Costly signals of skill
Costly signals of skill
Why the web3 version of LinkedIn endorsements is to get rid of them entirely
At every stage of my career, I’ve been told to go work for a named company where I can “build my brand and open doors.” That was good advice, but the kind that you give your kids to ensure they’ll be successful even if you took a riskier route and loved it.
Instead of following that path, I’ve worked for a number of relatively unknown startups focused on big and difficult problems I care about. I’m absolutely not complaining — it was via privilege that I could be in the position to make those decisions. But in order to allow more people to pursue entrepreneurial paths, we have to make the reputation system less biased, less based on wins and losses and more based on the quality of the work.
I am going to make the argument that the developing next iteration of the internet (or web3) will disentangle reputation from affiliation via more frequent and more granular endorsements. It’s also an answer to the question of “what’s the next iteration of LinkedIn” except the answer isn’t an iteration but a replacement of the existing system entirely. As Miguel Piedrafita said well — we need a first principles approach.
Put another way, we have to resist this skeuomorphic temptation to think that everything is better on the blockchain. Some things, when put on blockchains, just become the same as they were before, but on a blockchain — web3’s version of the “garbage in, garbage out” principle.
No such thing as a free vouch
LinkedIn endorsements make sense logically — in the “real world” we rely on our network to vouch for us, tell others what we are good at, and generally help us get new opportunities. But when those endorsements are brought online and into a massive social network they become tragically, and often hilariously, gamified. There is nothing stopping me from endorsing 1000 people in the hopes that some of them will endorse me back.
If we were to simply take this existing system and put it on a blockchain it would absolutely be garbage in, garbage out. An NFT of someone endorsing you is certainly verifiable, but it’s not like the issue with LinkedIn endorsements is that people assume they’re fake.
To get better endorsements, we have to make them costly. They need to cost time, money, or status. LinkedIn endorsements are intended to cost status, but compared to an effective status requiring endorsement — such as an email intro — LinkedIn endorsements hardly draw down your status capital at all.
Money and time tend to be purer signals of costly endorsements because they are fairly zero sum. I can’t really spend money or time without losing it. But the predominant time and money based endorsements in web2 are affiliation and role. “Early at Uber” or “Product at Stripe” vibes. In other words, Stripe (a great company) clearly thinks you're worthy of their time and their money, because they are paying you to work there. The issue is that if you only have two or three shots in the crucial early part of your career, and if reputation is based on working for “winning” companies, then you’re going to go to companies with existing brands. You’re going to be less risk tolerant.
What we need is more frequent and more granular data. We can’t evaluate a slugger after three at bats - we need three hundred. This is the one liner point of this article. People will get endorsed through actions, not words, because web3 data is granular, open, and trustworthy. To give some concrete examples, here are some on-chain signals that would make high quality non-gamifiable endorsements:
If you hire me for a project, and then you hire me for two other projects in the future, you probably believed that first and second project to be successful.
If we work together for many hours over a significant period of time, you probably feel I’m worthy of your time.
If you fund me multiple times, you clearly believe I have done and will do good things with that funding.
Etc (note that repeat signals of trust are key, and could happen via mixed sources like first we work together and then later you fund me)
At first glance this looks like the same system we have already. It is true that all of these signals are available in web2. I’m arguing that we can get them at far greater frequency and granularity in web3. There are a few key reasons for that:
There’s more data. The norm in web3 is to work in multiple DAOs, and to move fluidly between roles and projects. If we can get that data on-chain, it can be a fairly constant stream of activity.
It’s easier to aggregate. Composable identities (whether via dIDs or wallets) and interoperable platforms mean we can pull in data automatically. That means we aren’t limited by the kinds of signals that people are willing to manually add to the places they build their reputation.
Money is more open and free flowing. Flows of money are shockingly open in web3, and while there will be much more nuance built in the future we can expect there to be far more transparency when it comes to endorsements that cost money. This isn’t just about DAOs paying people salaries, it’s about the ability to make a $50 investment via an NFT crowdfund instead of the floor being a $50k seed investment. As Balaji said, web3 makes everyone an investor.
I’m not sure I can fully imagine purely programmatic hiring at least in the foreseeable future, but Greg Isenberg’s point here on the power of more granular data via NFTs is absolutely the direction I’m talking about.
What building blocks do we need?
How might we make the system we imagined above more of a reality? What building blocks do we still need?
Contribution and role data. Right now there is simply no way to know who works in what DAO, who is in what guild, or who has done what project. You could guess based on who is publishing on Mirror, listed as an admin on Snapshot, attached to a multi-sig on Gnosis, has a badge on Discord, etc — but you’d never quite capture it. We believe this will ultimately work via non-transferrable NFTs that are distributed to contributors with metadata that corresponds to their role, duration, team, and so on. The more granular the data, the better. If you’re working on or thinking about this problem, reach out — we have a cool group of DAOs and people jamming on an initial solution.
Clear applications for contribution and affiliation data. Valuable applications of this data create the incentives for bringing more of it on-chain. This is one of the primary reasons we launched Backdrop, but there will be many different platforms all building on top of this same data. For example, when a DAO airdrops tokens it might do so programmatically just to people holding a certain contribution NFT. In order to keep meetings, documents, or events private, a DAO might limit access to people holding that NFT in their wallets. The key variable here is just time - the space will mature and NFTs will increasingly be used as digital access cards/passwords/badges. Some of the best applications to watch in this space right now are Rabbithole (a good bet on where a lot of the data will be built) and Station (still in beta, but seems to promise to both read on-chain resumes and mint NFTs to help guide people to their next role).
Cleaner graph/standards. This will be an obvious point to anyone building in web3 at the moment, but the data is a (beautiful) mess. If you want to use a contribution graph right now — which people are connected to which DAOs and which projects and so on — you have to do quite a bit of manual work. For example, there’s no ENS type solution for projects so you can’t actually know which Snapshot belongs to which Discord server or Gnosis wallet. There’s also a mountain of data you might want but can’t currently get access to, like who has been getting the most tips via a system like Sourcecred or Coordinape — data that would basically tell you who the key contributors are in a given project or DAO. To allow more applications to build on top of this interoperable graph, we need open standards and more than likely a protocol to do some of this work and produce an open graph that others can use. For more on the topic of the tipping point to interoperable data read my previous piece “Mirror and tipping point to interoperable content, social, and economic graphs.”
All of the above is basically a specific example of the app-infrastructure cycle Fred Wilson described in "The Myth of the Installation Phase." We are at the point where we need the initial valuable apps to pull the data out and do the work of structuring it for many more products to be built on top. After all, the platforms that own this data aren’t hoarding it in private vaults - that was web2 vibes.
How open will we be?
One of the things I find most interesting about this general line of thinking is that there’s still such a long road to travel to fulfill the radically transparent vision put forth by many leaders in web3. As an example, almost a decade ago companies like Buffer made headlines when they made employee salaries publicly visible. Now in web3 there’s nothing really stopping us from knowing exactly how much every person is getting paid from the various projects they’re working with. After all, all of that data is flowing through publicly visible wallets. The issue is that right now you can’t actually know from the data what is a payment from a DAO - creating that salary graph would take work and the buy-in from the ecosystem to label the data in a consistent way.
These questions are all over the place. There is a big difference between data that is open by default and data that is open by design. I have no doubt that the ecosystem will trend towards more standards and more powerful open graphs, but it’s actually hard to know exactly where things will land. Will DAOs set themselves apart, like Buffer did, by making their work more publicly visible and allowing their employees to benefit from the points I made above? Or will that simply be the standard? Perhaps the biggest question is how will the ecosystem continue to incentivize interoperability and collaboration given that first party data is still very valuable?
Rather than simply offering questions, I can also give you my preference. I hope that work data is made open by design, and accepted as a standard across web3. That includes some level of salary data, in-DAO contribution data like Coordinape and the like, and of course project and role specific data. This doesn’t mean literally everything is transparent - there’s a huge amount of nuance in how to maintain people’s privacy as we push for the benefits that come with better contribution data. For example, it might be enough to simply know that someone is being paid by a DAO rather than exactly how much. I know from talking to many of them that the leading platforms are absolutely thinking about how to make their data open and useful in the right way. Much more thinking and many more articles by people better informed than I am will be needed to continue pushing this conversation forward.
What makes this exciting, and nerve wracking, is it’s one of the places where “the ecosystem” of web3 really has to operate like a signal organism, finding the right incentives and ways to work together to win — better than the sum of its parts. This could look as simple as token swaps between major platforms to incentivize open data sharing, but no doubt there isn’t a single silver bullet. This will take lots of work from lots of people over a long period of time. If we can make it happen, people will get endorsed through actions, not words. A robust contribution graph will enable powerful applications to be built on top of it, applications that decrease bias in hiring and compensation, increase the speed that valuable connections form, and help make the next iteration of the internet far better than what we’ve seen so far.
Big shoutouts to Rapha Menezes and Julia Lipton for the help with this article (and with everything I’m doing), and if you’re looking for the people whose thinking on this topic I’ve found inspiring (in addition to those I linked above) check out Gaby Goldbeg, Julia Rosenburg, and Rafa0.