Nic’s Orb

Software
• I am excited about the prospect of synthetic data or self-play in order to further scale LLMs (see: https://www.dwarkeshpatel.com/p/will-scaling-work), as we appear to be hitting capacity constraints in terms of the supply of human generated data on the internet.
• The emergence of multi-modal is key for LLMs – we need a form factor that isn’t typing text into an interface, but rather communicating with models in a more realistic and intuitive manner.

AI Safety

• The emergence of models that reject the contemporary AI safety doctrine, like Erik Voorhees’s Venice (https://moneyandstate.com/blog/the-separation-of-mind-and-state), or Meta’s open source Llama 3, which appears to be very performant relative to close source models. Keeping models open source is critical in terms of allowing AI development to exist free from the monoculture in Silicon Valley. Zuckerberg should be praised for this (see: https://www.dwarkeshpatel.com/p/mark-zuckerberg) . Firms that maintain the full inference stack and refuse to censor AI outputs are equally important. Hopefully, the nexus of crypto and decentralized AI can also help here, although these projects seem to be less performant than conventional inference setups at present. I am still evaluating the crypto-AI space, haven’t found anything I find super compelling just yet.

Hardware

• The emergence of NVIDIA H200s which have far better performance than their predecessor, the A100 (https://nvidianews.nvidia.com/news/nvidia-supercharges-hopper-the-worlds-leading-ai-computing-platform). We are seeing inference speed getting far better recently, and I believe this is partly due to H200s
• Massive datacenter scale, such as Microsoft’s planned 100b datacenter (https://www.reuters.com/technology/microsoft-openai-planning-100-billion-data-center-project-information-reports-2024-03-29/). So far, datacenters have been pretty small (the largest are ~100MW). New versions will be in the gigawatt range, but this infra takes time to build.

Products

• I’m excited about AI wearables like Rewind or Tab (I am an angel here). Although recent products like the Rabbit or the Humane pin were panned by reviewers, I think AI wearables that ingest multimodal data from your everyday life will be extremely compelling, as they are a kind of “external storage for your brain”. You can query them about anything that you did or said throughout the day. I think these will be very compelling products.
• Not LLM based, but the two new products that stood out to me recently are Sora (text to video) https://openai.com/index/sora/ and Suno (music) https://suno.com/about. Both truly remarkable products, especially Suno

Note, from the investment perspective, I am much more interested in AI applications rather than infrastructure. I believe models will be largely commoditized (see this post: https://press.airstreet.com/p/alchemy-is-all-you-need) , whereas applications can create enormous value without huge upfront fixed costs.

I’ll start with general commentary around LRTs and then get to the individual protocols.

General LRT risks

Liquidity crisis – if a lot of users are rushing for the exits, similar to a stablecoin, LRTs might trade at a discount to par, even if their fair value is par. If users have a leveraged view on LRTs (i.e. with looping) they are exposed to these depegs. Liquidity for LRTs has not been established yet, so we don’t know exactly how they will fare especially in times of distress.
Duration risk – LRTs have different duration profiles. Some amount of LRTs can be sold instantly, although perhaps not at par. Redemptions might be gated by the protocol, meaning that extracting the full value of the LRT might take some time. Eigenlayer for instance offers a one week exit period. If the exit period is shortened, the risk of depegging is much less. This contrasts with the StETH/ETH depeg which was caused by the lack of exit entirely pre-Shapella. I don’t expect depegs to be meaningful with the relatively short exit period.
Default risk – node operators could be slashed causing LRTs to trade at a discount. Eigenlayer has the ability to veto a slashing event if it looks like it will be too disruptive.
Smart contract risk – conventional risks that all crypto infrastructure faces

Since we still have limited data on LRTs and Eigenlayer itself, I can’t give scores above a 7 for now. However we’ll look at a few protocols.

Ether.fi.
• Ether.fi raised 23m in a Series A from Bullish and CoinFund. T
• hey have $3.2b locked.
• The largest LRT provider, they are focused on native restaking, which in my view results in less intermediation. They also seem to be the most widely integrated with other protocols. Strong investors, largest treasury. 7/10

Kelpdao
• Kelp has $748m TVL.
• Reading the sigmaprime and code4rena reports are impressive and comprehensive
• Unclear about investors or fundraise
• Non native restaking, additional layers of risk
• Discount for the above, 5/10

Swell
• 250m in TVL
• Native restaking
• Raised 3.75m from strong investors Framework, IOSG, and Apollo
• 6.5/10

Renzo
• 2.06b in TVL, second-largest provider
• Mainly differentiated through its support for ETH L2s, and support for both ETH and LSTs
• Raised a 3.2m seed from Maven11 with participation from others
• They face three layers of risk (LRT, LST, and Eigenlayer) compared with Puffer Swell or Etherfi
• Discount on their score due to the above and investors 5/10

Puffer fi
• 3rd largest LRT provider with $1.3b locked
• They raised 5.5m from Lemniscap and Lightspeed.
• Impressive traction with relatively little capital raised, although in theory you would want a larger treasury
• Score: 6.5/10

For now, it’s hard to make a final assessment because we need more evidence for how these function in the wild under a variety of risk conditions, especially when slashing and depegs are concerned. Overall I remain skeptical of LRTs. I do favor native restaking (Puffer, swell, and etherfi) versus non-native (kelp, renzo), and my scores reflect that.

Case #1

If Powell prints $20T (or roughly 1x GDP, 1.3x M1 or roughly 1x M2), or about 4.4x the number of Treasuries held by the Fed right now, inflation would skyrocket, treasuries would sell off (massively), yields would presumably increase (as investors would start to refuse to hold treasuries which are paying out in inflated USD).

A few things would happen. USDT would face some liquidity problems because yields would fall reducing face value of treasuries. USDT holders would presumably exit to Bitcoin or something similar (unless USDT hiked yields to holders), because dollar inflation would be Argentina level, so USDT would have to process a lot of redemptions. But they would have duration issues because their bond portfolio would have sold off, so they would not have the cash immediately to make large redemptions. If they redemptions took place over 90 days, they would have sufficient funds to give investors back their (devalued) dollars. Note: this is the same thing that happened to SVB, but in Tether’s case, we’d be dealing with a bigger shock and so even 3-month treasuries would be affected.

Inflation would become extremely high and variable. The US government would have to engage in Yield Curve Control because the market rate of interest that holders of debt would demand would render the government immediately insolvent. Interest payments on the debt would climb to 30-50% of GDP. It’s complex to determine what inflation would be exactly, the problem is inflation expectations would become unanchored (because dollar holders would lose trust in the Fed), so it’s likely that inflation would be higher than 200% for a one year period. Let’s hope Case 1 doesn’t happen.

Case #2

In this case, half the money supply is backed by ETH (via Liquity), half is backed by treasuries. I would agree that Liquity holders would be largely immune from inflationary pressures, however the Liquity protocol still tracks the dollar, so it’s worth exploring whether Liquity would choose to devalue their UoA as dollar inflation skyrocketed, or simply move to a flatcoin/inflation indexed Unit of Account. My expectation is that Liquity if faced with a dollar devaluation, would move towards a different UoA. The inflation discussion is obscure again, because these things are nonlinear and can’t be derived like normal equations. Just look at how hard a time the Fed has prediction inflation. Even in our simplified model it’s hard to predict.

However, because the economy in this case would have a liquid alternative to Treasuries as a collateral asset, in this example, I think capital would flee to ETH/Liquity, making inflation worse for dollars stuck in the legacy system (by driving up velocity). Because ETH in this example would be big enough to absorb this capital inflow, this would leave dollar holders worse off inflation-wise.

Case #3

The difference between case 2 and 3 is the scalability of Liquity and Ethena. Even though Ethena is a more interesting protocol to me (because it’s based on a tokenized funding trade plus staking versus mere overcollat ETH), it’s less scalable, since the yield is based on the demand for long-side leverage on ETH. One Ethena grows “too much” it suppresses yields on ETH staking and pushes ETH funding rates negative. So Ethena is fundamentally less scalable as a stablecoin than Liquity (keep in mind we’re talking about tens of trillions of dollars here though). But it certainly true that if Ethena becomes a total juggernaut (and I’m talking 50b+ supply), it would be stacking on a ton of short-side liquidity, meaning funding rates would go negative. Of course, normally, this would causes USDe yields to fall, reducing demand for Ethena, but in your example, Ethena is super massive. So in conclusion, in the scale of trillions, Ethena would have a negative interest rate, since it would have to pay to be short (assuming Ethereum also isn’t in this case worth $10T+ with abundant demand for leverage or staking). So Ethena doesn’t look as attractive relative to Treasuries/USD in this case relative to Liquity. I think it would also be perceived as a less safe asset than an overcollat stablecoin, so I think less capital would flee the dollar system == lower inflation.

So we are investors in Ethena and I’ve known Guy for a while now. I did a podcast with him before we invested (https://onthebrink-podcast.com/ethena/) and one more recently (https://onthebrink-podcast.com/ethena-ii/). Additionally, on the matter of synthetic stables, I was an outspoken Luna critic, I diligenced Basis Cash (basecoin) for an investment back in 2017, and we’ve invested in various previous iterations of the ‘nakadollar’ on Bitcoin. So I have been thinking about synthetic stables forever. (You can see I reference them back in 2020 in my whitepaper cryptodollars: https://www.castleisland.vc/cryptodollars)

Previous experiments with synthetic USD failed because expectations were wrongly conditioned (i.e. founders wanted to hold the token at $1 at all costs, and were forced to abandon the system or take a loss when the basis went negative). Ethena is somewhat different in that the long side of the trade also pays a yield, via ETH staking (previous iterations focused on Bitcoin). StETH yields vary, but they are in the 4-6% range (although you might expect that Ethena’s entry into the market would further suppress that). So that does give you some margin of safety if funding goes negative.

The other thing that’s different about Ethena is that it’s envisioned as a crypto native interest rate that grows and shinks its supply in line with rates. So when funding rates fall, Ethena supply shinks (since it offers a less attractive yield), thus lifting the shorts (and unstaking stETH), both of which have the effect of pushing yields back up. Ethena is not trying to target a fixed yield, and so because they are willing to allow the yield to fluctuate, the system can be adaptive to market conditions.

Historically, the basis trade yielded 18% in 2021, -0.6% in 2022, and 7% in 2023 (and it’s tracking much higher in 2024). With stETH yields overlaid on the funding, Ethena would have paid a positive yield in all historical phases, even the bleakest portions of 2022. Historically (and this is a little questionable because Ethena’s presence changes yield dynamics, so you can’t just plug in historical data and get a reliable estimate), Ethena would have paid 21% in 2021, 5% in 2022, 13.3% in 2023, and right now it’s paying 33%.

The actual reason the yield exists is as follows. On the ETH side, there’s a yield because ETH is a vibrant place to transact with plenty of MEV opportunities. That won’t change. On the funding side, the yield exists because there’s generally in crypto a long-biased demand for leverage on futures and derivs exchanges, and not enough supply to accommodate this demand. This is in my view a market structure issue. If there was a single extremely trustworthy futures or perps exchange, and traders relied on a single large prime broker giving them credit lines at a few points above SOFR, I don’t think you’d see the same basis we see today. So in a sense, the persistent positive funding in crypto is due to counterparty risk assessments of these derivs exchanges. No one really wants to leave a ton of cash laying on one of these exchanges forever, and so the big lenders aren’t arbing away the basis (by going short futs and long spot), even if it’s paying 10 points+ more than SOFR. Crypto is a uniquely fragmented global market, with very disparate regulatory approaches, and a still significant “offshore” market for futs and perps, so I don’t see this changing any time soon. However, if things do become more formalized and come more onshore, then I expect the basis does over time collapse.

Ethena currently has an insurance fund (https://ethena-labs.gitbook.io/ethena-labs/solution-design/insurance-fund) , but it’s relatively small ($12m) relatively to the supply of USDe ($828m). (https://app.ethena.fi/dashboards/solvency) In theory, Ethena could ‘fill up’ the insurance fund in this high growth portion of the protocol and grow its capital ratio by skimming off a portion of the net interest margin. It’s yet to be determined how exactly the insurance fund would be utilized and in what conditions. The token issuance would presumably bolster the insurance fund, and Ethena itself can raise equity.

On negative funding and risks: as Ethena notes in their documents (https://ethena-labs.gitbook.io/ethena-labs/solution-overview/risks/funding-risk), they have only seen one quarter in the last three years where the stETH yield and the funding trade aggregated to a negative yield, and that was during the ETH PoW issuance period which polluted the data (by persuading a lot of people to go long spot and short futs, flipping funding).

Another risk worth considering is depegs in stETH derivatives that Ethena uses to collateralize its short positions. If stETH depegs (like it did in mid 2022), even though Ethena isn’t using more than 1x leverage, its collateral could decline in value, causing its positions to be liquidated. However, post Shapella, stETH doesn’t meaningfully depeg (because unstaking is possible).

Based on my investments in the space, I’ve long felt a ‘nakadollar’ was possible. The introduction of the positive carry on ETH with staking is the unlock that I felt made it meaningfully possible. That way holders could benefit from a somewhat uncorrelated portfolio of long stETH and long funding, with the stETH yields covering up shortfalls in funding. Long term, I expect yields to compress, both because more credit will make its way into futures and perps exchanges, as crypto prime brokerage gets more sophisticated, and because Ethena itself (and perhaps competitors) will grow, adding demand to stETH and funding markets, thus suppressing yield. However, there’s an equilibrium where USDe holders refuse to accept xx% yields above SOFR (let’s say it’s 500 bps), and so they simply stop creating units of USDe if yields aren’t sufficiently appetizing. This acts as a natural ceiling on the supply of the asset. And regarding the funding markets, I think there will always be a risk premium associated with leaving collateral on an exchange, and I think most traders will generally continue to be long-biased, so I do see a residual positive funding rate in the long term.

Bittensor defines a marketplace for digital commodities. It doesn’t contemplate a specific digital commodity in its design but aims to be general and create a protocol that is sufficiently broad to accommodate many different types of computational resources, all relying on the TAO token for incentivization.

It has 32 different subnets, covering different types of computational or virtual resources, ranging from text prompting, to translation, to audio, to vision, to transcription, and many more. (see: https://taostats.io/subnets/).

Like PoW, miners are paid to create computational resources. However, in Bitcoin, miners just have to create a valid nonce proving a sufficient amount of arbitrary work done. In Bittensor, miners produce responses to queries (typically machine learning or AI related), so the “work” done is theoretically useful.

Tokenomics: it has a fixed supply of 21m, with 6.3m issued. The FDV is $14b and the free float market cap is $4.2b. It’s traded on a variety of exchanges like KuCoin and Gate, so not the tier 1s so far. You can also buy wTAO on uniswap and other defi venues. It’s trading near its ATH. 88% of free float supply is staked or delegated, and only 11.3% is freely circulating. It follows a similar halving cycle to bitcoin. Miners need to acquire TAO in order to participate in the network.

Network revenue: neither Messari (https://messari.io/project/bittensor) nor Dune (https://dune.com/spencerswanson/intersection-of-web3-and-ai-decentralized-compute) , nor Tokenterminal (https://tokenterminal.com/terminal/projects/bittensor) appear to have metrics on actual revenue or utilization figures. The official explorer taostats does not have a dashboard on this either. It seems that these must be de minimis. This is a significant concern. In theory, the token valuation could come from demand to stake TAO in order to be a miner, but you’d want to see meaningful revenue accreting to miners to justify TAOs valuation on that basis. Unless I’m missing something (and I looked thoroughly) I don’t see any network-wide revenue. Joseph Jacks says it’s generating >$150m in revenue (https://twitter.com/JosephJacks_/status/1751076985286369495) but I think that’s false. I’m guessing he’s referring to staking rewards, which is really just redistribution between existing holders.

Conceptually: I understand why an open marketplace for AI inference would make sense. However, the challenge would be demonstrating why this is better than a bilateral relationship with an AI cloud provider like Core Weave, Lambda, or any of the hyperscalers or large AI companies. Even if they do end up participating in Bittensor, this seems like an additional layer of intermediation (versus just buying inference from them directly).

Conclusion: it seems like TAO is being valued based on AI hype and excitement, which is considerable based on NVIDIA performance. It is one of the more “mature” AI blockchain projects, and it seems to be one of the perceived winners in the category. It doesn’t seem like it’s being valued based on network revenue or any more “fundamental” financial metrics. Conceptually, I have a hard time justifying it. I don’t see why it is a significant improvement over the existing, highly efficient market for AI inference. Based on the extremely advanced valuation (and, in my opinion, AI hype reaching a topping stage), I am not particularly constructive on the token in the medium term. I would want to see more transparent financial metrics before I could devise a fuller opinion and certainly evidence of meaningful network usage. However, since the valuation is mainly hype driven as far as I can tell, it can run up much further in the near term, because no one is actually doing a DCF on this thing – just buying into the excitement and the general blockchain/AI thesis.

Summary:

Narrative: 10/10
Tokenomics: 7/10
Ability to generate hype: 9/10
Network revenue: 0/10
Evidence of usage: 0/10

Do with this information what you will.

Justin provides a sophisticated analysis of a possible 51% attack on bitcoin.
He assumes that some entity is able to obtain 51% of hashrate and lays out a few scenarios in terms of ways they would be able to mess with the network. To be sure, most of what he is saying is true. If some entity was able to marshal that much hashrate, they would be able to impair Bitcoin significantly, and he talks about breaking Lightning, breaking WBTC, not processing deposits from exchanges etc. They would also be able to mine empty blocks and just do a DoS on Bitcoin. The one main mitigation I would see would be Bitcoin becoming a de facto proof of stake system, whereby large stakeholders (i.e. exchanges, custodians, etc) agree together to mine on a distinct chain and only “bless” that separate chain, periodically reorging out the blocks mined by the adversary. However I don’t see this as a great solution, because it’s clearly very centralizing, and the attackers still has a ton of hashrate, and you can’t exactly pinpoint their ASICs and zero them out (without changing PoW, which doesn’t really achieve much). You could in theory move to a new PoW algo but I think that should be considered an absolute last resort, because it would nuke the businesses of all the miners (assuming it goes to GPUs), and they would keep the ASIC PoW chain going anyway.
He is right that the security ratio is declining with time, although the security budget isn’t necessarily decreasing with time, in dollar terms. But the ratio only really matters if you can short bitcoin, and make the attack worthwhile. And as I explain below, that’s not really the most obvious assumption.
There’s a few specific things I disagree with:

1. I think his budget for the attack is quite low. He says “on the order of 1-3b USD”. The state of the art research (done by my team at Coin Metrics), says it’s more like $6b at an absolute minimum to buy all the ASICs to do the attack (see: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4727999) . And obviously, there’s a market impact as you buy ASICs and theyre not actually available for sale, so you’d have to manufacture them, which means working with a foundry etc. And then there’s the cost to build or rent data centers etc. Just fundamentally speaking there’s huge logistical barriers to actually marshalling ~10GW of power (see: https://labs.coinmetrics.io/), getting the data centers, and getting the ASICs.
2. He says an attacker can make money by shorting bitcoin. But any liquid exchange will not honor transactions which are clearly market manipulation and paired with an attack. That’s how it works. Especially not the major most liquid exchanges like CME. Until there’s a ton of liquidity for decentralized Bitcoin DEXes, which isn’t the case today.

And it’s worth noting that bitcoin mining is getting more distributed, and more connected to energy infrastructure with time. The future of mining is basically energy utilities doing it alongside their infrastructure to monetize it, and they are all pro bitcoin (naturally). They’re not going to sell or rent their hardware or infra to an attacker, since it’s against their economic interests. A good fraction of miners are public companies too, and they are obviously highly incentivized to stay in business (which means keeping the bitcoin network intact), and they are transparent. They won’t sell or lease their equipment or energy resources to an attacker.
My other views on this have to do with the motives of an attacker. I don’t see a state level attacker wanting to attack bitcoin, because the rogue states generally benefit from the existence of Bitcoin mining (see Iran, Russia, NK, etc). So I don’t see obvious political will there. It’s very costly to attack bitcoin from a consensus approach, and there’s cheaper ways to “deal with Bitcoin” if you are a nation state that hates bitcoin – namely, you simply ban it, and ban the exchanges dealing with it. That’s cheap, and basically achieves the same thing (mostly getting rid of bitcoin usage within your borders). So I don’t see why an attacker would bother to try and undermine the entire network, when they can just solve their “bitcoin problem” super cheaply.
In terms of monitoring, the approach of monitoring mining pools already exists, pioneered by James Lovejoy’s work. Coin Metrics also extended this with their Farum product (monitoring work being done by mining pools to get advance notice of any oddities or mining on shadow chains or competing chains). I’m sure other analytics providers do this too. Basically you join a mining pool and monitor the block headers that they give you, and you compare them to the other pools. A big shadow chain being mined by an attacker would be evident. So the attacker would have to solo mine their blocks to avoid detection, which means they’d need a huge amount of hashrate which they would have had to amass secretly. The other thing you’d want to monitor would be the foundries selling space to a new ASIC manufacturer that didn’t want to sell ASICs but would use them for an attack, as well as looking at the secondary market of ASICs to see if there was a huge amount of activity there. Firms like Luxor and others monitor this.

The other thing is that mining companies use the latest ASICs, which are more efficient each generation, whereas the ASICs available on the secondary market have less “bang for their buck”, as in they produce fewer hashes, so you’d need a lot more of them to compete with the state of the art ASICs. So it would be monumentally difficult to compete with the pro-bitcoin miners. Of course, if there was a massive die-off of hashrate and most of the ASICs were furloughed, then you’d worry. What matters is the share of ASICs owned by honest miners versus “mercenary” ASICs which are available on the 2ndary market or furloughed.

I gave a talk at MIT Bitcoin Expo in 2019 about these ideas (https://www.youtube.com/watch?v=AyOyNF-bCkA) where I talked about the stock, threshold model, and the flow models of security. I used to believe in the stock model (the same on Justin talks about), basically believing that as the value of bitcoin grew, the “prize” to attack it would grow (aka the value that could be extracted by attacking it, with shorting eg), and so the security would have to be “proportional” to the value of Bitcoin. However, I now believe in the threshold model, because I think attacks can be defrayed relatively “cheaply” (i.e., exchanges can not honor short transactions made by an attacker), and I think the threshold of energy and hashrate required to mount an attack is so significant in real terms. So I think the real thing that protects bitcoin is the sheer real world resources required to attack it, and they are equally insurmountable at a $1t or $10t market cap (even if the security ratio is worse off). So I am also as a consequence less worried about fees (than I was in the presentation), because I am not requiring that security spend maintains a constant ratio versus market cap.

So I don’t think Justin is wildly wrong in his analysis, it’s just that it’s very theoretical and presumes an actor can amass 51% of hashrate, which I don’t think is very practical. Empirically speaking, I see great challenges to pulling this off. And I also have a hard time envisioning the motives of a sufficiently large hostile actor, when they can achieve policy objectives without attacking bitcoin consensus.

On the Points Guard post

The thrust of the Arthur Hayes post on Points is that Points work, Points are here to stay, Points are a more mature evolution of ICOs or Airdrops/yield farming, and (this is implied) Points don’t violate Howey in the US.

Some quotes:

 Points combine the best aspects of ICOs and yield farming.
 Are points a contract between the project and the user for a tangible reward in the future? No.
 Is there any form of money, fiat, crypto, or otherwise that is exchanged between the user and the project for points or tokens? No.
 Whether you like it or not, every successful project, and by success I mean, token number go up, will enact a points program before their TGE.
 If points create better alignment between users and the protocols, LFG.

I would say, referring back to my prior post, that Points technically are more useful from the protocol’s perspective than a mere frontloaded ICO, or the more sophisticated single-shot airdrop. They provide more flexibility, they give the ability to not issue if regulatory trends are perceived to be adverse, and they give developers the chance to continually tune the offering, deprive certain users/sybil farmers, etc. Technically, they are better.

However, you know that meme “I consent; I consent; is there someone you forgot to ask”? There’s more than two stakeholder groups here. There’s the team; there’s the VCs; there’s the retail punters; and most importantly there’s the regulators.

For VCs, points are useful, since they provide a cheap way for protocols to “buy” usage early on (and solve the cold start problem associated with protocols that need liquidity) without giving up either tokens or equity. Tokens solved this initially, but points are even better, since they’re just a vague promise of a token.

For the regulators however, Points don’t help. They are super arbitrary, and encode massive information asymmetries by definition. Aidrops are generally fairer, since the team (generally but not always) lays out the rules, rewards the community, and then that’s it. Points are more of an ongoing thing and the rules can be constantly changing. If we review Howey, Points don’t look great.

1. Expectation of profit – clearly present, otherwise no one would be farming
2. Investment of money – arguably present. Points farmers are stumping up capital for a time in the protocol. Capital has an (opportunity) cost, so this is an investment.
3. Efforts of third parties – especially for points programs that are drawn out, this is clearly present, as the retail farmers are depending on the team to build something
4. Common enterprise – yes.

So I would say points are no better than airdrops, and arguably worse, since info asymmetries are worse in my view. Here’s where I disagree with Arthur. Just because the promise of a future reward is vague, doesn’t mean it’s not a promise. It’s still compelling retail to do something. And in fact, from the securities regs perspective, if you’re making a promise which induces grandma to put her life savings in your defi thing, and then you DON’T release a token after, you’ve basically tricked her – arguably, this is fraud. So any team that is going to release points and wink and nod to the community in the hope that they commit their own capital and effort to make the protocol valuable, and then they don’t get tokens in the end – that’s probably the worst thing you could do. That’s an easy way to get sued. An implied contract that isn’t an explicit contract is still a contract, in the US.

Why do Arthur and me disagree? We have the same set of facts at our disposal. However, Arthur doesn’t live in the US, so he doesn’t necessarily have to care as much about the letter of securities laws. It's possible other jurisdictions (whose laws I’m not as familiar with) will take a more benign view of points. Assuredly, many airdrops and new listings will continue to use points, and I’m sure it will “work” much of the time. But in the US, using points doesn’t save you from securities laws considerations.

On Lex Node’s comments

He has a few comments on points. I’m not sure which ones specifically you want me to address. I’ll paste a few here.

 Points are probably the stupidest detour crypto has taken in my time here. Also borderline scammy.
 I just have zero faith that most of these points programs will turn into a token. Likely result: "Our lawyers advised us tokens are illegal. Enjoy your consumer utility attached to our closed-loop points system; you've earned it." Plan accordingly.
 overall, while there COULD be a good use of points as a more transparent, fair, consistent yield-farming structure as described in Hayes' article, it seems that's more theory than practice at this point--in practice, they represent a reversion to web2 and a new opportunity for team/VC abuses and info asymmetries

The key problem with points isn’t a points problem, it’s an alignment problem. Fundamentally, VCs have the advantage because they buy BOTH tokens and equity (sometimes in different ratios, so you have 1:1 ratios or 1:3 ratios, i.e. the FDV is 3x the post money of the equity deal, and VCs get a smaller share of the token supply), and they can be indifferent in terms of value accrual. The community that is ONLY buying tokens does not have that privilege. So if it turns out the equity accrues value (which should happen most of the time, because the equity is senior to the tokens (remember, the equity determines who runs what at the company, and only the shareholders can fire the core team, not the tokenholders)), as happened with Uniswap, for instance, the tokenholders lose. The team and the VCs are incentivzed to get a free token option by implying the token will be worth something, selling their own tokens, and then actually redirecting the cash flows from the protocol/business into the equity. It’s a BAD system. The only way to solve it is by having a single unified cap table where everyone has even to tokens and equity (as in, the equity cap table is mirrored by the token cap table). But this doesn’t happen, because the point of tokens is broad buyin, whereas equity cap tables are very small.

Points don’t solve this, as lex node says, they just introduce a third, most junior level of pseudo-equity. So you have a capital stack that looks like: preferred equity, common stock, tokens, points. Preferred equity gives the VCs the ability to set the board and fire the CEO etc, and it has preference in liquidation or a sale. There’s a lot of explicit rights that come with it. Common stock has fewer rights but still lots of codified legal protections. Tokens give you a firm right to whatever the token gives you. Points give you maybe a claim on future tokens. So you see how at each stage of the stack the rights get more diffuse and weaker.

The point of securities laws is to set retail investors on an even footing with the most sophisticated firms. This is why securiites laws are alllll about disclosure. If you go public, you have to reveal everything material about the company, and do it on a cadence and not insider trade on that info. Points are the opposite of this. They give you virtually no information, but youre still meant to buy in to the project anyway. So on the spectrum of Arthur Hayes to Lex Node I’m much more sympathetic to Lex. Although, I do think Points will remain the meta for a long time, at least until courts catch up with them.

Points are an airdrop precursor. They developed through a competitive ‘arms race’ dynamic between protocol founders and users of these protocols/airdrop farmers. They have recently become preferred to conventional airdrops because they give developers far more flexibility and discretion, and allow developers to defer an airdrop for as long as they need in order to sort out sybil issues as well as regulatory concerns.

Airdrops should be thought of as paid user acquisition. Instead of users being paid with cash, they are paid with pseudoequity in a protocol token. They purport to solve the ‘cold start’ problem whereby a protocol is only useful with liquidity. Thus, airdrops incentivize early usage of a protocol and get liquidity running, as users believe they will be compensated in the future for their past efforts. Aidrops basically pull future usage into the present by rewarding market participants in an anticipatory way and getting them to commit economic value to the protocol in anticipation of future rewards.

There are analogies to this in traditional finance. For instance, for certain exchanges, you have similar earned equity schemes where market makers can have the opportunity to earn equity in the exchange based on their proven market making volume. Airdrops are similar, except they are generally open to even more participants, hence the rise of the airdrop farmer.

The problem is of course that airdrops became an arms race between entities that attempted to sybil the protocol, and the protocol engineers, who want “honest” and non-mercenary liquidity. So it’s a never ending loop of sybil resistance mechanisms versus airdrop farmers trying to intuit what constitute “valid” liquidity. Of course the more intricate this arms race becomes, the more insider knowledge can be utilized, since knowledge of the anti-sybil mechanisms can be effectively used to gain an advantage. This means that airdrops frequently suffer from insider information. And because an airdrop is often a one time thing, developers are often compelled to honor the historical usage traits, even if an airdrop has been gamed. But this is rigid and the distribution is frozen in place after the airdrop occurs which means that developer often give out disproportionate rewards to sybillers, and they can’t do anything about it post airdrop even if they discover this historical behavior.

Another problem with airdrops is securities laws. In the US, it’s unclear as to whether an airdrop is sufficient to make a token “not a security” according to the SEC. Many airdrops are now ex-US. In my view, an airdrop alone doesn’t allay the concerns of a token being a security. Some developers want to take a wait and see approach and therefore don’t want to commit to an airdrop before we have more clarity in the US (since the US is still the largest market for crypto).

For all these reasons, the meta changed and developers moved towards airdrop precursors. They accomplish much the same – incentivizing early liquidity – while solving some of the above listed problems with airdrops. They are more mutable, so points can be arbitrarily altered throughout the points gathering process. They aren’t a firm commitment to an airdrop, more of a vague promise, so they don’t necessarily require an airdrop to occur (although users would be disillusioned once some of these points fail to convert to airdrops). And they allow developers to defer an airdrop until they feel comfortable with the anti-sybil dynamics and the security status of the token. Overall, they give developers a lot more flexibility and control, and dont force them to commit to token issuance models that might not make sense. They represent the power shifting back from airdrop farmers or sybilers, to devs, and potentially increase the share of ‘honest’ liquidity on a protocol.

I believe the first points system was Blur, or at least the first major one. Now they are virtually ubiquitous in defi especially. Anywhere where there’s a desire to get early liquidity and solve the cold start problem. Some good additional further reading can be found here: https://archetype.mirror.xyz/bbLSBkiAQY7gpSTzmvhlXgW93QSMx2goojhXrjOSWVg

Versus BTC

I would argue that it’s not really ETH’s fault that the ETHBTC and ETHSOL ratios had a difficult 2023. Both Bitcoin and Solana had major exogenous developments that propelled them over the year.

For Bitcoin I believe the main determinant of the rally from 15k to 45k+ was the ETF, naturally, at least the last leg from the 20s to 45k. The remainder was mean reversion after the credit crunch worked itself out and the bankruptcies were sorted, as well as the end of the hiking cycle and the re-injection of liquidity into the markets. A Bitcoin ETF will only come once and it’s the biggest development in the history of Bitcoin (in my opinion). The market is clearly anticipating structural long term flows that are unlocked by an ETF. These are indeed quite material and I think we could see $20b in net new flows into Bitcoin ETFs (not counting flows OUT of GBTC) this year. This will take some time as the big wirehouses, financial advisors, and brokerages need to do their diligence on the asset, see its volatility profile, validate client demand etc. This generally takes a while so don’t expect those flows immediately (although – we did have a very hot launch that was successful by virtually any metric). The ETF really does represent a meaningful legitimation of Bitcoin, it’s a massive victory over the SEC (perhaps the first), and it gets multiple trillion dollar asset managers on our side, some of whom had not been Bitcoin boosters historically. All of this is to say that this is a massive development for Bitcoin, so it doesn’t surprise me that ETHBTC sold off throughout the process of the market coming to grips with this reality.

ETH may get an ETF of its own later in the year, but I believe it will also require litigation (the SEC was very careful to specify that Bitcoin was the only cryptoasset they felt was a commodity). An ETH ETF is also more complex as it involves staking, and ETH is also just factually more centralized than Bitcoin (even if I personally think it’s “sufficiently decentralized” at this point). So the ETH ETF battle will be a long one. But ETH doesn’t need this as much in my opinion, because they’re not really chasing institutional acceptability or the “sovereign” buyer and they’re not really trying to be a SoV the way Bitcoin is (sorry Bankless guys).

Overall the two narratives are very different but I think ETH has a more active set of near term technical catalysts which people could get excited about. It will be hard to match Bitcoin’s post-ETF flows this year. To a certain extent it really depends on whether activity heats up in crypto and we see a big explosion of usage and tx fees. In that case, ETH will do very well.

Versus SOL

SOL recovered well this year because it shed the SBF affiliation/hangover and surprised to the upside: people thought it was basically a dead chain, and when it wasn’t, it had a nice recovery. SOL is also able to sell itself effectively against ETH because it has lower fees and it has a monolithic structure which means liquidity doesn’t get fragmented between a bunch of L2s. Even a big swimming pool can eventually be filled though, and I think SOL will one day reach physical capacity and will have to renege on its implied promise of low fees permanently. I also think there will be a dissilusionment with the monolithic approach, especially as the rollup-centric roadmap is executed on. This really is the fundamental question of blockchains though, which I don’t think people have the answer for – is the composability loss associated with the L2/rollup approach worth the tradeoff? This is one of the hardest questions I think that exists, but my intuition is that “biting the bullet” (aka pursuing rollups and L2s) is the right thing to do. Take your medicine, accept that it’s painful, but scale in a more sustainable way long term. This is the same tradeoff that Bitcoin made as well (instead of just naively increasing block size). I align with that, which is why I also prefer ETH’s design philosophy.

This isn’t to say I’m bearish SOL - in fact we have been pretty active there recently too - but I think the narrative is a little played out and ETH has been a little hated this year. I think the bigger threat to ETH is newer L1s that have a ton of venture backing and lots of excitement. I think it’s an easy bet all things considered for ETH to have an up year versus SOL.

General ETH commentary

I think the biggest problem ETH has is that it’s caught between two minds. Trying to provide a seamless, cheap and fast user experience, versus trying to create a cashflow machine and one that returns capital. These are at tension with each other. I wrote about this here: https://www.coindesk.com/policy/2021/09/17/ethereums-design-choices-are-inherently-political/

I wrote at the time that the imposition of EIP1559 (which I do find elegant) set the interests of users against those of tokenholders. High fees and lots of MEV = value returned to ETH tokenholders. But those of course are not “good” things for users. So there’s a tension there, post 1559. I wrote at the time that ETH competitors like Solana would do well relatively speaking, and we’ve certainly seen that play out. The other thing is the monolithic approach which is certainly more convenient from a DeFi user perspective than an ETH user that is constantly shuttling between various different rollups and L2s.

I think ETH is at a time of transition, where leadership is asking themselves whether they want to cultivate an efficient low fee experience, to combat the pressure from Solana and others,

The other big problem is cannibalization of ETH value from L2s. They all have their native tokens, and I believe these are siphoning significant amounts of value that would otherwise be resting in ETH itself. The ETH stake yield isn’t sufficiently attractive to move the needle here, especially not relative to an L2 token that could rally by 100%s of percent in relatively short order. It’s a weird feature of the ETH ecosystem that everything has a token (unnecessarily so in my opinion) and so this creates an idea of mercenary capital that’s constantly chasing the next thing, whether it’s an ETH L2 or a new L1 like Celestia or Monad etc. The other problem is that while L2s on ETH do make the transactional experience better, they are kind of parasitic in that their tokens redirect flows that might have gone into ETH, and they don’t pay super high fees (and this is also going to be reduced with EIP 4844), so they don’t burn a lot of ETH.

All of that said, if we do re-enter a crypto bull market, the fees and MEV will juice yield and get capital flows to enter ETH, and ETH is still a prime form of collateral on DeFi (even if stablecoins are taking over). The emergence of ETH-backed stables like Ethena is interesting though, and a clever way to use USD while holding ETH.

Conclusion

I do expect ETH to do well against both BTC and SOL this year. I think Bitcoin has lots of tailwinds from the ETF, but I think ETHBTC and ETHSOL can both end positive this year. The key is for the market to realize that ETH has a distinct and realizable roadmap, and that it’s worth the short term pain of the non-monolithic roadmap in order to eventually achieve Vitalik’s rollup-centric vision (which I think ETH is executing on). From my perspective, ETH is the only L1 that has really found a way to charge for usage and return that capital to tokenholders. While they are having a bit of a midlife crisis right now in terms of whether to lean into capital return or to lower fees overall to compete (and it seems like they’re going in the low fees direction now) they have been executing well on their roadmap for longer than anyone and remain the dominant smart contract L1 and de facto center of the universe.

Unfortunately, these matters are very important for me and I have dealt with a number of security threats in my life. A committed adversary can always defeat a strong security setup, but you can make things very difficult for them.

A lot of this info I gleaned from this guide by Jamison Lopp although he goes really far. https://blog.lopp.net/modest-privacy-protection-proposal/. The rest of it I got from friends that go through their own challenges and from my own experiences.

Physical security

Firstly, I chose to live in a place where self-defense is legal, and the police and courts are tilted in favor of law-abiding defenders, as opposed to attackers. This isn’t the case in most of the US. Not to get all political, but this rules out all blue states (in my opinion). There, even if you validly defend yourself, you are liable to face prosecution whether criminal or civil. Florida has excellent self-defense rules and of course firearm ownership laws. That is a huge part of the reason I live here. An armed society is a polite society. This extends to conceal carry. Obviously, this doesn’t really apply to non-Americans, but it bears mentioning. If you do own firearms you have to train with them (in real world situations), it’s not enough to just have them in your house or car.

Sadly, the place in which you live is like 70% of the battle. If you live in a functional country with strong property rights that affirms the right to self-defense, you are in a good spot. As a resident of Florida I feel like I made the best possible choice. But I had to go through a lot to actually end up here. Ultimately it’s your life and there’s no substitute for feeling safe and comfortable. Miami is actually more dangerous than NYC but it really depends on the neighborhood.

Next, I live in a high-security concierge building. Obviously not everyone can do this, but this is a great start. No one gets in without being on camera.

Third, I don’t recommend owning property directly. If you do buy property, in the US, your address can be found trivially through real estate records, so I recommend buying through a trust or an LLC. That’s essential, because property is a long-term commitment and you don’t want to be an easy target. And of course, you don’t want to get SWATted, which is a real threat these days.

Also, kind of a silly one, but being in good physical shape and knowing the basics of combat sports is important (if you’re a man). I train boxing when I have free time and I know that I can hold my own in a physical confrontation with someone my size. Good cardio is also important. You can always run away! I believe that just by presenting as someone that’s in shape and more physically imposing you’re much less likely to get into a confrontation in the first place. You just don’t look like a victim.

Use a PO box for orders and deliveries. This is cumbersome and annoying, but necessary IMO.

Digital security

I use a yubikey and physical two factor authentication for everything where it’s possible. I use two factor everywhere. I use a cell provider that is SIM-swap resistant. SIM swapping is basically a guarantee so this is one of the most important things. Using an authenticator is also essential. I am maxed out with my security on all exchange accounts etc. And with everything else like dropbox and email.

In terms of storing crypto, use a multisig provider with multiple different hardware devices like Casa for long-term cold storage. For short term crypto holdings, a hardware wallet with a browser wallet is necessary. I think if you do these things you are basically good. The problem is when you get sloppy and just use a browser wallet with no backup or no hardware device to sign.

A standard antivirus stack like malwarebytes and various browser extensions that disable ads and javascript etc is also necessary. Being smart about what links you click in emails. Assuming sketchy emails are phishing and confirm with people if they send you a dropbox link that you’re not expecting.

Best practices

Never post publicly about where you are at a certain moment. I am not that good about this. Holidays especially. You don’t want people to know when you’re traveling or away.

Never, ever, ever, ever post about family, significant others, or other people close to you. As a public person, I live under the eye of sauron and I’m used to the normal volume of threats, attacks, etc, but regular people aren’t, and they didn’t sign up for that. The few times I’ve posted about family, I have always regretted it, as they subsequently became targets. I have never to my knowledge posted about a significant other, and I don’t think I ever will.

When you’re traveling and have a driver, use a fake name. That way, gangs can’t canvass your driver and realize it’s you, replace your driver, and abduct you and hold you to ransom.

Don’t have your real name connected to your airpods or your laptop. Use a fake name. That way no one can identify you from your digital footprint on bluetooth or airdrop.

When traveling I wear sunglasses and a hat so people don’t recognize me as easily. It’s weird having people come up to me in the grocery store. Flattering, but also scary every time.

Don’t wear crypto merch in public. I used to be really bad about this but I’m better now. I almost never wear identifying stuff. Obviously it depends where you are and live. In some places it’s normal to run into tech and crypto people. It doesn’t worry me as much in affluent places. But when traveling or in a place I don’t know well I never wear any kind of merch that identifies me as a crypto person.

You also don’t have to tell random uber drivers what you do for a living. Probably best not to ever tell strangers what you do.

Don’t do silly stuff like go to the club and brag about how much crypto you have or put yourself in a compromising situation especially in a foreign place. This is just general good advice. But you are much more likely to be robbed when you’re drink in a foreign club or bar naturally.

Obviously, don’t share info regarding your net worth or holdings with anyone. Maybe just your family or SO. No one needs to know this info. It’s a liability.

I think my opsec is like, 99% percentile, but unfortunately this kind of stuff is necessary for a crypto person who is fairly well known. I view all this stuff as basically the cost of making a living in this industry. Crypto is lucrative, but it’s a very costly industry to operate in in many ways. The opsec risks and constant paranoia are the “cost” of operating in crypto.

The Solana phone I don’t worry about too much, because it looks like a normal phone.

Since the question is posed ambiguously, let me interpret it two ways and respond to each.

First formulation

The first is the straightforward question of labor; that is, what happens if the human workforce is rendered obsolete by AI models that vastly outperform them within a decade. How do you react?

So I basically believe this will happen, albeit not on the timescale proposed. Already, relatively primitive LLMs have already completely changed the game (as in, undercut the equivalent of human labor by 1/100th of the cost, or performing equivalent tasks in 1/100th the time) in a few domains. Off the top of my head:

• Translation
• Transcription
• Stock photo generation
• Graphic design
• Data cleaning/preparation/manipulation
• Essay composition
• Programming
• Copywriting
• Vehicle operation
• Summarization
• Legal advice
• Radiology/imaging

I’ve used AI for most of these use cases and it continues to utterly shock me in terms of how much human labor it eliminates. And this is with only a year or two of experimentation with newer models. Eventually, as we lose regulatory shackles, this will trivially extend to other white collar professions like accounting, law, tax, and medicine. At which point the AI will be incredibly disruptive and will be highly resisted at the government level.

There’s one school of thought that the AI won’t destroy that many jobs, because technological innovations have always created new jobs where they’ve made others obsolete. But this is to misunderstand what AI does. Prior technological innovations were generally narrow, as in, they helped people harness new forms of energy or communicate more efficiently. AI isn’t a narrow innovation. It’s broad. It touches on everyone aspect of human behavior and work, because it involves general cognitive tasks. There’s nothing that AI doesn’t pertain to. There’s only a few things that the steam engine or the mechanical loom help with. AI is a superseding technological development because it provides anyone a cognitive overlay on top of the world that can be applied to any form of knowledge work (and a lot of analog tasks too).

There’s another point to be made here. The industrial revolution was the most analogous situation because it allowed us to harness brand new forms of energy like coal, natural gas, and oil, allowing us to develop modern mechanized industry and basically, civilization. Before that we were effectively neolithic still. But no one was put out of work by the industrial revolution right? Wrong. It was the horses and other forms of animal labor that were “put out of work”. Except they weren’t around to complain about it. They were just slaughtered. This time, it’s the accountants, doctors, and lawyers that are the “horses” of the latter day industrial revolution. Once an AI model reaches parity, there’s no reason to use a human for the task that’s thousands of times slower and more expensive. Of course, these professions will resist and will try to stop the flow of progress. But logically people that want to consume cheap, AI-based medical or legal services will simply choose to do it in the jurisdictions that are open to the idea. So the market will route around these kinds of barriers. And the outcome will be the same. No one will want the inferior human version.

(We don’t have to consider the risk that AI models stop improving, or fail to reach human parity in the key fields I mentioned, because the question presumes that they will).

So how do you approach life? Very simple. What this AI revolution will do is massively empower capital at the expense of labor. And not just working class / blue collar labor, but the professional classes (in particular). Labor is the biggest cost line item in most corporates today. If you eliminate that, you massively increase profitability. So I think you get a massive productivity boom while simultaneously thinning the workforce significantly. AI has made it possible for the solopreneur to emerge, a single entrepreneur building a business reaching over 1m ARR with no employees at all. This is a massive benefit to founders and the entities that fund them, and generally very bad for everyone else.

So my strategy is to do what I’m doing – invest aggressively in AI at each layer of the stack (compute/datacenter, hardware, and applications), both directly, and by LPing into VC funds. Under no circumstances would I want to be on the other side of the equation – selling my labor on a linear basis, even if it’s consulting or lawyer type billable hours. Or even working as a tech employee, because headcount requirements are simply going to be reduced as AI models improve. Already in my day to day life I’ve made fewer hires than I otherwise would have because of AI.

So my advice is to try to own as much equity in AI-benefiting businesses as possible (frankly, they don’t even have to be AI companies per se, since in theory virtually all capital will benefit from AI verus labor), short of starting one directly.

In terms of picking a field that’s “AI proof”, I think VC is actually one of the last to go, because it’s highly personality based and not systematic at all, and it’s more about access than it is merely evaluating deals. But I don’t have that many good ideas, because I’m pretty pessimistic on the ability of any field – whether professional, manual labor, or artistic – actually avoid the ravages of AI over the next decade or so.

Section formulation

The second formulation of this question, is that the singularity is achieved within 10 years. By this I mean we reach a stage in which AI becomes recursively self-improving, and achieves a state of superintelligence, vastly eclipsing the sum of all human intelligence. At this point humans begin to see biological life as futile perhaps, and upload their consciousness to merge with the AI in some sort of AI rapture. (The latter doesn’t follow the former but you can grant me some creative license). This is the kind of “positive transhumanism” believed by folks like Grimes. How do you live, knowing that this will be the case?

First of all, I’d point out that we don’t have a ton of market signal at least that this is all that likely. For instance, if you read this paper (https://basilhalperin.com/papers/agi_emh.pdf), they point out that if either aligned or unaligned AGI were imminent (on the order of decades), that would be reflected in high real interest rates. Which is not the case. So at least, the most liquid markets on earth are not at all pricing in the probability of the emergence of hyper potent AI within a decade.

But let’s assume it happens anyway. Let’s break it down into unaligned AGI and aligned AGI. In the unaligned case, the AI needs our atoms for something else, and simply vaporizes us/turns us into goo. Let’s say its malicious turn is relatively abrupt too (based on the question, we know that AGI is coming, but we don’t know if it’s aligned or not), and we don’t have plenty of forewarning. In this case, the best choice would simply be to live life as normal. Presumably, any approach you could take to the imminent extinction of humanity aside from following the ordinary course of business would be -EV. In the aligned case, I would take a similar attitude, although I would prepare by keeping my health and mind as intact as possible ahead of the sublimation and merging with the AI. So that would require being extremely health conscious, very cautious in terms of risk taking, etc, because a possible biological-machine hybrid eternal life waits at the end of the decade.

Note that this is not investment advice but a thought experiment about portfolio construction.

Assuming that compute, in particular, AI compute, will be very valuable in the future, I would include the following categories in an “evergreen” portfolio (I am not limiting myself to public markets, but including also large or late stage private companies where acquiring a stake might be possible on the secondary market). For our evergreen portfolio, we will allocate equally to each area of the stack, from components in the chip supply chain, all the way to AI model developers. For the sake of this thought experiment, I’m going to assume that obtaining secondary equity in private names is possible (or that they may list in the coming years, like ARM).

The five tranches, and their suggested weightings within the portfolio are as follows:

• Extended chip supply chain (10%)
• Chip manufacturers (20%)
• Datacenters (10%)
• AI clouds (35%)
• Foundation models (25%)

I’m focusing mainly on AI clouds, as I believe they are the ultimate picks and shovels play for AI, and there are a lot of pure play AI names in the sector. Foundation models are important, but we haven’t encountered clear winners in the space yet. Certain chip manufacturers also have meaningful AI exposure.

Extended chip supply chain:

The standout here is TSMC. They are exposed to geopolitical tensions, but they are the standout in high quality chip manufacturing, with 57% market share in semis, followed by Samsung, UMC, Global Foundries, and SMIC.

TSMC’s critical supplier is ASML, the Dutch company that provide the photolithographs that TSMC uses in their foundries. You would also look at Applied Materials (manufacturing for chip fabs), Lam Research (wafer), Tokyo Electron Limited (assorted semis equipment), KLA, BASF, DuPont, and Merck KGaA (chemical supplies)

Based on AI as a share of their business, I would allocate within this tranche of to TSMC (50%), UMD (20%), Global Foundries (20%), ASML (10%).

Chip manufacturers:

The standout of course is NVIDIA. They dominate GPUs for modern clouds and for contemporary models. These chips, like the H100, or the forthcoming H200, tend to have a relatively long depreciation period because if you train a model on a specific chip, you will want to do inference on that same chip. The other competitors here are ARM, Intel, AMD, Qualcomm, and the hyperscalers like Apple, Samsung, Google, Microsoft, and Huawei which are all building their own chips.

Because the hyperscalers are trading richly already, and chips are a relatively small part of their business, we will focus less on them here. Within this tranche, I would allocate to NVIDIA (50%), AMD (20%), Intel (20%), and ARM (10%).

Datacenters:

We are interested in brick and mortar datacenter operator that are able to build tier 4 datacenters. This is a vital part of the AI supply chain and likely the bottleneck in the coming 24 months. The most important ones are Equinix, IBM, NTT Comms, Switch, KDDI corp, Sungard, Global Switch, and T Systems (dutch telekom). Here I allocate 50% to Equinix, 30% to Switch, and 20% to NTT comms, based on their respective AI focus and geographic exposure.

AI clouds:

This is one of the most interesting tranches. We have the megaclouds run by Microsoft, Oracle, Amazon, and Google, followed by boutique names like Core Weave, Lambda, Paperspace, Cirrascale, and Vultr. Additionally you have crypto names like Akash and Render.

Microsoft gets 20% of the tranche since they are so AI focused and have OpenAI exposure. Core Weave (disclosure: I am an investor from seed) is the standout in the boutique name, so they get 40%. 5% apiece to Lambda, Paperspace, Cirrascale, and Vultr (acquiring stakes in these names might be harder), and 10% across the major crypto render tokens as a hedge, in case decentralized rendering takes off. Overall, I believe this category will create the most value in a long term AI portfolio.

Foundation models:

This one is harder to predict. Major players like Google have yet to meaningfully enter the game. Meta is open sourcing their models. There’s still a huge amount of churn in foundation models and it’s very unclear who the winners will be. But nevertheless we need exposure. The big names here are OpenAI, Google (Deepmind), Anthropic, Huggingface, Midjourney, Meta, Stability AI, Microsoft, Inflection, and Cohere. Since we already have Microsoft exposure, and Google and Meta aren’t really AI companies, I would focus on OpenAI (30%), Anthropic (20%), Huggingface (10%), Stability (10%), Inflection (10%), Stability AI (10%) and Cohere (10%).

Exchange risk

I think Binance clients can generally feel pretty comfortable with the custodial risk of the exchange. They have completed 12 consecutive PoRs, and recently moved to a ZK model in terms of proving liabilities. Their most recent PoR had 33 assets, by no means all the assets on the platform, but by my estimate this covers 95%+ of the platform assets, so I’m happy with the coverage there. They do publish a list of addresses but as far as I can tell (and I would welcome more clarity on this), they aren’t actually signing those addresses, so we don’t have cryptographic verification that they own the addresses. I don’t have a huge reason to doubt the veracity of their claim regarding on-chain assets held, but they could go further here. I feel pretty comfortable in saying that they actually do have the $65b+ in assets that they claim to have.

In terms of actually verifying the liabilities, they claim to have a third party auditor, but they don’t actually share the name of the firm or make the report available, so I’m discounting that entirely. We still don’t have strong confidence that they are faithfully reporting (i.e., not understating) liabilities. They need to improve here.

As far as legal and contractual protections go, it’s not clear from their ToS that client assets are held in FBO trust accounts (i.e., bankruptcy remote), and it’s also not clear what jurisdiction and legal system they are based on. These are vital pieces that are necessary for a fully specced out PoR in my view.

So the PoR is useful, but it could be much, much better. Specifically transparency around the liability reporting, on-chain verification of assets, incorporating a third party auditor (more obviously), and creating legal and contractual clarity are necessary for me to trust it.

However, returning to a more basic heuristic, Binance has historically been solid from a user custody perspective. They don’t have a track record of mishandling client funds. They have a strong new CEO with a regulatory background who has a very strong incentive to do things properly, and they will be supervised by the DoJ as part of their settlement.

Also, it’s worth noting that institutional Binance clients can trade on the exchange with third party custody using something like MirrorX (https://www.binance.com/en/blog/vip/keep-your-institutional-assets-in-thirdparty-custody-when-trading-on-binance-with-mirrorx-7578047330993855911) or one of the emerging prime brokers. So larger clients don’t necessarily need to trust them. I’d want to see more integrations with the likes of Copper’s clearloop for instance though.

Lastly, if they had been misappropriating funds, I believe the DoJ would have included that in their indictment.

The Future of Binance

I believe that Binance will gradually lose market share, as they have been doing (they’re at 44% of the ex-US exchanges according to the Block (https://www.theblock.co/data/crypto-markets/spot/no-usd-support-exchange-volume-market-share) versus 67% a year ago), but perhaps not as quickly as some expect (i.e. after Bitmex’ settlement they fell off quickly). The reason for this, is that post FTX and Binance, newer exchanges will have a reduced motive to quickly gain market share by starting with no or light KYC. Everyone knows now that if they have weak KYC, the US will find a way to get to them. That was historically the way that newer exchanges gained traction, but executives understand now that it’s not really workable. Binance has some very sticky qualities, which is 160m users onboarded on the platform, tons (100+) of relationships with market makers, and deep spot liquidity even for the long tail. So it will be very hard for other exchanges to create the same value proposition quickly. That said, I expect that OKX, Derebit, and Bybit likely gain market share at Binance’s expense.

It’s worth touching on the SEC indictment. It looks like they’re looking for a gotcha (https://cointelegraph.com/news/sec-looking-evidence-potential-fraud-binance-us-changpeng-zhao) with the Binance US entity. I think it’s unlikely they find a smoking gun there, and Binance US is virtually irrelevant today. And as we know, the SEC has a patchy record when it actually comes to winning cases in crypto, so I’m not that concerned about it. This seems more like a fishing expedition than something that could threaten the broader Binance apparatus.

The Broader Crypto Ecosystem

I firmly believe that the DoJ action against Binance was the “end of the beginning” of the era of rollicking offshore crypto exchanges. The end started with the Bitmex criminal case, and the end ‘ended’ with Binance. As I said, newer exchanges can’t compete any more by going no-KYC. Newer exchanges will focus on regulatory compliance (whether in a light-touch offshore jurisdiction, or by having full KYC from inception). This is an absolute necessity to operate today. Especially if CZ gets an 18 month prison sentence. No one wants to risk that. I’m sure CZ would trade most of his wealth for the chance to be totally in the clear here.

What this means is that the institutional flow of capital in the US and major developed markets has a much easier time getting comfortable with the industry. The ETF was held up partially because the SEC hated the fact that the BTC spot markets were 60%+ dominated by Binance, a largely unaccountable exchange. Onshore exchanges felt hard done by that they had to compete with Binance and FTX, and it’s my view that the onshore exchanges felt that they had to engage in riskier listing and product strategies to compete.

Now that the threat from offshore exchanges has been mitigated, onshore exchanges can behave a bit more responsibly without feeling that they are losing a ton of market share. This also exposes them to less liability. So you will see fewer cases against the likes of Coinbase and Kraken going forward in my view.

I think we will also continue to see a move towards the segregation of exchange and custody, which has already been happening. Larger firms are very attuned to counterparty risk. This accelerates that.

And optics-wise, this gives people the impression that the bad actors and criminality are being cleaned up in the industry, which I believe is an unlock for sidelined capital to come in. I have a more moderate view on Binance, but I can see why some would see this as a strong positive.

I think it’s also very good for Proof of Reserve. Binance’s PoR, as weak as it was, helped eliminate any kind of bank run following the DoJ indictment. I think that’s extremely positive, and will be seen by many other exchanges as an extremely nice feature, turning a PoR from a cost center to something with a clear benefit.

Tether I’m less worried about. They are verging on being a systemically large holder of US treasuries, they work actively with law enforcement, and they are very active in freezing suspected illicit accounts. They also settled with SDNY a while back and were supervised for a while. They also have known US counterparties as part of their structure (Cantor). So I am not that concerned about a future issue with USDT.

So interestingly I was on a livestream with Richard Heart in summer 2017 when he teased HEX – back when people thought of him as a straightforward Bitcoiner – and I thought to myself “oh, that’s a shame, RH is launching an altcoin, I think he’s smart so it’s silly he would do that”. Well what I should have done was buy HEX instead of dismissing it, lol.

That said I’ve never owned HEX or dedicated a lot of thought to it – I outsourced all my understanding to Eric Wall who has always summarized the latest goings-on to me.

On to your question. I read the SEC complaint and I’m somewhat familiar. HEX is very much unlike Ripple, because Ripple has/had:

• A seasoned team of executives with tons of experience
• A huge legal budget and a massive headstart from selling billions of USD worth of XRP for about a decade
• A ton of credibility from all the partnerships they had struck with various financial institutions
• a meaningful “product” with real traction (even if I think it’s a stupid one)
• motivated backers and VCs who will go to bat for the firm. these are entities that have a ton in the idea of Ripple and won’t roll over without a fight. so they are highly motivated
• a decade of operational history and a bunch of momentum that goes with that- and a lot of serious stakeholders

Not to sound harsh, but HEX doesn’t have that. HEX is really just… Richard Heart. And he only has so much bandwidth and resources. And while I think XRP really was a security, there’s enough wiggle room in the case to see how the SEC might run out of steam and give up. But I do think the case will actually be fatal to Ripple in the end, just a pyrrhic victory for the SEC, since it will have been so insanely costly and such a huge drag on their resources and time. And they lost a bunch of intermediate court decisions along the way, which reduced their political capital.

By contrast, the SEC’s case against HEX/Pulsechain is a slam dunk, I’m sorry to say. I read the complaint and it clearly satisfies all the prongs of Howey. Additionally, RH marketed HEX to ordinary folks (Ripple did too, but RH was more working-class focused), and the price action is dismal, so the SEC is going to very easily be able to take the moral high ground and claim a huge number of ordinary folks were injured by RH.

While the SEC’s case is not criminal, when you see cases like this, it normally entails a parallel prosecution from the DOJ / law enforcement. This is such an open and shut case, and RH /HEX had such a US presence, that you will very likely see this. Even without a criminal case, the SEC is looking for disgorgement and fines, which means RH will have to cough up everything he made from HEX/Pulsechain and then more for fines. So he wont have capital left over to continue the project

On the matter of service. The SEC invested a lot of resources in the case, so they won’t give up easily, even if RH can dodge service for a time. Speaking from personal experience, a judge will let you extend service indefinitely if you ask them, and it’s clear the individual is dodging. What will actually happen is the SEC (or the DoJ, when the time comes), will hire investigators to find out where RH is and make sure is ultimately served. He may head to a non-extradition country, but the SEC case won’t just go away, so I think it will be fatal for the project (since no real exchange will deal with HEX in this state, no one will work for the company, etc).

I think HEXicans are emotionally invested in this (not the first time I’ve seen this in crypto, obviously I have been outspoken about Bitcoiners not being able to see any negatives with Bitcoin either), and so they refuse to admit that the bad news is really bad news. But it really is in this case. These kind of fraud cases are layups for law enforcement, and I don’t see any real ambiguity here. If the SEC was willing to destroy LBRY (a pretty solid, functional product that delivered a real service), you can bet they will show no mercy to HEX.

If litigation does develop, that will tie up all of RHs time and resources, and he won’t have any time to promote the protocol or develop the product. Based on founders who I know who have dealt with this, I can say this for sure. You can operate relatively unencumbered normally, but when you are dealing with discovery, and litigation, and the threat of massive penalties and fines, let alone possible criminal outcomes, you won’t have any ability to run the business as you ordinarily might. From my perspective, it looks like it’s curtains for HEX.

The most probably outcome, by far, in my estimate, is actually jail time for RH (after perhaps, an extended amount of time on the run/living in a non-extradition country). I give that a 70% chance. I don’t say this to be cruel or harsh, this is just my honest assessment. The 2nd most likely outcome, is a huge settlement which terminates HEX and Pulsechain, and causes the project to be wound down. I give that a 25% chance (keep in mind the first outcome is also fatal for HEX). By far the most unlikely outcome is that nothing happens, and the SEC forgets about their case. I would give that about a 5% chance.

First of all, thanks for being a great Orb holder (now that it has sold). I’m going to answer anyway though. I had a lot of fun with the questions and they were highly stimulating and relevant to my interests.

I think one of the most interesting possible topics for me is simply the topic of writing. (I have a few things I’d say I have “expertise” in, but we’ve touched on them recently in prior Orb answers). One question I get a lot is “what is your writing process” / “how do you write”? So I’ll share a bit more about how I do it. I’ve written around a million words publicly in blog posts, journal articles, and op-eds since 2017. Some of them I’m very proud of.

For more about my personal history and why I like to write so much, see this article: https://medium.com/@nic__carter/on-writing-8d90cd63c2e0. The main answer is “because I love the artistic process of writing”. I find a delight in words, etymology, syntax, and turns of phrase. The second answer is “because I feel a sense of urgency about taking ideas out of my own brain and conveying them into the brains of others. Almost all of my writing projects come about because I have an idea that I feel incredibly restless about, and feel a burning need to impart to others. And that’s why I write. Writing is painful, and it’s only worth it if you think there’s a lot of upside in putting ideas in brains, at scale.

I’ll add more on my specific process though. The first thing to know is when not to write. I don’t write when I’m not feeling inspired. This is why a column can sometimes be negative in terms of a writing career. There’s no quicker way to lose interest in your art if you feel that it’s your job. So for me, I rarely accept compensation for writing things, or in the case that I do, it’s not the main reason why I write. I feel lucky that writing can be a hobby for me, and it doesn’t have to be my career. (Even though my actual career involves a ton of writing).

So my first piece of advice is wait to strike until you have something you feel you really need to say. Writing for the sake of it is the equivalent of “junk reps” at the gym. Obviously, you need to be skilled at the art of writing (just cause everyone can write words in order doesn’t mean they’re all equally good), so you need a lot of practice in the first place. As a kid, all I would do was write short stories, and read voraciously. Don’t expect to become a good writer overnight. It’s like anything else, you have to work at it. But assuming you’re a decent writer, you will need to be judicious in terms of when you choose to write. If you don’t care deeply about the topic, you won’t me motivated to look deeply into sources, do the right research, talk to the right people, get feedback, and revise the story until it’s perfect. So my best tip for writing is don’t write when it’s not the right moment. When you do write, you want it to matter.

The way it works for me is I marinade on an idea for a long time. Sometimes days, sometimes weeks, and sometimes even years. I put my brain to work subconsciously developing an idea and it works on it in the back of my head over time.

Then, gradually, or suddenly, an impression comes to me. It might be a visual impression. It might be a fragment of text. Normally it’s when I’m doing something active, like running or walking. Charles Darwin went for a walk around his garden every day to gain inspiration. Physiologically, you’re more likely to be creative if you’re doing moderate exercise, for whatever reason. One of my best-known articles came to me as a sentence fragment – “bitcoin is a most peaceful revolution”. I thought about it for a bit, rushed home, and wrote the entire thing in about 30 minutes. This is a best case scenario of course. It’s not always like this.

But generally it’s a similar pattern. I find something that I think others are wrong about, or find out some key insight or piece of data I desperately want to share. But I don’t know the structure yet. So I think about it, try to clarify my thoughts (the first draft in my head is always hopelessly muddled), and when I finally feel confident, I sit down (normally late at night or early in the morning, when the noise of the day has died down) and try to release the thoughts in a torrent of energy. Actually, I’ve done a lot of my best writing on airplanes when there’s no wifi and no distraction. This is funny, because you are always a little hypoxic on planes, so your brain isn’t necessarily working at its best. Sometimes, I land, read the draft, and think to myself “wow, this is total gibberish”.

Then I take that first draft and sit on it, usually for around 48 hours at least (but sometimes, for months or literally years). Unless it’s super time sensitive. That way, I can return to it with fresh eyes and be very critical, like it’s someone else’s writing. Then I perform the difficult process of totally eliminating portions that aren’t efficient. You have to be extremely critical with yourself. The problem is you feel a huge “sunk cost” in terms of your words - you know exactly how long it took and how hard it was to write that page or that paragraph, and you’ll make up any excuse to yourself for keeping it in the final thing. But you have to be tough with yourself, and be super surgical. Your first draft is never good, so you can’t be happy with it right away.

After this initial stage is over, I spend a few days or weeks gathering the remaining data, facts, or interview content that I need to complete the story. Usually I am doing data-driven stuff, so this is often the most time consuming. But it’s easy at this point, because I tend to know exactly what I need. The last step is feedback. I’m lucky enough to have a bunch of subject matter experts at my disposal that are also great writers, and are happy to review my work meaningfully. (I’d estimate at least 50 people that I occasionally send me stuff to). In return, I’ll review their work. This part of the process is essential. Good third party feedback is absolutely essential - I don’t publish without it. This is a network that has to be cultivated over time. Normally, people are nice and don’t want to criticize harshly. This is a poisoned chalice. You deserve great feedback, and if you are willing to offer it to them, they’ll do the same for you.

This process has a strong filter. From top of funnel “idea stage” to finished article that I’m happy with, probably only 5% of articles make it. The fact that I don’t publish on a fixed schedule is what allows me to do this. I can select for my best ideas and be ruthless with my mediocre ones.

Hope you found this interesting!

We are in the post-truth era. But really, we’re in the pre-truth era.

So at this point we know we can’t solve epistemic uncertainty introduced by AI by using “deepfake detectors”. The only way is via negativa, aka committing to everything that is real, and assuming that everything not committed to is fake. Already, social media is an epistemic warzone, as we now have plenty of evidence of cheap fakes tricking people, but also, true real information being discredited because people plausibly allege that it’s AI. So we’re in the very worst part of the cycle right now, where people haven’t fully recalibrated their credulity yet (because they are still operating under the assumption that fakes are costly) and so AI content can arbitrage that. Soon enough, people will misbelieve all content online and assume it’s a priori fake. This isn’t good either, as no one will be able to convince anyone of anything. At that point hopefully we will begin to create additional layers of verification to address content authenticity. I’ll give a few examples of different media types and how I expect blockchain to assist in bringing about the era of verifiable content.

Personal content

So one problem people face these days is accusations that you did or didn’t do something, said something untoward, etc. If you are a high profile person whose value depends on your reputation this matters a lot. Currently, the way to prove that you didn’t do something, if accused of a crime, for instance, is to prove that you were somewhere else at the time, i.e. with surveillance footage. But with AI, the cost of creating arbitrary videos or audio of some third party is trivial, especially if there’s enough training data (which for anyone that appears on podcasts there is). This creates new problems. One product I expect to emerge is a personal verifier, namely a device that you wear that records video/audio 24/7, and location data. This dataset isn’t published anywhere, but hashed (every 10 min let’s say), and registered on a blockchain. If someone later claims you did or said something on x date, you can prove to them after the fact precisely what you did, and when, by selectively revealing a portion of the video or the transcript (plus location data). The first versions of this are the AI pendants like Tab or Rewind (https://medium.com/dare-to-be-better/ai-wearables-the-next-big-thing-a84ad82e4132). I actually preordered the Tab because I was so interested in the concept. I don’t know if they have plans to put the hashed transcripts onchain, but I’m going to ask them to. It means I’m effectively “committing” to everything I do and say, but I have the free choice to reveal it or not. It still puts all the power in the hands of the person. I do think this is a huge, immediately addressable use case.

Text content in the press

Something I’ve been thinking about for a long time is simply the need for versioning with media enterprises (see https://twitter.com/nic__carter/status/1714664977414369613 and https://www.coindesk.com/tech/2020/07/13/version-control-can-help-the-media-win-back-reader-trust/). This isn’t an AI problem, it’s an internet problem. Media companies are now used to the ability to change their article content post-publication to fix typos or completely revise stories in response to new events. But this introduces a new, bad paradigm of news. They are no longer required to “commit” to the articles that they write, allowing them to be much sloppier. And the practice of stealth editing (as we say infamously with the NYT this week regarding the hospital attack in Gaza) infuriates readers and undermines trust. The obviously solution is to hash article text and upload it onchain when you publish. When you change something, simply upload a new hash. Thus, readers can get a sense of changes to articles as they occur.

Blockchains specifically fix this, and not centralized DBs, because we’re dealing with trust. Many have pointed me to Wikipedia’s own version control product and said it fixes this, but it doesn’t, because the whole problem is we don’t trust the NYT, or whomever, to faithfully administer a DB governing their own edits. We need the data to be on a public ledger for public scrutiny. This is such an easy win, but no legacy publisher will adopt it because it makes their lives more difficult and forces them to return to a pre-internet model of journalism. I expect newer players that are trying to compete on trust will be the ones that adopt this. One company that does this is called Wordproof (https://wordproof.com/).

Image content

The next thing is proving that images and videos aren’t fake or AI derived. You can’t know this from looking at the image itself, especially as AI gets more sophisticated. As I said before, the only way is via negativa, by committing to everything that is true, and assuming everything else is fake. One way to do this would be to embed HSMs in smartphones and have the phone sign each photo. (And since you logged in with biometrics the phone assumes it’s you taking the picture. Optionally, you could also re-authenticate for each photo or video). Uploading a hash optionally on chain with location and identity metadata would allow you to situate a specific photo or video to a specific person, device, and time. That’s quite powerful, and while not perfect (I think people plausibly allege that HSMs are not bulletproof), it’s still a long way towards verifiable content. I do think this has to happen at the point of origin though. It doesn’t really work to have someone take an existing image and sign it and commit to it on chain. One of our portcos does this actually – Attestiv (https://attestiv.com/technology/#hfaq-post-3523) – and they’ve found traction in the insurance space (because insurers want to know that an image of, say, a fallen tree on a car, is recent and created by you). They use an app interface, so they issue a “challenge” to create the image, and then the user reacts to that and create the image. So they know that it didn’t exist before a specific time.

The problem is manipulating an image, because people like to edit their images and video naturally. This is where ZK proofs come in, I think. This is still new terrain, but I think it’s possible to issue a proof that you created some image data, and applied a transformation to it (say, cropping, or changing levels or contrast, or compressed it, or changed the format from HEIC or RAW to JPEG), which corresponds to the final image file. This seems like a very hard technical problem, but not an impossible one.

One other problem is taking an image of another image. Let’s say you wanted to “prove” you had just witnessed a rocket strike a building, and you took a picture with your smartphone of your laptop screen with an image on it, and represented that other image as your own image. That’s solved with “liveness” challenges in my view. Similar to how KYC firms require you to actually prove that you have a 3D face, rather than just uploading a static pic. Your phone could require you to waggle the phone around a bit so it knows it’s taking a picture of a dynamic, 3d surface, rather than just a flat image.

Another problem is ensuring personhood. What we’re really proving here is that a specific device created an image at a specific time and location. That actually doesn’t guarantee that a human created the image, just that an image was created on the device. This is why biometrics are so important in my view. A lot of hardware is going this direction. You now even have smart guns that wont fire unless the owner who is registered to the gun authenticates with a fingerprint. I think most cameras will eventually add some biometric verification too. Smartphones are the obvious place to start here since they already have it embedded.

Regarding the limitation that you mention, you could always adopt a protocol such that you have to sign the image content twice over 4 hours or something, such that a point in time credential isn’t sufficient. If the device is lost and a malicious third party uses it to create content, the requirement that you re-authenticate with biometrics upon each image presumably takes care of that.

Summary

I think this is going to happen. There’s no real limitation, software or hardware wise. Just a question of desire. If I had to make a guess, I believe Apple will be one of the first to react and will create something like what I’m envisioning (optional image metadata blockchain upload) within 12 months. The problem is simply too urgent. Eventually, I think people will look back on our current era of unsigned content and find it barbaric. So ironically, AI is going to push us through a primitive era of truth generation (based on the cost to create fake content), into a more credible era, as we all start to actually sign and take responsibility for the content we create.

I'm going to focus mainly on traditional audit, as in CPA services. For crypto native audit you have Certik, OpenZeppelin, Peckshield, Halborn, Chaos Labs, Trail of Bits, Quantstamp, Slowmist, Hacken, Spearbit, among others. We are Halborn and Chaos Labs investors so I have a preference for those naturally. To a certain extent, these crypto native audit firms are somewhat fungible, although they may have different focus areas. Some focus on crypto-economic design whereas others focus on pure code audit. I would go for references here. I think what's trickier is evaluating a crypto-native CPA firm (as in, traditional audit and assurances).

The first distinction would be defining “good auditors.” Depending on your circumstances, “good” might mean different things.

For example, the Big 4 accountancy have the people, process, and technology, to perform extremely comprehensive audits and attestations. They have huge national offices, with thousands of partners, quality control personnel to ensure the quality of the audits are “good.” However, the flipside of this is that B4 audits can be extremely invasive, time consuming, extremely expensive and often perceived as “overkill,” especially if you are a startup. There are also mid-tier firms, and smaller boutique shops, all with their pros and cons depending on your specific list of considerations.
So, if you were looking for a “good firm” for a FS audit, I would consider the following

• Comprehensive and thorough quality control and audit processes
• Experience w/ the subject matter, both from a technical accounting AND technical crypto standpoint
• An understanding of the true risks of the business and an approach to Identify and mitigate those risks without going overkill
• People you like working with – you will be spending a lot of time with them.
• Size: sometimes your investors or an outside party may require an auditor of certain size or reputation.

As you go through the Request for Proposal process, you will have an opportunity to learn about your prospective auditor’s approach to these considerations. During these introductory calls and throughout the process, I would:

• Test your auditor on technical components of the crypto audit. For example,
• Ask them how they plan on proving ownership of your digital assets. Are they going to use digital signature methods? Or send to self-transactional methods?
• Ask them how they would approach covering extended public keys? Can I just prove ownership of the first key? Or do I have to prove over all the keys in the derivation path?
• What about cDai and receipt tokens? How would you value these? At the value of cDAI on CMC? Or the principal and interest accrued?
• Your goal is to assess the audit firms technical understanding of crypto. They should have at least one person on the team give you a legit and coherent answer. If not, they might be bluffing and are going to cost you a lot of time, effort, and headache trying to get them up to speed.
• To get a sense of the firm’s thoroughness, ask them for their client onboarding documents and process. Additionally, ask them for any resources they might have to prepare for the audit.
  Ask for REFERENCES:
• Who are their existing clients in the space that are most like you? Did they do a financial statement audit for them? Ask them about the pain points in those audits and what they would do to make the process less difficult during your audit.
• Do a “background check” on your auditor
• Look up if they have any outstanding enforcement actions (for PCAOB registered firms): https://pcaobus.org/oversight/enforcement/enforcement-actions or privCo firms: https://us.aicpa.org/forthepublic/disciplinaryactions

In the US, the Accounting firm industry has been slow to service the space. However, as time goes on, more and more firms are building up the expertise and risk-tolerance to service the space.
From an audit perspective, these firms are audit either public companies, issuing public attestations (a la stablecoins) or are known to service the space in an audit capacity in some shape or form. This is not a comprehensive list, but should help you get started:
B4: Deloitte (USDC/Coinbase), EY (EY I consider very technically competent but as with all of the Big 4 they are politically constrained from doing crypto stuff)
MidTier: Grant Thornton (previously Coinbase), Marcum (which merged in Friedman – a few miners, Withum (Paxos), Aprio, Mazars, BPM, RSM, Cohen & Company
Smaller/Boutique: The Network Firm, Prescient Assurance

I know the principals at The Network Firm and trust them. They have a ton of experience given their history with crypto-native audit at Armanino.

From an international perspective, this a lot more challenging. Firms can operate as independent entities, even though they are under the same “name.” for example, KPMG UK may operate largely independently from KPMG Malta. The latter may service the space, but the former may not depending on their office/national offices approach to the industry.

(thanks to Jeremy Nau for his feedback on this answer)

So I’m going to limit myself to nonfiction books in this section. The majority of my knowledge comes from books. I know this is a bit unusual in a world of podcasts and twitter, but the stuff that really seeps in for me and that I retain ordinarily comes from reading plain old fashioned books. I am including only nonfiction although I have read and enjoyed a lot of fiction, I don’t see that as salient here. Also, I’m limiting it to books that gave me some brand new insight and that stayed with me meaningfully. I’m also limiting the list to the most important themes (in my view). The ones in bold are those that I consider particularly important. Aside from books, I get a large portion of my knowledge base from papers, so we can always address those in a future section, although that’s much more specific to the subject matter. And re Vitalik, the cohort of orbers keeps growing impressively so I think he will see the merit eventually, especially as it makes usage of the harberger tax which he has championed in the past. That said, I have instructed Eric as the chief orber to sell him on the idea when the time is right.

Crypto

• Digital Cash by Finn Brunton
• Bitcoin and Cryptocurrency Technologies by Narayanan
• Grokking Bitcoin by Kalle Rosenbaum
• Bitcoin Clarity by Kiara Bickers
• Applied Cryptography by Wiley
• The Blocksize War by Jonny Bier
• Crypto by Steven Levy

Finance / entrepreneurship

• Bad Blood by John Carreyrou
• Billion Dollar Whale by Tom Wright
• Adaptive Markets by Andrew Lo
• VC, an American History by Tom Nicholas
• The Mystery of Capital by De Soto
• Shoe Dog by Phil Knight
• Elon Musk by Walter Isaacson

Monetary history

• Free Banking in Britain by Larry White
• Payment Systems in the US by Benson Loftesness & Jones
• Bagehot by James Grant
• Competition and Currency by Larry White
• Good Money by George Selgin
• The Gold Standard (collection of essays) ed, Earle and Luther
• Money Free and Unfree by George Selgin
• The Theory of Monetary Institutions by Larry White
• When Money Dies by Adam Fergusson
• Monetary Regimes and Inflation by Bernholz
• Debt, the first 5000 Years by Graeber

Energy

• Energy and Civilization by Vaclav Smil
• Shorting the Grid by Meredith Angwin
• The Grid by Gretchen Bakke
• The Moral Case for Fossil Fuels by Alex Epstein

History / biographical

• Homage to Catalonia by George Orwell
• Down and Out in Paris and London by Orwell
• Ages of Discord by Peter Turchin
• Human Smoke by Nicholson Baker
• Mine Were of Trouble by Peter Kemp
• Always with Honor by Pyotr Wrangel

Philosophy / politics

• A World After Liberalism by Matthew Rose
• The World Turned Upside Down by Yang Jisheng
• Main Currents of Marxism by Kolakowski
• Bronze Age Mindset, BAP
• UR volume 1 by Curtis Yarvin
• Sun and Steel by Yukio Mishima
• Harassment Architecture by Mike Ma
• The Theology of Liberalism by Eric Nelson
• Democracy – The God that Failed by Hoppe
• The Myth of the Rational Voter by Caplan
• L’Etranger by Camus
• Superintelligence by Nick Bostrom

Health, longevity, and nutrition

• Outlive by Peter Attia
• Lifespan by David Sinclair
• Deep Nutrition by Cate Shanahan
• The Big Fat Surprise by Nina Teicholz

Misc

• Eiger Dreams by Jon Krakauer
• Into Thin Air by Jon Krakauer
• Disunited Nations by Peter Zeihan
• Mere Christianity by CS Lewis
• The Information by Gleick
• The Right Stuff by Tom Wolfe
• Skunkworks by Rich and Janos
• Touching The Void by Joe Simpson

Coinbase

So Coinbase has not done a PoR nor do they seem likely to (despite my many entreaties). The TX PoR legislation HB1666 (https://legiscan.com/TX/bill/HB1666/2023) which would ordinarily cover them (since they meet the criteria for having operations in the state) gave them a carveout since they are publicly traded company that adheres to a high standard of audits. So I can’t apply my PoR methodology because they don’t do a PoR.

Coinbase did write a blog post a while back about why their approach of having formal audited financial statements surpasses the assurances you get from a PoR (https://www.coinbase.com/blog/how-crypto-companies-can-provide-proof-of-reserves). Interestingly, the SEC’s much-maligned rule change, SAB121 (https://www.sec.gov/oca/staff-accounting-bulletin-121), which forces custodial firms to account for customer cryptoassets ON balance sheet (as opposed to off balance sheet, like Coinbase used to), means that Coinbase is now listing customer assets as a balance sheet item in their quarterly filings. [Note: SAB121 is a horrible policy, because it means banks have to take a capital charge against cryptoassets they would seek to hold for clients, making it completely uneconomical for banks to custody crypto. But in this case it’s helpful from an accounting perspective]

So Coinbase won’t provide a PoR, but they have to provide audited financials (Deloitte is their auditor) on a quarterly basis and as a public company they are held to a high standard. If you read their Q3 filing (https://d18rn0p25nwr6d.cloudfront.net/CIK-0001679788/9dcd01f7-3e99-4ad7-8d9c-fb39e00db165.pdf), as of eom June 2023, they had $124.3b worth of crypto assets held on behalf of their customers. This is a sufficient “proof of reserve” in my opinion as it gives, in my view, even stronger assurances than a conventional cryptographic PoR, since Coinbase is legally compelled to tell the truth here, and these numbers are audited. I still think they should do a higher frequency cryptographic PoR though. They don’t really provide a lot of detail here, no breakout by asset type. Even though it’s highly credible, it’s unspecific and low-frequency.

Coinbase also has a balance sheet they can use to remediate a loss they suffer if they are hacked (for instance in their Q3 filing they hold $760m of non-customer cryptoassets on their balance sheet, and $5.2b of cash), and they have the ability to obtain debt or equity financing if they run into trouble. From what I understand, Coinbase has strong controls in terms of security and IT, so I don’t expect them to be hacked. They also have a 10-year track record of not being hacked.

However, the risk of a hack is always unavoidable, and can never be fully mitigated. Given all of the above, I give Coinbase a 98% chance of being able to safely custody all customer deposits for the next four years.

Circle

As a stablecoin issuer, Circle doesn’t do a PoR either. A proof of reserve is applicable when a custodial firm isholding cryptoassets on the asset side, with client liabilities on their own database. The PoR endeavors to reconcile the two in a credible way. A fiat-backed stablecoin is the opposite - the liabilities are on the blockchain, and the assets are generally fiat assets in the banking system.

Circle has improved their disclosure framework. They provide monthly attestations signed by their auditor, Deloitte. This is not a “financial statement audit” the likes of Coinbase is required to procure though, it is an “examination” of management’s “assertion”. Still, Deloitte has a reputation to uphold so they have a very strong incentive to provide a reliable examination. Circle’s July report lists $26.4b in USDC liabilities, and $26.4b in assets held in reserve (https://www.circle.com/hubfs/USDCAttestationReports/2023/2023 USDC_Circle Examination Report July 2023.pdf). The reserves consist of $8.3b worth of short dated US Treasuries, $15.9b worth of reverse repos, $1b of cash in their reserve fund, and $1.9b of cash held in banks. The Treasuries are listed individually by CUSIP. The reverse repos are overnight cash lent to financial institutions, backed by Treasuries held by those borrowers. The cash is mostly held at BNY Mellon, the largest custodian in the world. The reserve assets, aside from the cash position, is bundled as the “circle reserve fund”, which is managed by Blackrock. Because each of the service providers involved is highly credible - BNY Mellon, Blackrock, and Deloitte – I have no concern about the asset quality here. Even if 3-month treasuries sell off (i.e. yields rise), and decline in face value, Circle can simply wait to collect the coupons and receive the full nominal payout. So even if their Treasuries decline in value in real terms, all they care about is the nominal payout, since their liabilities are also denominated in USD.

Although Circle wants us to think the assets are bankruptcy remote, in my opinion they aren’t really. They operate under state by state MTL licenses, rather than a state trust, like Paxos does with PYSUSD. It’s not clear that client assets would be protected in the case of bankruptcy or liquidation. Paxos chirped Circle for this a while back (https://paxos.com/2021/07/21/a-regulated-stablecoin-means-having-a-regulator/). Circle isn’t really regulated as a custodian of assets by a state regulator like NYDFS. The USDC terms don’t say anything about bankruptcy remoteness or treatment of assets in the case of Circle’s liquidation (https://www.circle.com/en/legal/usdc-terms). They do mention that USDC assets are segregated from Circle corporate assets though.

However it’s pretty hard to imagine a situation where USDC goes bust or takes on a ton of superseding liabilities. Circle itself also maintains a balance sheet from capital raised, although we don’t know its exact size. Circle had the snafu with their cash being held at SVB back in March 2023, but they have since changed the configuration. As of their most recent report, 88% of their cash was held at BNY Mellon. They have 12% at “other regulated US financial institutions” which they don’t name, but I expect they are using a more prudent model of using multiple banks to get “super FDIC” insurance (although we don’t know this for sure).

Overall, Circle assets can be considered highly secure, despite the lingering issue of questionable bankruptcy remoteness. I don’t expect any other SVB type situations to develop now that the Fed has reacted and created liquidity facilities like the BTFP.

Since the USDC asset side is in regulated traditional finance instruments, there’s no real risk of hacks here either. I give Circle a 95% of safely custodying all customer assets for the next 4 years.

Tether

Tether lists on their USDT transparency page $86.4b in assets corresponding to $83.1b in liabilities, with a $3.2b in listed shareholder capital as a cushion. They do quarterly reports, currently undertaken by BDO Italia (not a top tier auditor). Their latest is June 30, 2023 (https://assets.ctfassets.net/vyse88cgwfbl/63oJePOHqIvrcnXWMPZ1M0/4cfaf2e7cdf80c30b17fdc70faaf741f/ESO.03.01_Std_ISAE_3000R_Opinion_30-06-2023_BDO_Tether_CRR.pdf). Their reserves consist of $55b in 3-month Treasuries, $8.8b of reserve repos, $8.1b of MMMFs. They don’t list the Treasury CUSIPs and they don’t disclose where the assets are custodied. The weirder stuff is the following: 115m in corporate bonds, $3.2b in precious metals, $1.6b in BTC, $2.3b in unknown ‘other investments’, and $5.5b in secured loans (loans to crypto entities, if you believe press coverage, overcollateralized by liquid assets).

Each of these is problematic - the corporate bonds may not be liquid (we don’t know), the precious metals don’t correspond to the liability (gold could sell off against USD), BTC is volatile and clearly doesn’t correspond to the USD liability, there’s $2.3b of complete unknowns, and there’s $5.5b in loans to unknown third parties of unclear creditworthiness. On the face of it, this is pretty weak - we have to believe Tether’s relatively unknown auditor, they are doing quarterly and not monthly reports, they aren’t listing counterparties, custodians, intermediaries, and we don’t know where they are regulated or what their domicile is at all. The charitable interpretation is that Tether is opaque by design since they want to be as offshore as possible and they would expose their counterparties to governmental pressure if they named them. With the auditor, it’s likely that no big 4 auditor would agree to audit them since they have had issues in the past. And Tether maintains an asset liability mismatch on a liquidity and a returns basis.

However, Tether also makes a ton of money. $4.7b annualized at a 550 bps rate if they put everything into short term treasuries (and for their loans they are probably charging more). With this, they are able to build a reserve cushion. They also have a ton of breakage in their liabilities - likely a decent chunk of Tethers are simply lost on chain and will never be redeemed. (This is similar to how gift cards work, where some just get lost and never get redeemed). And Tether liabilities are very sticky, because a lot of their holders may not have access to the traditional finance system, so in a sense they are trapped in crypto. If you went into Tether specifically to move your funds away from tradfi, then you may not ever want to redeem. So Tether’s ersatz portion of their portfolio only becomes an issue if they have massive redemptions, which seems pretty unlikely based on history. Even throughout a brutal bear market USDT hasn’t seen that many redemptions, in fact they’ve been growing and their supply is at ATH.

Obviously we know nothing about the bankruptcy remoteness, so we have to assume USDT holders are unsecured creditors of Tether limited, which is registered in the BVI (not exactly a high quality jurisdiction). The main risk that I would be concerned about would be some kind of fine or enforcement action which impairs their ability to operate. If somehow a court decided they had to be shut down and unwound, likely there would be a long settlement process where USDT holders would be able to claim their pro rata share of the reserve.

Unlike the Tether truthers, I don’t think Tether literally doesn’t have the funds, but I can also acknowledge that they have clear and apparent deficiencies in their setup. I don’t worry too much about their lending activity, because that’s a relatively small portion of their reserves and they can handle a ton of redemptions before they get to that. The main concern would be enforcement, especially given their very patchy track record. It’s very unclear what is going to happen here, and it mostly depends on whether US enforcement agencies gain the political will to truly go after Tether. In the case of an enforcement action I think the settlement process will siphon some portion of client funds, similar to how FTX is being drained of its reserve by fees taken by the administrators.

Because of the poor quality of auditors, unclear legal domicile and legal status, questional reserve composition, and patchy track record, I give Tether a 60% chance of being able to fully honor client redemptions over the next 4 years.

Thank you for asking me a question where I am uniquely qualified (most likely top 10 worldwide) to answer. So last year I developed my PoR evaluation framework (see here: https://medium.com/@nic__carter/the-status-of-proof-of-reserve-as-of-year-end-2022-48120159377c). It evaluates PoRs on the basis of credibility. It consists of six criteria:

• Cryptographic asset verification
• Asset coverage
• Frequent ongoing PoRs
• User verification of liabilities
• Liabilities credibly accounted for
• Auditor oversight

These criteria cover the narrow PoR procedure. As I mention in this article (https://medium.com/@nic__carter/proof-of-reserves-for-policymakers-ae59c4b1f917), PoR doesn’t cover a lot of things, like segregated client and operating capital, the existence of any large superseding liabilities, the official bankruptcy remoteness of client deposits, and operating in a jurisdiction with strong property rights and a functional legal system. These are accounting, contractual, and legal concepts which cannot be satisfied by a technical procedure like PoR. So I will add another three prongs to the above analysis:

• Bankruptcy remoteness
• Segregated client and operating capital
• Stable and functional regulatory domicile

Now running through the various PoRs, starting with Binance

Binance PoR

• Cryptographic asset verification: It’s not clear that Binance is signing these addresses. It looks like they are just publishing a list of addresses. 0/1
• Asset coverage: their coverage is comprehensive. Well done 1/1
• Frequent ongoing coverage: they are doing PoRs on a monthly basis, which gives them full marks. 1/1
• User verification of liabilities: they are using the ZK method, having moved on from the merkle method. This means users can verify their inclusion in the liability set without leaking data. This is the state of the art, although it’s more black box-y. 1/1
• Liabilities credibly accounted for: this means being clear about how they are accounting for things like loans against the business and margin accounts on exchange (which have to be treated with negative values for the PoR ordinarily). Binance is not clear about these. 0/1
• Auditor oversight. The website claims ‘third party audit’ but there doesn’t appear to be an audit report, and certainly no Big 4 auditor would engage with them. It isn’t clear who they are using and it’s not easy to find any kind of audit report, whether from a CPA firm or a security company. 0/1
• Bankruptcy remoteness: Binance is unclear about the treatment of client funds, most likely clients are unsecured (aka junior) creditors of the exchange. 0/1
• Segregated client and operating capital: Binance is notoriously sketchy about this, there is no question they have comingled in the past and likely do today. 0/1
• Stable and functional regulatory domiciles: it is unclear where they are regulated. 0/1

Overall, while Binance has come a long way since their early efforts at PoRs and should be lauded for that, they still score poorly on my rubric, getting a score of 3/9.

The main problems are the obvious lack of a regulatory jurisdiction, the apparent comingling of exchange and client funds, and the apparent lack of an audit. Some of these issues are easy to solve, like cryptographic attestations to the attets and being more clear about how they are dealing with different types of liabilities. Regarding the comingling, I will say that Binance does appear to have a meaningful amount of assets which are growing stably. You can see this on Nansen (https://portfolio.nansen.ai/dashboard/binance) or Cryptoquant (https://cryptoquant.com/asset/btc/chart/exchange-flows/exchange-reserve?exchange=binance&window=DAY&sma=0&ema=0&priceScale=log&metricScale=linear&chartStyle=line). In fact, they’re almost at an ATH in terms of their BTC reserves. So It’s far fetched to think this is all operating capital and not client capital. The concern is simply that they may not be segregating the two in an accounting or custodial sense. The probability I give them of maintaining client funds on a 1:1 basis over the next 4 years is 70% (this is based on the strength of their balance sheet and likely ability to fill a possible hole, however the level of regulatory scrutiny they are under is hampering their operations).

OKX PoR

I happen to know OKX leadership personally, so consider that a disclaimer. (Granted, I’ve also interacted with Rana at Binance who was on a panel I hosted, and I think highly of her).

• Cryptographic asset verification: OKX signs all the wallets they control. 1/1
• Asset coverage: comprehensive. 1/1
• Frequent ongoing coverage: monthly. 1/1
• User verification: they are using the ZK method, partly at my suggestion actually. Clear instructions for clients on how to verify 1/1
• Liabilities credibly accounted for: they are very clear about their “constraints” - namely that their assets are the sum of all user balances, balances cannot be negative (that’s one way to cheat a PoR), and that every users balance is included. This is clear and helpful. However, there’s still a certain amount of trust here, that would be ideally extinguished with a third party audit. So they get partial scores here. 0.5/1
• Auditor oversight: Not present. 0/1
• Bankruptcy remoteness: OKX has licenses in Bahamas, HK (I recently confirmed with the regulator the quality of this license) Brazil, and Seychelles. I took a cursory read through the ToS and did not find language saying assets were bankruptcy remote, held in a trust for the ultimate benefit of users, senior in the case of liquidation, etc. I may be missing something important here, and I will actually clarify with OKX leadership to see what the situation is. Given my understanding of HK regs, their digital license there should be fairly sound with respect to user assets, but that may only apply to HK persons. I give them partial scores here and recommend they make this more explicit in their ToS. .5/1
• Segregated client and operating capital. I assumet they do this, but the ToS again isn’t clear on this front. 0/1
• Stable and functional reg domiciles. Partial marks here. .5/1

OKX gets a 5.5/9. Very good marks. They can improve by clarifying their ToS with respect to bk remoteness and segregated client / operating capital. Obviously the other next step would be getting auditor coverage which would further improve their score, but naturally that’s extremely difficult as auditors are notoriously leery of the PoR space. The probability I give them of maintaining client funds over the next 4y is 90%. This is based on their credibility, my knowledge of leadership, their demonstrated commitment to transparency (they did a PoR in 2015!), and the overall quality of their PoR.

Bybit PoR

I previously gave Bybit a 4/6 on the PoR score (that’s the first six tests) (https://medium.com/@nic__carter/the-status-of-proof-of-reserve-as-of-year-end-2022-48120159377c). Let’s dive in.

• Cryptographic asset verification: Bybit uses the send to self method 1/1
• Asset coverage: very comprehensive. 1/1
• Frequent ongoing coverage: every two months. This is better than quarterly, so I give them full marks, although it could be better. 1/1
• User verification: Bybit uses the merkle approach. This could be more private, but this is still an acceptable method. I recommend they consider ZK. 1/1
• Liabilities credibly accounted for: Bybit is quite clear what the verification applies to. However without audit coverage it is hard to know how exactly they are treating margin accounts etc. 0.5/1
• Auditor oversight: Not present. 0/1
• Bankruptcy remoteness: There’s no clear language in the ToS about client deposits being senior or protected in bankruptcy or liquidation. Also, their legal domicile is unclear. They mention Singapore in their ToS but I don’t think they’re officially licensed there, although I could be wrong. 0/1
• Segregated client and operating capital. ToS unclear on this front. 0/1
• Stable and functional reg domiciles. Again, unclear. Mentions of BVI, Singapore, Dubai in the press. Work to do here. 0/1

Bybit gets a 4.5/9 on my expanded framework. They need to bring in an auditor and clarify client asset status and find a stable domicile for a license. I give them a 65% chance of maintaining client assets 1:1 over the next 48 months.

Lastly, on your question regarding exchange risk. I think the glassnode tools mentioned are usefully indicative, including especially the metric measuring turnover relative to exchange held supply. Rapid inflows or outflows are also indicative, although failures in tagging could add noise to this data (for instance if they rotate to an untagged address). Also, the rise of 3rd party custodians complicates this analysis, as exchange assets are held elsewhere. With the rise of copper, clearloop, hidden road, and others, expect this to get murkier.

This is a great and very tricky question.

I had to look into why rebasing tokens don't work in Uniswap. As far as I can tell, changing the quantity of a token in a pool while keeping the other one invariant violates the core notion of the constant product. Interestingly, stETH fails in different ways for uniswap v2/3. The proceeds in 2 are donated to the pool, in 3 stETH doesn't work at all.

I think the answer hinges on whether uniswap and other such defi products can be made to accommodate rebasing tokens. I think this is unlikely, even if stablecoins move to fully interest bearing. Already, defi collateral moved from ETH to stETH, but this wasn't sufficient incentive for major defi protocols to natively incorporate the unwrapped version. Even though I do expect most stables to move towards interest bearing (and I expect these will be rebasing, naturally), I don't think they will be able to change the core logic of defi on their own. So I think we will be stuck with the wrapping approach for now on existing major defi products.

That said, I don't think there's any fundamental reason new composable defi can't work with rebase tokens, it's just that those assumptions have to be built into the protocol from inception. The problem was that many of these defi protocols emerged prior to the creation of rebase tokens, so they didn't contemplate that eventuality. Because both stETH and yield bearing stables will be such important collateral types going forward, I would imagine that future defi protocols would specifically try and determine how to support rebase tokens.

This is a fascinating question so I’m glad you asked me. I think it’s actually one of the biggest enduring misconceptions in the crypto space, that you can plan for a distributive outcome by setting some initial parameters and letting the system go from there. Well, this idea is far greater than crypto of course – it’s the foundation of, for instance Rawlsian liberalism (you can try to design the political economy of society in such a way that ultimately minimizes inequality). I would say the notion of liberalism in general ultimately boils down to: humans have equal capacity and dignity, there are no natural hierarchies, and the political and economic structure of society can and ought be set up such that each individual is put in a position to thrive. More perniciously, we have collectivist concepts which have been attempted for over a century whereby the distributive outcome in society is considered too unfair to be fixed incrementally, and all existing power structures must be overturned in order to derive a new distributive outcome. This is socialism, or as it is actually politically imposed, communism. The alternative school, which I belong to, is that you can’t plan for specific egalitarian outcomes, and that we have 100 years of evidence of tinkerism that it doesn’t work. Instead, you give people the tools to be meaningful economic agents themselves and have ownership of their labor. From there, wealth is created. It’s undeniable that the capitalist system which is responsible for industrialization and modernization is responsible for the vast majority of scientific and technological progress, whereas socialist systems did relatively little to add to that total. As far as I’m concerned, the ingredients for success (and thus, strong distributive outcomes) are political stability, cultural coherence, a respect for property rights (so that individuals are incentivized to build things, as they know they will be able to retain them), free trade and commerce, and the existence of capital markets. But even with these pillars present, you can’t plan for a specific distributive outcome – you merely have a society that experiences growth and collective advancement, and hopefully the government at that point has the resources to take care of the least well off, within reason.

Now what does this have to do with crypto? Well the question you’re asking me – about a technical system that can cause resources to be allocated, persistently, in a specific way (in an egalitarian manner) – is the same question that we have been asking ourselves for hundreds of years, especially as the battled has raged between collectivists and free marketeers. The answer is of course… you can’t plan for an egalitarian outcome. To do so would only be possible through tyranny, because humans are so diverse and have different inherent abilities to generate economic output. A setting where wealth settles through society in an egalitarian manner is only possible if you forcibly deprive everyone above the mean of their wealth – but that’s not a recipe for a healthy, productive society as we know. The main motive which drives progress, and R&D, and innovation, is the entrepreneurial spirit, and that only exists where founders know they will be able to reap the fruits of their labor.

I’ll be more concrete. When the Soviet Union collapsed, a question existed as to how we might distribute equity in all the factories and companies that were previously owned by the government. In the end, it was decided that the workers themselves would get vouchers which represented stock in these companies. So if you had been working in the factory throughout the USSR, you were now eligible for your pro-rata share of that factory itself. So today, Russia is one of the most egalitarian societies out there, right? Wrong. The vouchers were bought up for pennies on the dollar by local oligarchs or opportunistic hedge funds, leading to more concentration of wealth. Most people didn’t know what they were worth and had no thought of owning equity. For them being able to liquidate the vouchers and get a small immediate payout was worth it. Plus, they had no trust in the state. Who knows if the voucher would be worth anything in 10 years? Might as well sell it now.

But doesn’t the initial distribution for a new cryptocurrency really matter? Even though this is a strong part of Bitcoin’s founding myth, I’d say no. As you note, it really was insiders (basically people that read the cypherpunk mailing list or various niche libertarian and cryptography forums) that got a huge fraction of all the BTC “for free”. 50% of all BTC were issued in the first 4 years, for virtually nothing. What I would say matters is the direction of travel, not the initial conditions. Bitcoin has the very neat feature whereby the minters of the coins are not particularly privileged. Mining is almost perfectly competitive, and no miner really dominates – not for long, at least – and miners have thin margins and frequently go bust. This means that miners tend to be only temporary custodians of those newly issued Bitcoin, and none of them gain enough stature whereby they can use their protocol access to get super rich and then change the protocol in their favor. Staking, by contrast, is largely the opposite (in my opinion). The direction of travel for PoS, ceteris paribus, is weakly concentrative. Because the set of privileged entities that can stake at scale is a regulated function – basically, exchanges, asset managers, and custodians. And in those lines of business there are returns to scale, and it does pay to be large and politically important. So to summarize my view, you can improve things at the margin by making sure the business of seigniorage (creating the new units of currency) is a competitive one. This makes sure you don’t create a permanently privileged class that is able to harvest wealth from the protocol and turn it into political power. PoW accomplishes that – any open, competitive process does. But aside from that, you can’t make a grand plan to have your token be held equally by everyone globally, no matter how intricate your airdrop. This is for the same reason that you can’t make everyone the same IQ, height, skin color, attractiveness, and so on.

What about Worldcoin? Well Worldcoin will likely fail for two reasons: 1) the world doesn’t need a new free-floating SoV coin (this is obvious), and 2) just initially distributing the coin in a sybil resistant way is no guarantee that people will care about the coin and hold on to it and keep using it. 1 is trivial, Worldcoin has no theory whatsoever of why it should be valuable beyond “investors want to make money”, 2 is a bit more subtle. Recall the Soviet voucher privatization. Most people just sold theirs for pennies on the dollar because they didn’t know what it was worth, they weren’t in the business of speculation, and they had a shorter time preference. Foreign hedge funds came in and bought a lot of the vouchers because they were more sophisticated and understood the arbitrage. Oligarch robber baron types also ended up with a lot of vouchers through intimidation and coercion. So consider Worldcoin. I’m a Somalian and I scan my eyes. I get $20. Am I going to hold it in some complicated free-floating crypto that I don’t understand the value or prospects of, or am I going to sell it for USD cash, which I can use to buy cigarettes with today? The latter obviously. Just because Worldcoin is able to bribe people to scan their eyeballs doesn’t mean that they can force everyone to gain an equal understanding of the value and trajectory of the token and to gain sufficient technical competency to hold and use it. So Worldcoin thinks they’re solving the problem of sybil resistance in the issuance of scarce economic units, but what they’re actually trying to solve is the problem of people having different wants and needs, different economic circumstances, different time preferences, different risk tolerances, and different desires to speculate. That latter “problem” can’t be solved. It’s just natural human diversity.

So how would I distribute a new crypto token? The truth is, it doesn’t matter. If it becomes economically consequential, early insiders will sell out and the supply will naturally disperse that way, as the general population decides it’s worth owning some. Of course, this is a messy process – look at spontaneous dollarization in countries with failed currencies, the people who realize the direction of travel first end up doing best – but it can’t be any other way. If there is to be another SoV currency, some early insiders will get rich. That’s not something that can be planned out of the system.

One coda on proof of personhood though. I am actually intensely bullish on the idea, but not for distribution of new coins. Rather, as a way to solve the epistemic uncertainty that derives from the rise of cheap AI content. Historically, it was relatively hard to forge content, especially audio and video (and even photos). This is now trivial – effectively costing 0. So just in the last couple years, we have entered a new age of uncertainty. Some may worry that no digital content will ever be reliable ever again. But this is wrong. In fact, we’re merely in a brief interlude between two ages of certainty. And the coming age of certainty will be far more concrete than the previous one (which was dependent on the cost to doctor content). The way we will achieve this is not by proving that some content is AI generated (this is impossible), but rather by committing to all content that we want to later prove is authentic. This function will be incorporated into your devices, so you will simply have the option to sign a photo with your biometrics when you take the pic, and to insert a hash of the photo on chain. Video might be more technically challenging but still doable. Text is trivial. So biometrics and proof of personhood will be very important, but not for currency distribution. Rather, for selectively proving that certain content is authentic and was created by a human (or at the behest of a human). This will probably be a reality within 12 months.

 
So as written this question specifically asks about a Bitcoin-centric rollup on Ethereum. This means we can avoid a discussion of rollups on Bitcoin specifically. But I’ll dwell on this briefly. I have been an advocate of distinguishing the Bitcoin asset and the network (i.e., tokenizing Bitcoin and allowing it to settle on a variety of transactional media – whether that’s LN bitcoin, Bitcoin-native rollups, other blockchains, or other databases (like the Coinbase database)). I would analogize this to the USD network. Dollars don’t just settle as cash, or as push payments in the Fedwire system, but in all kinds of different settings – ACH, debit, credit, Fednow, wires, correspondent banks, money orders, checks, interbank settlement networks, gift cards, internal databases like Paypal, Venmo, Zelle, etc. Similarly, I want to see Bitcoin liberated as a Medium of Exchange from the tyranny of its own transactional network. This doesn’t mean reducing the number of mainchain settlements, but simply vastly broadening the utility of Bitcoin by making it usable in many other domains, thus ultimately driving up fees on the base layer itself (as these non-Bitcoin-network transactions would have to eventually settle to Bitcoin). The objective is to increase the economic and semantic density as much as possible and I believe making Bitcoin as a collateral useful in many other domains ultimately increases the security of the network on net, without overly adding to reorg risk. (One quick sidenote: as Vitalik I believe has mentioned, Proof of Reserve ideas ultimately intersect with rollups, and since Bitcoiners are generally pro PoR (at least the smart ones are, that realize we will inevitably have custodial intermediation somewhere), you could imagine that they would eventually see the value for a ZK rollup.)
 
If I were marketing a new rollup to Bitcoiners I would first and foremost aim to build it on the Bitcoin L1 specifically. Most Bitcoiners do not share my perspective around tokenizing BTC and using it on a variety of transactional domains, but I believe that many Bitcoiners are open to the idea of Bitcoin native L2s (even non-LN ones). A Bitcoin native rollup would require a new ZKP verifier to Bitcoin Core as we know. This would mean a soft fork. There are firms working on this and I am optimistic about this pathway. A sovereign rollup should be doable on Bitcoin too as well using the blobspace made available via inscriptions, and wouldn’t require a soft fork. But a sovereign rollup is ultimately not long term satisfactory, since there is no trustless bridge between mainchain assets and the rollups. But a sovereign rollup is appealing because it could migrate to a new L1, so you can always start one on Ethereum (if you’re so inclined) and move it to Bitcoin as a full validity rollup if a soft fork comes through.
 
But this wasn’t the question! The question is about a Bitcoin rollup on Ethereum. This is a good approach, since many rollups exist on Ethereum and we’re not starting from scratch. I believe ZK rollups are more appetizing to Bitcoiners than optimistic rollups, given Bitcoiners’ historical aversion to fraud proofs (granted, LN uses fraud proofs but this fact isn’t commonly acknowledged) and general desire for technically complete, even if difficult solutions (as opposed to worse-is-better cludgy solution). So I would go with a ZK rollup. Since we’re marketing to Bitcoiners, I would want three conditions to obtain:
 

• The collateral in the rollup has to relate to BTC itself – i.e. as liquidity is deposited in the rollup it has to ultimately be accretive to the value of Bitcoin
• The rollup has to be as trust minimized as possible
• The rollup has to have a value prop that aligns with Bitcoin values
• (Optional): the rollup has to avoid using ETH specifically

 
On the first point, this means that we have to tokenize Bitcoin somehow and port it over to Ethereum. Here we immediately run into a conflict with point 2, since there aren’t many trustless tokenized Bitcoin systems. tBTC is one option but it is small for now. I find it complex, and capital inefficient. One interesting way to go if we’re not too concerned about capital inefficiency would be a Bitcoin-backed stablecoin. There’s two ways to do this semi trustlessly and still have value accrete back to Bitcoin: a system like the original DAI (except backed just by Bitcoin rather than just by ETH), which I think Reflexer labs has suggested they want to work on. Alternatively, a synthetic Bitcoin stablecoin, which several folks are working on. You would achieve this exposure by combining a spot long with a short position (the basis trade) and in theory you are generally getting paid to keep the position on. I think the DAI approach is best. Bitcoiners are generally supportive of stablecoins so I don’t see this as too hostile to the Bitcoin philosophy.
 
Regarding Bitcoin values, the rollup on its own scales Bitcoin without making unacceptable trust tradeoffs so I think that aligns with Bitcoin values. If you can create private p2p cash-type transactions with Bitcoin collateral in a way that ultimately augments fees on the Bitcoin mainchain, that will satisfy all but the most hardcore Bitcoiners. (Stablecoin-skeptic and tokenized-bitcoin-skeptic Bitcoiners will fall in line within the next 12 months in my opinion, when the sheer force of market evidence becomes undeniable. There isn’t a big discussion of rollups in the Bitcoin community right now but I do expect that moderates will be on board soon enough).  
 
Regarding the optional fourth point, no current rollup on Ethereum as far as I know supports gas payments in wrapped bitcoin, so we would have to envision a new, theoretical rollup. Technically I don’t see any dealbreakers here. As far as I understand it’s possible to pay for L1 fees in non-ETH, so I think the entire rollup could use wrapped Bitcoin as an ERC20 even if it’s settling to Ethereum. (I think there must be a catch here because I see fee abstraction as hostile to ETH’s value prop but if there is I am missing it).
 
In terms of L2 stacks, I would favor the Starknet stack, since it is Cairo based and thus inherently designed to be ZK proven. This seems more in line with the Bitcoin design philosophy – doing it right with a purpose-built programming language, rather than inheriting the technical debt and zk-hostility of EVM/Solidity. Alternatively, in terms of achieving a positioning that is likely to be favorably received by Bitcoiners who value privacy, you could model it off a privacy-focused rollup like Aztec network. I don’t believe Aztec could work off the shelf with tokenized Bitcoin collateral though.
 
In terms of market positioning, I would market this to moderate Bitcoiners (the hardliners will reject anything built on Ethereum at all) and pro-Bitcoin crypto users generally. I would try to identify who is actively transacting with Bitcoin and is looking for cost savings or efficiency gains that come with a rollup. So I would look at people that use wBTC and may be looking for a less trusted product but are nevertheless comfortable with the Ethereum ecosystem. Exchanges, market makers, prop funds aiming to settle Bitcoin with each other with faster settlement could also be worthwhile anchor tenants. But overall I think it’s hard to forecast who will get the most benefit out of this because the use cases tend to emerge once the transactional space is created, and it’s hard to predict who would derive the most value out of a fast settling, private form of Bitcoin.