# Polis experiment lessons

> This HackMD is for the Polis Exploration group to leave insights / comments about things they learned while doing our initial EOF Polis Experiment. Especially insights that could inform future product directions.

## Dionysuz thoughts

- Restricting comments to a single statement is important. You want more statements to say "yes"/"no" instead of pass, so if a comment includes a nested comment, it's better to disentangle them into separate comments.
- The process of disentangling statements and creating higher quality ones seems possible to use as an input for training an AI.
- There is a large burden on curators creating seed comments / initializing discussion. This is a blocker into generating more Polis conversations especially at scale. If it could "form" on its own through simpler abstractions that would be awesome.
- Direct argument is useful for challenging ideas and generating insight, but maybe not the best for showing as news (or, insight can be seen separately).
- It is **easier** for humans to verify if a comment is good / high quality (and determine their agreement/disagreement/etc) than for them to come up with them.
- It seems somewhat hard to weight relative importance..? Like while one may agree with a (set) of statements, or have consensus on them, those statements don't necessarily encode the relative importance against other ones (but it seems this is more for driving decision making).
- Raymond noted that the Polis team made lots of little refinements for their algorithm, instead of this being a reason to stay with the original Polis codebase, it could be argued as a reason to diverge from it. It is exerting almost too much centralized / opaque control over the process, when it would be cool if it was more modular / separated / easy to reason about.
- I genuinely don't know if the incentives are strong enough to participate. Even I barely feel like participating with no incentive. Because I'm not even necessarilyyy confident this will help make things more productive. I worry others will just not participate, like why would they, and it's also all anonymous so they can't even ego flex.
- I feel sometimes statements don't capture the full nuance of how I feel. Like for EOF: "EOF improvers the developer experience for Ethereum developers". I feel like yes and no so I can't express my nuance, because you could have EOF on an L2 but not on L1. On the other hand it may break composability, so it's kind of nuanced / depends on where we go with L1/L2 scaling (more native or not). I feel like perhaps Polis encourages you to disregard nuance in order to come up with easier statements to vote on but this could lose important details.
- One example is this post from pcaversaccio: https://warpcast.com/pcaversaccio/0x6a050edd -> you can see how much nuance is captured in this longer form statement. I believe there exists insight in this level of nuance, but unfortunately this level of complexity is hard to work off of (so I think Polis is still good in a way to manufacture broader / simpler things we can agree on), but what I mean is the "input" may benefit from more entropy/nuance.
- Pol.is doesn't have auth anymore and I predict this will be a big issue. We don't know about bots or anything. We probably need to solve this, in other words fork Polis. Based on my research, I think their team moves too slow and the codebase is too antiquated, strengthening the argument to restart from scratch.