ZH version is available. Content is displayed in original English for accuracy.
Howdy all. I'm Zack :wave:. I've been thinking about the problem of misguided AI pull requests and figured I'd throw a possible solution out there for feedback. Basically, CleverCrow lets supporters give tokens to a GitHub repo (or set of issues in that repo) for the maintainers to use to build/fix stuff. The fun implementation challenges have been around implementing the pooling dynamics and keeping the maintainers in charge while the backers are motivated to support their work.
Discussion (42 Comments)Read Original on HackerNews
The AI provider gets paid, the platform gets paid (20% is a lot in my opinion!), and the maintainer gets more unpaid work: another PR to plan, review, revise, merge, and then maintain... that's a lot of work.
If people are willing to fund an issue, why should that money mainly cover LLM tokens rather than maintainer effort? Or at least, why doesn't the leftover money go to the maintainer instead of back to the donors?
Regarding rewarding maintainer effort, I'm shooting for the value prop of "free AI", this only works if reconciliation is per-phase and liquidity is accessible across as many repos as possible. So if I had each reconcile drain the pool, there would be a lot of stalled work and human intervention required.
That said, there are probably some maintainers that don't want "free AI" and that's okay.
sounds like you do not support other people that have nothing to do with the code that you like
We are moving into a weird time where people are assuming that now we have to pay machines churn out code.
Somehow they packaged up our own ability to think and are selling it back to us. If they can get us to forget how to do it we'll be the perfect customers, dependent forever.
Do you have a source for this? I believe “personal subscription” plans on OpenAI and Anthropic are likely ran at a net loss or close to it, but all indications elsewhere are that API pricing for these companies and likely Google as well are profitable per API call [1][2]. I would definitely believe that the Chinese players are operating at a loss though if that is what you mean.
[1] https://newsletter.semianalysis.com/p/anthropic-growth-and-b...
[2] https://martinalderson.com/posts/are-openai-and-anthropic-re...
I suppose this is the most common scenario - I doubt features are not getting implemented because maintainers are lacking tokens.
I like your approach of pooling resources around specific issues. That seems a practical missing piece for aiding the maintainers.
If all you’ve got is relative pocket change they probably aren’t going to agree but if you put real money behind it and it doesn’t go against their vision of the project then most people would be willing to accept actual contracting work to expand their project.
If you like a project enough to donate to them, give them the money directly and let them decide how to spend it. This is just convoluted, weird and vaguely dystopian.
I think there's a trust continuum that culminates at giving money no strings attached, but it starts somewhere else. If my goal (as creator of CleverCrow) is to get more people to support maintainers, then I have to open the funnel wide and connect with where people start (transactionally, I'm assuming).
Its efficiency also relies on being better than whatever other platform/harness the maintainer is already using. It's limited to whatever the harness the platform provides, and they're taking a 20% platform fee on top. So I have to, instead of taking $10 from a donor, i take $10 worth of tokens, which may or not be spent more efficiently than me just going in with my claude subscription and fixing it, and I get $8 of those to run in a platform I don't control? In what world as a FOSS maintainer would I sign up for this? It just seems strictly worse than just having a platform that can back resolution of issues with real money... which already exists.
I think it's a cool idea, don't get me wrong. But it has to be a very good solution to get adopted, like, it would have to significantly streamline the operations of getting bugs resolved by a FOSS maintainer, and I think it's going to be tough for you to try to beat "fire up my favorite agent in my terminal with an already optimized setup and give it this issue that has $X attached" rather than "I have perhaps inefficient token spend from a platform I have no control over and I have to take 20% less of a donation for that privilege"
In other words, I think you've built this solution for donors, and not FOSS maintainers, but really the bottleneck and problem and who you would be selling this solution to is not donors, but rather FOSS maintainers, and that's who you need to solve for if you want a platform like this to work. The donors have the easy job: they throw money at the problem to help it get solved faster. The FOSS maintainers have the hard job: They have to understand, accept the issue, propose a sensible solution, build it out, test it, etc. And your solution just makes it harder for those end users, because now they have two paths in their development workflow to getting issues resolved that they have to maintain, the non paid path and the paid path. So you're significantly increasing the overhead burden on these people and the material gain promised to those end users as the tradeoff is not convincing at all.
This kind of makes sense in that space while the subsidies (if true) last?
Unrelated, "tokens" feels very like... back-then blockchain to me. All the craze.
It’s just an obvious example of market segmentation by charging enterprise customers many times more than personal users while selling the same product
As an aside, when you do login, CleverCrow shows projects that you've starred on GitHub to help find things you might want to support.