When you look at the list of projects on the Tea network, itās difficult to find something useful. Most projects are copies of āNext.js starterā and something similar with virtually no changes. They have no value and are essentially spam.
There are also projects that, based on one repository, create dozens or hundreds of npm packages and use each other as dependencies. And maintainers of these projects are in the top of Leaderboard. For example āhello-*ā projectsā¦
And these projects have millions of Total Staked by Community (TEA)
And of course, I think such kind of developers use multi-accounting.
If this remains the case in the future, then it will be more profitable to develop smart bots for spamming Tea Network than to develop useful Open-sourceš¤
We are aware of these issues. This is also a āvaluableā aspect of the testnet, as it allows us to identify techniques used to abuse the tea protocol.
At the same time, we have implemented new anti-spam measures based on these findings. Moreover, more major updates are planned to ensure that such abuses are not possible on the mainnet.
It will be difficult to confirm whether a project is real or not without human review. Unless the scoring involves many aspects such as repository age, activity, number of commits, number of stars, number of downloads, number of contributors, number of mentions on social media, whether it has documentation or not, etc.
It seems that now it is more influenced by the number of dependent projects which is easier to manipulate.
It will be difficult to confirm whether a project is real or not without human review
100%
repository age, activity, number of commits, number of stars, number of downloads, number of contributors, number of mentions on social media
All of this can be simulated by bots if thatās what it takes to make money.
There are already tens of thousands of new repositories with garbage on npmjs.com thanks to Tea)
Maybe a LLM could be able to tell just by looking at the codebase if a project is BS.
Beside, stars are not so easy to fake, if the account age of the pepoles starring is took into account.
Iām not saying that itās an easy problem to solve, just that itās not nessesarely a lost battle.
@garroneJoseph
The age of the project is not a good indicator. There are plenty of half-dead projects with a huge age. There are archaic technologies like jQuery that are also very old. But there are a lot of young and promising projects. Reliance on age may demotivate developers to open-source their very old developments because of the small age of new repositories.
LLMs can certainly help in some ways. But you have to realise that it is a rather simple technology, without any magic, with a limited range of applications. LLM builds the missing parts of an āimageā within a given context. LLMs are good for search, classification, dialogues, but not for decision making when they are being purposely tricked.
For example, someone will add 20 lines to an old useless project and make it useful, and someone will add 20 lines of working but useless code to the āNext.js starterā. If you let the LLM decide which of these projects to keep, the cheat projects will have a much better chance than the real ones. And itās only a matter of time before people realise what coefficients need to be influenced to please LLM. And a huge number of really useful projects will fly into the spam filter.
As I understand, the goal of the Tea network is to support the most useful open-source projects. And my opinion, the degree of usefulness cannot be understood by stars, dependencies and other measurable indicators. All metrics can be easily tweaked. There is no bots or market for buying stars on github now, but it will be)
Only recommendations or contributions from real people in the network can determine usefulness.
Maybe there should be some rounds and sandboxes for projects, maybe voting, I donāt know, havenāt thought much about it
And itās true that any algorithm put it place will be subject to a lot of attempt to break it and this is undesirable.
We donāt want NPM to be flooded to generate fake download count, we donāt want bot created to open fake issue and we donāt want an economy around buying github star to emerge.
(Although it already exists)
That being said I was refering to the age not of the project but the age of the github accounts of the pepoles that have starred the project.
Age of the github account is also very synthetic indicator. It didnāt have very strong correlation with open-source project value or usefulness and can be a very coarse filter that will reject a lot of useful projects
Correct, but I think that we can at least esablish that if not all impactfull project have a large amount of star from senior devlopper at lease all the project that do are impactfull to some extends.
In my view itās a beter trade of to close the gate to some legit projects if itās the price to pay to avoid having a ton of spams like itās currently the case.
Only stars from people who have stars and have had stars for a very long time is not a bad filter. But this is closer to proof of authority and recommendations. And then it is reasonable to build Tea-rank around recommendations among network users without relying on third-party stars.
I have concerns about an internal system of recommendation/voting, as there could be financial incentives to gatekeep or to admit non-legitimate projects.
There are numerous options to explore for objectively quantifying the impact of an open source library. For instance, we could agree that all packages depended upon by the top 1000 libraries of a stack are automatically accepted.
However, all the conditions I can think of are sufficient but not necessaryā¦
I have concerns about an internal system of recommendation/voting, as there could be financial incentives to gatekeep or to admit non-legitimate projects.
By using stars and number of dependencies you are just using an external recommendation system. Itās the same thing, but you also lose control over it
For instance, we could agree that all packages depended upon by the top 1000 libraries of a stack are automatically accepted.
Really not a bad rule, but weāll only reward those who donāt need the award already.
By using stars and the number of dependencies, you are just using an external recommendation system. Itās the same thing, but you also lose control over it.
Yes, but it could be made more objective. Losing control over it is the point.
Btw are you part of the team or just an OSS maintainer? What are the project youāre working on?
Btw are you part of the team or just an OSS maintainer? What are the project youāre working on?
Iām not part of the team and Iām new to open-source.
My main job right now is developing a PaaS for web application development.
This project just seems interesting to me and I decided to open-source some of my old development and toy projects add it to Tea and see how it works.
I have accumulated a huge amount of code for everything over the last 25 years, but I have no motivation to transform it into reusable libraries and maintain it.
I think a project like Tea and the possibility to get rewards could be a good motivation to open-source useful code for people like me.
