[Personae] trait growth

As we get close to launching Traitcaster, I've been reflecting a bit on traits in general and how they relate to Personae's long-term strategy.

What is a trait?

I think of a trait as a human-understandable label for an anonymity set.

relative value of traits

Not all anonymity sets have meaningful traits. It's certainly possible to pick a set of public keys that have no cultural relevance.

Not all traits are equal in 'value' either. Aside from the purely quantitative value of a trait (i.e anonymity set size), traits vary wildly in their 'usefulness' to Ethereum/Internet users.

How do we define this 'usefulness'? I think it's related to 'cultural relevance' to the internet. Traits are interesting if they speak to things that humans care about at a point in time on the internet. If humans think in traits rather than individuals, they're useful.

Some traits probably have evergreen value, while others are very temporal. The drama of the week (i.e. DAO governance drama, SBF trial) generates traits that matter a lot for short periods of time, but there are also historical events (i.e. beacon chain genesis, accomplished smart contract dev) that create traits that matter a little bit all the time.

measuring the value of a trait

This is still theoretical. Obviously as a strategic entity, we'd like to move towards the highest value traits and not waste too much time on the low value ones. How do we think about what traits are valuable and what aren't?

Another way to pose this question: if we had an oracle on all human activity on the internet, how would we measure the value of a trait s.t. we make good decisions about integrating said trait into our system?

I posit that the general statement here is something like "people make financial or social decisions in the context of the trait."

Apps like traitcaster give us (if properly implemented), a way to examine a social action: follows. If we see in our analytics that people often search for certain traits and/or follow individuals based on certain traits, we have some inclination in realtime of what traits matter.

Side-note: still think it's interesting to consider what sorts of financial things can be built on traits. I.e. lending collateralization ratios based on traits.

towards a 'trait engine'

Looking back at the growth challenges of creddd (so far), nouns nymz, and heyanon/heyanoun, a common thread seems to be that we had a static set of traits that didn't grow/adapt to the market. I'd posit that this factor has also limited the growth previous attempts at chain identity like degenscore.

I believe that to succeed, we need to build a 'trait engine', i.e. processes by which we can measure trait effectiveness and auto-update our trait DB as necessary.

measuring trait supply vs. trait demand

Trait supply is what traits addresses in our network have.

Trait demand is what traits humans in our network desire to make decisions around.

I think we need to measure both in understanding what traits to add, but lean towards thinking that marketplaces are easier to bootstrap on demand (and so it's more fruitful to focus on demand).

prior art in traits

There's been a lot of prior art in traits:

  • dune labels
  • nansen labels
  • etherscan labels
  • degenscore attributes

Among these, I think the most used are Nansen labels and etherscan labels. No others seem to be used that widely anymore.

Not sure what to make of this, but it's obvious to me that we should bootstrap our own engine on data from others and contribute back our learnings as well.


More thoughts soon on data infrastructure, etc