write at scale is now worth writing for an audience of one.
I was really confused how this could be possible for such a seemingly simple site but it looks like it's storing + writing many new commits every time there's a new review, or new financial data, or a new show, etc.
Someone might want to tell the author to ask Claude what a database is typically used for...
json in git for reference data actually isn't terrible. having it with the code isn't great, and the repo is massively bloated in other ways, but for change tracking a source of truth, not bad except for maybe it should be canonicalized.
It's not a terrible storage mechanism but 36,625 workflow runs taking between ~1-12 minutes seems like a terrible use of runner resources. Even at many orgs, constantly actions running for very little benefit has been a challenge. Whether it's wasted dev time or wasted cpu, to say nothing of the horrible security environment that global arbitrary pr action triggers introduce, there's something wrong with Actions as a product.
Can you? My understanding is that AI cannot claim copyright and my assumption would be that copyright law immediately extends authorship to the user operating the AI (or their employer).
So you're suggesting that an AI translation of, say, a novel removes human authorship from the result? Unless a human goes in and makes further "substantive transformations" to the AI generated work?
And if that's not what you are saying then how are you determining that prompts to and AI are not copyrighted by the author of the prompt? The results are nothing more than a derivative work of the prompt. So you are faced with having to determine whether the prompts themselves or in combination are copyrightable. Depending on the prompt they may or may not be, but you can't apply a blanket rule here.
(Notwithstanding that Claude inserts itself explicitly as co-author and the author is listed on the commit as well)
> The Copyright Office affirms that existing principles of copyright law are flexible enough to apply to this new technology, as they have applied to technological innovations in the past. It concludes that the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements. This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts.
People have tried over and over again to register copyright of AI output they supplied the prompts for, for example, in one instance[1], someone prompted AI with over 600+ cumulative iterations of prompting to arrive at the final product, and it wasn't accepted by the Copyright Office.
> In March 2023, the Office provided
public guidance on registration of works created by a generative-AI system. The guidance explained that, in considering an application for registration, the Office will ask “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.”
You are going to have to prove that things Claude stamps as co-authored are not the work of an "assisting instrument". It's certainly true that some vibe-coded one-shot thing might not apply.
I would also note that the applicant applying for copyright in your linked case explicitly refused guidance and advice from the examiner. That could well be because the creation of that specific work was not shaped much by that artist's efforts.
I wouldn't read too much into that when discussing a GitHub repo. It really will depend on how the user is using Claude and their willingness to demonstrate which parts they contributed themselves. You need to remember that copyright extends to plays and other works of performance. Everything the copyright office is saying in your linked ruling suggests that an AI-implementation of a human-design is copyrightable.
Yes, but the point is that the AI output could still be covered by the definitely-human copyright in the prompt, just not a new copyright in the output.
For example, machine-translating a book doesn't create a new copyright in the new translation, but that new translation would still inherit the copyright in the original book.
Stars measure popularity, not output. The more interesting signal would be: what percentage of those repos got to v1 at all? My guess is AI tooling dramatically increases the ratio of 'ideas that actually shipped' to 'ideas that stayed in a notes file.' That's not captured by stars.
It's a great idea to track releases. Won't be a perfect measure since not all things are packaged on Git, but adding it to the roadmap. Thanks for that.
I have many GH repos, most have no stars. Probably because most of what I write is not very useful to other people due to quality or use case. I would say this is true of most fully human-created repos on GitHub.
It looks like my one-star repository [1] came close to making this person's leaderboard for number of commits (currently 5,524 since January, all by Claude Code). I'm not sure what that means, though. Only a small percentage of those commits are code. The vast majority are entries for a Japanese-English dictionary being written by Claude under my supervision. I'm using Github for this personal project because it turned out to be more convenient than doing it on my local computer.
Thanks for the recommendation. I didn’t know about forgejo.org.
The main convenience of Github for me is the ability to send preprepared prompts to Claude through its web interface or the mobile app and have it write or revise a batch of dictionary entries in the repository. I can then confirm the results on the built website, which is hosted on Github Pages, and request changes or reverts to Claude when necessary. Each prompt takes ten to thirty minutes to carry out and I run a dozen or more a day, and it is very convenient to be able to do that prompting and checking wherever I am.
When I have Claude make changes to the codebase, I find that I need to pay closer attention to the process. I can’t do that while sitting in restaurant or taking a walk like I do with the prompting for dictionary-entry writing. The next time I start a mostly (vibe) coding project, I’ll look into Forgejo.
Thanks! The dictionary should be more or less finished in a few months. If you or anyone else might find it helpful for studying Japanese, feel free to use it, copy it, and adapt it however you like.
I'm one of those zero star repos. I've been using Claude Code for some weeks now and built a personal knowledge graph with a reasoning engine, belief revision, link prediction. None of it is designed for stars, its designed for me. The repo exists because git is the right tool for versioning a system.. that evolves every day.
The framing assumes github repos are supposed to be products.
Wait a minute! Ha, just saw this. The knowledge graph I mentioned is a separate project (heartwood on my profile). Different angle from propstore but I think we're circling the same problem, conflicting claims that shouldn't be silently resolved. Added my email to my profile now.
I used Claude code to build a custom notes application for my specific requirements.
It’s not perfect, but I barely invested 10 hours in it and it does almost everything I could have asked for, plus some really cool stuff that mostly just works after one iteration. I’ll probably open source the code at some point, and I fully expect the project to have less than two stars.
Still, I have my application.
For anyone that’s interested in taking a look, my terrible landing page is at rayvroberts.com
Auto updates don’t work quite right just yet. You have to manually close the app after the update downloads, because it is still sandboxed from when I planned to distribute via the Mac App Store. Rejected in review because users bring their own Claude key.
One downstream effect of "agents can publish code" is that the trust signals weve relied on for years (stars, maintainer reputation, issue history...etc) got noisier. I don't think that means the ecosystem collapses, but it could mean we need to separate provenance from popularity.
If an automated system is going to generate and then publish artifacts at scale, you gonna want a verifiable chain of custody. Like which principle authorized the pub, what policy constraints applied (I mean like license scanning, dependency allowlist...etc), an then what checks passed (tests, static analysis, supply-chain provenance). Without this the default consumer posture becomes "treat everything as untrusted," whidh is expensive and slow adoption of legitimate work too.
I suspect we end up with something like "signed built receipts" becoming normal for small projects as well, not because everyone loves ceremony, but becauses the alternative is an arms race of spam and counterfeit maintainers.
Absolutely! I think the real stats will far exceed what we can see on public GitHub. That said, going through some of the top "performers" by commit and line count - I am surprised by how many people have all their code in public repos.
Maybe because people are using claude to to write code for themselves, to scratch their own itch, and upload it to the world just because. The value of code can't be measured in star counts.
I think the value right now in LLM code assist tools is in small projects: small reusable libraries or proof of concept “I want this app even though almost no one else does” types of projects.
For libraries: still probably mostly useful for personal code bases, but for developers with enough experience to modularize development efforts even for personal or niche projects.
I am bothered by huge vibe coded projects, for example like OpenClaw, that have huge amounts of code that has not undergone serious review, refactoring, etc.
I am not sure it's meant to be a negative thing. Obviously, a lot depends on the context here.
But, I've developed a dozen or so projects with Claude code. I am meant to be the only user.
I am maintaining a homelab setup (homelab production environment, really) with a few dozen services, combination of open source and my own - closed sourced - ones.
I had tons of ideas of how to set things up. It evolved naturally, so changing things was hard. Progress was quite slow.
Now, I have a pretty much ideal end-state - runs on auto-pilot, version bumps mostly managed by Renovate, ingress is properly isolated and secured (to the extent I am familiar of).
I was able to achieve things I wouldnt've otherwise in that time. I skipped parts I did not care about and let LLMs drive the changes under supervision. I spent more time on things I did care about, and was interested in learning.
Yeah, most of my LLM code is sitting closed source and that's by design.
None of your original products are actually closed.
If you go on other account and ask LLM about your projects you'll pretty much get all the code you wrote using LLM again.
That's my gripe with LLMs, most of my friends across the globe are working on similar projects. They are 90% same. You are burning tokens thinking they are doing some new innovation.
I pretty much google for things if they exist, i don't build them.
I'd like to see projects in spaces where nothing exists like a good CAD for 3d printing etc...opensource. but nobody is building all that.
I have developed five applications for my Mac using claude that I made for myself. I'm perfectly happy with the setup. The projects help me solve daily problems. I could never have had the time or the skill to do them myself.
The framing here is interesting — low stars doesn't mean low quality or low usefulness. A lot of internal tooling, personal projects, and niche libraries have zero stars but real users. The metric conflates 'AI is helping people build things nobody uses' with 'AI is helping people build things that aren't publicly popular.' Those are very different claims.
The idea with Claude writing code for most part is that everyone can write software that they need. Software for the audience of one. GitHub is just a place for them to live beyond my computer.
I cannot understate how much of an improvement that is. If I had a dollar for all the shit I made myself, the old fashioned way, that got 0 attention at all? I'd have enough for a month or two of claude
Even if that stat were compared directly to the base rate (human output), it could easily be explained by correlating strongly with Claude usage skewing towards new repos.
the more interesting signal in that data is about intent, not quality. most of these low-star repos probably aren't failed open source attempts - they're personal tools that were never meant to be shared.before ai-assisted coding, the effort-to-build ratio was high enough that most personal scripts stayed on a laptop or in a private gist. pushing to a public repo implied an implicit claim that someone else might want this. now the build cost is low enough that people just push things to git for their own version history and move on.what's actually happening is that git is becoming a personal dev journal as much as a collaboration platform. stars were always a weak proxy for value, but they're especially wrong for this use case.the 90% number probably also undercounts the real extent of this - most serious claude code usage is on private repos and internal tooling that never touches public github at all. the 50b lines stat would look very different if you could see total token output vs just github-public-linked output.
It would be very interesting to see how much of this is the "audience of one" type of project - i.e. personal scripts - vs new developers/vibe coders trying to start an app. I have definitely been surprised by the scale of some of the repos that seem to be vibe-coded. People who seem to have no history in development are building game engines, and payroll systems, and Broadway review websites.
Unfortunately that type of analysis would take a bit more work, but I think the repo info and commit messages could probably be used to do that.
The security implication of this shift is underappreciated. A repo that was never meant to be shared was also never security-reviewed. Personal tools built fast tend to have hardcoded API keys, credentials committed during a "just get it working" phase, and file system access patterns that weren't meant to be public.
The 50B lines across those low-star repos isn't just an interesting metric about usage patterns. It's a significant amount of unreviewed code sitting in public repositories. Stars were never a quality signal, but they were at least a proxy for "someone other than the author looked at this." That selection effect disappears entirely when the build cost drops to near zero.
I hate everything about this headline and metric. As a lifelong graphics programmer from Pentium U/V pipeline assembly optimisation days: so fucking what.
I have never cared about LinkedIn or GitHub stars or any of those bullshit metrics (obviously because I don't score very highly in them), and am enjoying exploring a million things at the speed of thought; get left outside, if it suits you. Smart and flexible people have no trouble using it, and it's amazing.
Rather measure how much I've learnt and created recently compared to before, and get ready for some sobering shit because us experienced old dudes can judge good code from bad pretty well.
This is just base rate neglect though. Something like 98% of all GitHub repos have <2 stars regardless of how they were made. If 90% of Claude repos have <2 stars that actually means they're outperforming the baseline...
Yeah, but all these internal and not so internal tools I baked with it are great - they solve my own problems - and without LLMs I would never have a chance to implement even 20% of that.
How long does it normally take projects to get stars though? You're not going to have a project with 100+ stars overnight or even within a month, you have to promote the project?
Depends widely on the target audience. In my case, targeting Julia developers who want to package their applications into installers to reach 100 stars took 2 years - https://peacefounder.org/AppBundler.jl. If I were to target Python developers, I would have many more stars.
It depends on how much you promote your repo and how big it is. I know when my repo gets posted somewhere because I'll get a little burst of stars for a few days and then it'll calm down until it's posted somewhere again. Much larger repos will get stars at a more constant rate as they reach a critical liftoff velocity.
The HN headline is at least misleading, because I suspect a majority of Claude usage is at the enterprise level (deep pockets), which goes to private GitHub repos.
I mean, most of the code that I have written to Github with normal human intelligence also goes to Github repos will less than two stars. They're usually repos that I create and no one else touches.
Is this surprising in any way? People who let Claude Code attribute commits to itself are probably vibe coders who delegate all the work. It's expected that there will be a growing number of new projects.
Who goes out of their way to “star” a GitHub repo, and what does that even mean? Is it a “like” button, a tip jar, or a bookmark? Does this bare any relationship to the importance, quality, novelty, trustworthiness, or any other property of the repo other than number of stars?
At a glance this may read as “most of this code isn’t valuable to others” but reality is probably complected with “this type of code is reducing the need for shared libraries”.
The 2 stars or fewer metric may show one thing. We’re moving from an era of 'open source as a digital monument' to 'open source as a disposable scratchpad.' Not that the code is slop, it’s that the cost of creating a repository has dropped to near zero.
Stars ceased to be relevant a long time ago, around the time Github went from a beloved pillar of the open-source community to just another facet of the Microsoft behemoth.
I wonder if there's a critical failure mode / safety feature of our species for some percentage of the population to always dislike whatever some other large percentage of the population likes.
As if it's to prevent the species from over-indexing on a particular set of behaviors.
Like how divisive films such as "Signs", "Cloud Atlas", and even "The Last Jedi" are loved by some and utterly reviled by others.
While that's kind of a silly case, maybe it's not just some random statistical fluke, but actually a function of the species at a population level to keep us from over-indexing and suboptimizing in some local minima or exploring some dangerous slope, etc.
Did we democratise software engineering? Seriously, I created a bunch of tools that I find useful without the bloated framework issues that are present in software nowadays. Jokes on me if something does not work.
Toggling the stars shows 50b lines of code created across all projects, only 5b on projects with 2+ stars since Claude Code launch. Kind of eye opening where these Claude Code tokens are going.
What percentage of GitHub activity goes to GitHub repos with less than 2 stars? I would guess it's close to the same number.
reply