Extremely loud and incredibly fast
a departure from our regularly scheduled programming
I've been deep diving in crypto for the last few months and have drafted various pieces about it, but things move extremely fast and takes get stale really fast, so I'm going to try and write stream of consciousness without editing. Feedback is always appreciated, whether or not you wanted crypto content -- though I’m trying to write for “some” crypto knowledge.
Crypto changes too quickly for me to edit my articles, and it also moves too fast for most platforms to build durable moats.
OpenSea, the most popular NFT marketplace, is a clear example of a product which had an absurd amount of usage, clear product market fit, raised a ton of money, and yet very likely will get torn apart by competitors who take lower margins, such as LooksRare. LooksRare is a brand new competitor with a number of new features, but the most attractive benefit has to be their lower fee structure combined with paying out fees to token stakers. This is basically now just a lower margin business, as they now must pay out revenues, and the platform is hoping to make up for that with higher volume. Being "paid to trade" is also a nice spin, sure. I don't have a view (or position) on if LooksRare will succeed, but their team is clearly willing to give up margin for market share, and that is probably the correct path.
In theory, marketplaces should have strong network effects, and this drove huge amounts of capital into Web2 marketplaces. Some marketplaces had high switching costs (cross-listing on Airbnb and VRBO is challenging, for example) while others offered significant incentives to disincentivize switching (riders and drivers "can" switch freely between Uber and Lyft). In crypto, switching costs are often significantly lower or non-existent: the same NFT can be listed on OpenSea and on Rarible or any other place, for just the approval transaction fee. It would be interesting to see a marketplace experiment with retention incentives to exclusively list on their platform, offering frequent traders lower fees, etc.
DeFi dexes, where individuals can swap tokens for other tokens, however, cannot race to zero fees. Liquidity on a DEX is provided by individuals who (quite reasonably) demand a fee for their service, and for higher risk pairs, if that fee is too low, will not provide liquidity. This is reflexive too, that people will trade more on platforms with better liquidity and lower slippage, since that volume attracts more liqudiity, and so on. For lower risk pairs this really is a race to the bottom, though, as seen with Uniswap's new 0.01% stablecoin fee tiers, cheaper than Curve. In some stable pairs, Uniswap went from < 2% before release to more than 80% of the market today (h/t Tom Schmidt, chart below forked from kroeger0x). The release was live on Nov 12 and you can see pretty much immediately how the ecosystem changed.
There is probably some "correct" fee tier for each pair based on its risk and the markets on each exchange will stabilize to offer that pricing. Some random tail coin pair might not go to 1 basis point but I wouldn’t be surprised if a common pair such as WETH-USDC goes to 1bp, especially now that people treat Uniswap V3 as a real market making system and can use new tools such as Squeeth to hedge impermanent loss.
I would expect the end game for the various exchanges (token and NFT) to be that most end users interact with them through an aggregator (such as Matcha or 0x) with extremely low fees. Even in this space, I don't think that aggregators can expect to sit on their laurels -- Paraswap had a lot of volume but after botching their airdrop by being too conservative, their usage has dropped off significantly. Matcha also announced that they had no plans for a token, and I would think a new competitor could get a lot of volume by simply offering a token. But in each of these cases, you see that the usage is tied to both the airdrop of a token (diluting project ownership) and the existence of the token (reducing project margins).
(does this line keep going up?)
A very well-funded organization could build out an aggregator and run it at basically zero margin (take and return fees) to completely disincentivize any competition here, basically "the Amazon approach," if you would. It wouldn't surprise me to see various protocols consolidate/bundle into a single very large DAO with multiple related businesses and run them all at extremely low margins at scale. You are already starting to see centralization in DAO's, e.g. through the Rari-FEI merger. There are obviously similarities with how consumer banking evolved, both in pre-tech era and in the fintech era; one thing crypto will never change is that the only two ways to make money are to bundle and unbundle.
The projects that survive the longest have, in my mind, a team premium. It is clearly easy to copy any project and any successful IP will just be ported over across chains or whatever quickly, by a team that doesn't also have to support your existing customers. Capital is extremely fluid and happy to support anyone for the right "incentives." The second mover advantage is real -- many teams launch a product after tons of work, finding great customers and allowlisting them for the launch, slowly refining as they roll out from beta -- and then somebody else just copy pastes the thing with a different logo, no allowlists, and a flashier APY. Success is then often not driven by who comes to market first, or who comes to market best, but rather who continues to iterate and sustain and outcompete the other.
There's lots of good reasons why OpenSea didn't go for a more decentralized or token-based approach, e.g. easier to build and raise from traditional investors, but those are structural constraints which apply to the first mover. In particular, since they needed to iterate and hire lots of people to do so, they could not afford to pay out platform revenue as dividends or in buybacks; they needed to reinvest that revenue into building the platform. So with LooksRare, that team could win lots of market share in the short term, but OpenSea has more capital, probably better talent (certainly more of it), and the willingness to reinvest their revenue in growing the product. They can still "win," they just can't remain complacent. Even if LooksRare flops soon, it's clear that many traders are mercenaries for whichever platform offers them the best incentives, and other platforms such as Coinbase should be licking their chops, knowing they too could seize market share when they enter.
That, or the first mover locks up a ton of capital and doesn't allow you to sell. This is a bit like how mutual funds and closed end funds tend to work - significant transaction costs on the way out to incentivize holders. Vanguard doesn't allow you to hold their lowest fee tier Admiral funds if you sell too often either. These projects then become bets on the team and that the team will be able to deploy that treasury effectively.
The "treasury utilization bet" was the underpinning of Olympus, not the high APY ponzi or game theory stuff. I mean, the ponzi stuff definitely mattered for getting retail adoption, but fundamentally you were buying a pool of capital trading at some multiple of cash and betting that the team would do a good job managing it. Olympus Pro was a good example of generating revenue through that treasury, as were their various stablecoin farming efforts. But at that point, why would you buy Olympus at such a high multiple when you could also farm those same stablecoin pools? When I realized that, I sold out and threw the proceeds into similar farming pools. I'm not claiming Olympus was a rugpull or ponzi (they never lied about this, but plenty of people truly believed they were gonna buy lambos), just that it reflected too high of a premium for what I felt the team could execute on.
For similar treasuries to trade at a premium, you need trades which are hard for an individual to execute; this is the same argument for why hedge funds say they don't just buy SPY, since it doesn't justify their fees. So treasuries doing stablecoin farming or covered call selling (lmao) are acting rationally, but you need something else to justify that premium. Olympus had a diversified portfolio of project tokens at discounted rates (challenging to pull off without access and lots of capital) and a pretty good team. It makes sense to me for this pool to trade above the cash price but who knows what kind of multiple it truly deserves, probably less than a top-tier private equity firm (the closest public equity comparison I can think of) but also probably more than 1.
I wonder what the playbook is for a crypto business today. Your expected lifespan must be shorter than the expected lifespan of a similar traditional tech business. The instant liquidity of tokens vs illiquid shares compensates for that, so maybe the expectation is projects will live 4 years on a normal vesting schedule? Is that why each crypto cycle takes about 4 years? Or that NFT projects tend to die after the initial sale -- if there isn't a huge community response and expected future resale revenues, the logical incentive is to collect the mint fees and move on, rather than try and dig the project out.
On tech debt
Not gonna link to the thread, but a guy on Twitter pretended to be a big tech engineer and made a bunch of dogwhistle claims about how bad the companies are. The only substantive claim was one about how the company is running on years old tech stacks and how there is mind boggling amounts of tech debt. I have a couple thoughts here:
Lindy principle -- old code has survived until now and therefore is probably "fine," otherwise it would have already been rewritten. Banks are running on COBOL from the 1980's and it's because the code *works* and the consequences of rewriting and introducing a bug are just way too high. That being said, this is where ownership becomes an issue, as many of the programmers who know COBOL are getting quite old and it's not being used by new engineers, so places are willing to pay *lots* for consultants with COBOL experience.
Ownership -- the average tenure of an engineer in Silicon Valley is about 2 years. So you should expect the average lifespan of a service that that engineer owns to be about 2 years and a couple months as well. Aspects of engineering such as "documentation" or "on-call runbooks" are rarely reflected in promotion packets and performance reviews, at least less than "shipped a product."
There will always be another scrappy L5 looking for their system design checkbox willing to tear apart your monolith into kubernetes microservices. And a few years later, there will be another L5 frustrated with how many microservices there are, dependency management, and API drift; their solution will be to build a singular monolith that does everything their team needs. Perhaps the only ways to get promoted are to bundle and unbundle services.
There is no alternative -- obviously crypto moves at lightspeed, but tech companies today have lots of the same accelerationist characteristics - notably, easy access to capital and low capital moats. So any competitive space (red oceans, if you have an MBA and somehow are still reading) will move quite quickly and converge on the same ideas, anything that works will get copied. That's how you end up with about 80 different fintech apps which all have the same features and end up competing on brand. Product iteration is therefore far more important than code quality. This point does not apply as much to the true big tech monopolies, who can outsource product development to startups and instead spend time re-engineering their systems into beautiful code. Of course, getting to that point required a ton of accumulated tech debt too, and these companies are probably too large to enforce any singular set of standards on.
Short and long term problems
More broadly, tech debt is a symptom of how humans approach short term and long term problems. I theorize that we are extremely good at solving well-defined short-term problems but are extremely bad at long-term problems. You could extend this to say that we underestimate our ability to solve short-term problems and overestimate our ability to solve long-term problems, though I am usually a bit too optimistic for that.
The atomic bomb is an example par excellence of humans solving tremendous problems in a very short amount of time, by locking genius physicists in a room and not letting them come out until they had something to show for it. While we were able to solve a short term problem, the consequences were extreme and many of the involved scientists distanced themselves from the work.
Organizational behavior researchers generally find that people are most successfully motivated by either money or by mission, but in different ways.
Mission is the most effective motivator for creative work, but it's very difficult to find people motivated by your mission.
Money is a very effective motivator for well-defined problems with clear outcomes, but not as much for creative work, where you have to decide what the desired outcome is.
I am consistently amazed by how much iteration there is in crypto and how quickly things move. I can see both incentives, of money and mission, at play -- you have mission driven folks working on new tools for decentralization and cryptography (e.g. ZK proofs, coordination problems) and you have money driven folks iterating on protocol design (e.g. automated market makers). There are of course charlatans and frauds, but we still allow Chamath on CNBC, so how is this any different from public markets?