People used to care about decentralization—or pretended to, at least. When I first started working on decentralized finance, onchain perpetual futures markets were the cutting-edge. Layer two scaling solutions were being rolled out for Ethereum so transactions could include enough compute power to process trades at a reasonable cost. Dynamic funding rate mechanisms were engineered so liquidity providers could earn trading fees without price exposure. Decentralized oracle networks were being developed to provide prices associated with specific timestamps to prevent front-running.

The goal here was to create onchain derivatives—freely tradable tokens that track the value of any price feed. The derivatives markets that generated the most trading volume on these protocols were the most volatile crypto assets. Users willing to use cutting-edge financial products had a high risk tolerance and were seeking a way to trade long-tail assets with leverage. In other words, the early product-market fit was degenerate shitcoin casinos. If the tech could handle this, building derivatives for “real world assets” like foreign currencies, commodities, and equities (with exponentially higher liquidity and less volatility) would be a walk in the park.

At the time, the vast majority of crypto trading was happening on centralized exchanges like FTX, Binance, and Coinbase. This was a major embarrassment. The entire innovation of this technology is to be able to hold and trade digital assets without middlemen. But instead most people still send their money to private businesses and pay them fees to trade within their closed systems.

Fast forward to 2025. We’ve made virtually no progress. Here are some recommendations I have for anyone involved.

Because of the leveraged shitcoin trading, I think many shy away from perps (read: derivatives) as a use case in general. But stablecoins are just derivatives of US dollars! I think the consensus is that the global adoption of stablecoins is the biggest success story of crypto so far. It used to be that only the very wealthy could opt out of their local currency, but now anyone can.

Unfortunately, the most widely used stablecoins are unapologetically centralized. Like banks, they use deposits to earn yield (which they keep), are trusted to handle issuance/redemption, and can freeze funds arbitrarily. Everyone seems to have lost interest in making scalable decentralized stablecoins. Ethena became the name brand for a promising approach that could be built on decentralized perps exchanges, but implemented it with centralized perps exchanges. Meanwhile, the standards for what constitutes decentralized perps have rapidly regressed in the name of “user experience”.

Reject that big suites of upgradeable contracts are an anti-pattern. [1] How many lines of code is Binance’s trade execution engine? I assume they’re regularly shipping improvements. Practically, it’s reasonable for protocol engineers to retain upgradeability for now, especially as dependencies like sequencers and oracles are under active development. This is just a much more difficult engineering challenge than something like Uniswap; you can’t build it in a “lab setting” effectively. The north star is renouncing upgradeability on the whole stack for these systems—we’re just not there yet.

In addition to decentralized stablecoins built on perps, I would love to see more flavors of tokenized perps positions. [2] No one wants to learn about margin accounts. Everyone already understands how tokens work. (Tokens also make accounting straightforward.) Make every wallet application a perps interface.

You might dislike that oracles introduce new trust assumptions, but otherwise you’re just trusting multisigs. Designing decentralized oracle networks is an interesting problem space and it’d be great to see more startups working on this instead of Ethereum knock-offs. The world computer could use more fun and weird peripherals.

UMA works but isn’t perfect. Most of DeFi still relies on prices pushed by oracle networks, which are useless for trade execution and have no mechanism for price consumers to compensate providers. (This is fixable.) Liquidity providers for decentralized perps continue to get wrecked for having stale data on asset risk. There have been some attempts at specialized risk oracles, but it should just be standard to have liquidity-related metrics available with prices to use in functions for parameters like maximum open interest.

I’ve never met venture capitalists more cynical—or just willing to say the quiet part out loud—than those in crypto. (I prefer it to the stereotypical smarmy tech VC, for what it’s worth.) Figuring out how to capture value while investing in systems that are made to remove middlemen is a challenging job! It should be unsurprising that so much capital finds its way into strange projects.

Meanwhile, retroactive public goods funding, futarchy, and other experiments around philanthropy are neat, but I’d like to get a better sense of what’s being learned from them. Open source should be a prerequisite for digital public goods, but let’s not be afraid to stay opinionated about what projects should receive donations. [3]

Governance power isn’t a perk—it’s a job with responsibilities. Cultural value is real (see memecoins and NFTs) but all the relevant tech is downstream of derivatives anyway.

Buybacks and dividends give tokens value. If a protocol performs a function that users will pay a fee to execute, some or all of that fee can benefit tokenholders. It’s confusing to me that this seems debatable.

There won’t be any mainstream adoption of this in its current form because people don’t want their coffeeshop to know their net worth after buying a latte. At the same time, the financial system is a critical tool for law enforcement (whether you like it or not).

It looks like some form of onchain KYC is going to be the short-term solution, but I’m not sure moving your financial freedom into Brian Armstrong’s hands from Jaime Dimon’s is what we’re here for. [4] The Tornado Cash case was interesting, but I’m not convinced society will accept money laundering while replacing the status quo. (Though perhaps it’s inevitable? I’d be surprised by every possible outcome here.)

For a while, I thought the creation of a truly meritocratic and free market for consumer financial apps would drive mass adoption. I was wrong. Being able to use any block explorer, any wallet application, any custom script, etc. is inarguably better than being at the mercy of my bank’s dashboard. But it’s not so much better that I’d expect people to move over all of their finances, even if privacy were allowed.

Also, how is there no wildly popular open source crypto tax software yet? This should be a quintessential public good.

Add a thin offchain layer to your protocol with functions including descriptions that an open-weights large language model can understand. If you can get an LLM to successfully interface with your protocol, LLMs will be able to generate GUIs on demand—if not just take care of users’ needs directly.

Wealthy people rely on other people to manage their finances by chatting with them, never touching GUIs. This will soon become available to everyone. I believe this is the 10x improvement for end users that will drive mainstream adoption. Everyone has the ability to do things on DeFi rails that they can’t with TradFi and AI agents will make this obvious.

Even with LLMs objectively assessing financial infrastructure options and managing money on behalf of users, the brands surrounding the technology will remain important. There are dimensions of identity and reputation that come with deciding how you handle your money. This isn’t going away.

At this point, I think “crypto” is unsalvageable as a brand, associated with crime and presidential memecoin pump-and-dump schemes. “Web3” is lame. (“Decentralized compute” is the best I’ve got.) The brand for the runner-up world computer is trash. [5] Ethereum is the real deal. It’s not a business. It’s Linux. It’s weird. It’s cypherpunk. Let’s keep it that way.

Ethereum is only ten years old. When the world wide web was this age, Pets.com was collapsing, and the idea that the internet could replace cable television was fringe. If this is really going to secure the world’s financial system, it’ll take a little while to get right. It starts with global access to Wall Street via centralized stablecoins. Replacing Wall Street is the end game.

The recent breakthroughs with large language models is serendipitous. They’ll cut through all the bullshit. This is bad news for traditional finance and crypto grifters alike.

So go put your energy into building financial infrastructure that’s appealing to my future superintelligent money manager. Thanks!

[1] You might think that you can progressively make a perps dex immutable, but this is a mirage. Either the contract holding the collateral is upgradeable or it isn’t. If integrated subsystems are immutable, that’s nice but you can just change them out on the contract that’s holding the funds.

[2] I would’ve guessed tokens would be the “primitive” that you’d build perps on, but it turns out it’s the other way around. Go figure.

[3] I say this with love. Development continues on Cannon (an open source dev tools project I contribute to which supports multiple teams building decentralized perps) largely thanks to the generosity of Optimism.

[4] I’ve seen “compliance-friendly” used as a buzzword here. This would’ve been parody just a few years ago.

[5] Yes, they retracted the ad. But was anyone that surprised they shipped it in the first place?

People used to care about decentralization—or pretended to, at least. When I first started working on decentralized finance, onchain perpetual futures markets were the cutting-edge. Layer two scaling solutions were being rolled out for Ethereum so transactions could include enough compute power to process trades at a reasonable cost. Dynamic funding rate mechanisms were engineered so liquidity providers could earn trading fees without price exposure. Decentralized oracle networks were being developed to provide prices associated with specific timestamps to prevent front-running.

The goal here was to create onchain derivatives—freely tradable tokens that track the value of any price feed. The derivatives markets that generated the most trading volume on these protocols were the most volatile crypto assets. Users willing to use cutting-edge financial products had a high risk tolerance and were seeking a way to trade long-tail assets with leverage. In other words, the early product-market fit was degenerate shitcoin casinos. If the tech could handle this, building derivatives for “real world assets” like foreign currencies, commodities, and equities (with exponentially higher liquidity and less volatility) would be a walk in the park.

At the time, the vast majority of crypto trading was happening on centralized exchanges like FTX, Binance, and Coinbase. This was a major embarrassment. The entire innovation of this technology is to be able to hold and trade digital assets without middlemen. But instead most people still send their money to private businesses and pay them fees to trade within their closed systems.

Fast forward to 2025. We’ve made virtually no progress. Here are some recommendations I have for anyone involved.

Because of the leveraged shitcoin trading, I think many shy away from perps (read: derivatives) as a use case in general. But stablecoins are just derivatives of US dollars! I think the consensus is that the global adoption of stablecoins is the biggest success story of crypto so far. It used to be that only the very wealthy could opt out of their local currency, but now anyone can.

Unfortunately, the most widely used stablecoins are unapologetically centralized. Like banks, they use deposits to earn yield (which they keep), are trusted to handle issuance/redemption, and can freeze funds arbitrarily. Everyone seems to have lost interest in making scalable decentralized stablecoins. Ethena became the name brand for a promising approach that could be built on decentralized perps exchanges, but implemented it with centralized perps exchanges. Meanwhile, the standards for what constitutes decentralized perps have rapidly regressed in the name of “user experience”.

Reject that big suites of upgradeable contracts are an anti-pattern. [1] How many lines of code is Binance’s trade execution engine? I assume they’re regularly shipping improvements. Practically, it’s reasonable for protocol engineers to retain upgradeability for now, especially as dependencies like sequencers and oracles are under active development. This is just a much more difficult engineering challenge than something like Uniswap; you can’t build it in a “lab setting” effectively. The north star is renouncing upgradeability on the whole stack for these systems—we’re just not there yet.

In addition to decentralized stablecoins built on perps, I would love to see more flavors of tokenized perps positions. [2] No one wants to learn about margin accounts. Everyone already understands how tokens work. (Tokens also make accounting straightforward.) Make every wallet application a perps interface.

You might dislike that oracles introduce new trust assumptions, but otherwise you’re just trusting multisigs. Designing decentralized oracle networks is an interesting problem space and it’d be great to see more startups working on this instead of Ethereum knock-offs. The world computer could use more fun and weird peripherals.

UMA works but isn’t perfect. Most of DeFi still relies on prices pushed by oracle networks, which are useless for trade execution and have no mechanism for price consumers to compensate providers. (This is fixable.) Liquidity providers for decentralized perps continue to get wrecked for having stale data on asset risk. There have been some attempts at specialized risk oracles, but it should just be standard to have liquidity-related metrics available with prices to use in functions for parameters like maximum open interest.

I’ve never met venture capitalists more cynical—or just willing to say the quiet part out loud—than those in crypto. (I prefer it to the stereotypical smarmy tech VC, for what it’s worth.) Figuring out how to capture value while investing in systems that are made to remove middlemen is a challenging job! It should be unsurprising that so much capital finds its way into strange projects.

Meanwhile, retroactive public goods funding, futarchy, and other experiments around philanthropy are neat, but I’d like to get a better sense of what’s being learned from them. Open source should be a prerequisite for digital public goods, but let’s not be afraid to stay opinionated about what projects should receive donations. [3]

Governance power isn’t a perk—it’s a job with responsibilities. Cultural value is real (see memecoins and NFTs) but all the relevant tech is downstream of derivatives anyway.

Buybacks and dividends give tokens value. If a protocol performs a function that users will pay a fee to execute, some or all of that fee can benefit tokenholders. It’s confusing to me that this seems debatable.

There won’t be any mainstream adoption of this in its current form because people don’t want their coffeeshop to know their net worth after buying a latte. At the same time, the financial system is a critical tool for law enforcement (whether you like it or not).

It looks like some form of onchain KYC is going to be the short-term solution, but I’m not sure moving your financial freedom into Brian Armstrong’s hands from Jaime Dimon’s is what we’re here for. [4] The Tornado Cash case was interesting, but I’m not convinced society will accept money laundering while replacing the status quo. (Though perhaps it’s inevitable? I’d be surprised by every possible outcome here.)

For a while, I thought the creation of a truly meritocratic and free market for consumer financial apps would drive mass adoption. I was wrong. Being able to use any block explorer, any wallet application, any custom script, etc. is inarguably better than being at the mercy of my bank’s dashboard. But it’s not so much better that I’d expect people to move over all of their finances, even if privacy were allowed.

Also, how is there no wildly popular open source crypto tax software yet? This should be a quintessential public good.

Add a thin offchain layer to your protocol with functions including descriptions that an open-weights large language model can understand. If you can get an LLM to successfully interface with your protocol, LLMs will be able to generate GUIs on demand—if not just take care of users’ needs directly.

Wealthy people rely on other people to manage their finances by chatting with them, never touching GUIs. This will soon become available to everyone. I believe this is the 10x improvement for end users that will drive mainstream adoption. Everyone has the ability to do things on DeFi rails that they can’t with TradFi and AI agents will make this obvious.

Even with LLMs objectively assessing financial infrastructure options and managing money on behalf of users, the brands surrounding the technology will remain important. There are dimensions of identity and reputation that come with deciding how you handle your money. This isn’t going away.

At this point, I think “crypto” is unsalvageable as a brand, associated with crime and presidential memecoin pump-and-dump schemes. “Web3” is lame. (“Decentralized compute” is the best I’ve got.) The brand for the runner-up world computer is trash. [5] Ethereum is the real deal. It’s not a business. It’s Linux. It’s weird. It’s cypherpunk. Let’s keep it that way.

Ethereum is only ten years old. When the world wide web was this age, Pets.com was collapsing, and the idea that the internet could replace cable television was fringe. If this is really going to secure the world’s financial system, it’ll take a little while to get right. It starts with global access to Wall Street via centralized stablecoins. Replacing Wall Street is the end game.

The recent breakthroughs with large language models is serendipitous. They’ll cut through all the bullshit. This is bad news for traditional finance and crypto grifters alike.

So go put your energy into building financial infrastructure that’s appealing to my future superintelligent money manager. Thanks!

[1] You might think that you can progressively make a perps dex immutable, but this is a mirage. Either the contract holding the collateral is upgradeable or it isn’t. If integrated subsystems are immutable, that’s nice but you can just change them out on the contract that’s holding the funds.

[2] I would’ve guessed tokens would be the “primitive” that you’d build perps on, but it turns out it’s the other way around. Go figure.

[3] I say this with love. Development continues on Cannon (an open source dev tools project I contribute to which supports multiple teams building decentralized perps) largely thanks to the generosity of Optimism.

[4] I’ve seen “compliance-friendly” used as a buzzword here. This would’ve been parody just a few years ago.

[5] Yes, they retracted the ad. But was anyone that surprised they shipped it in the first place?