After 2+ years in the robotics data space, we are shutting @Eidon_AI down. The thesis was right. But the business is brutally hard. We close this chapter by open-sourcing everything we built and sharing lessons for anyone venturing into the space.

I've been working on a project to fix cold outreach. Introducing Instant Expert: Reach out to anyone in the world, and only pay if they respond. They keep the money or donate it. I've been thinking about this problem for a while. 1/

Big news! Last Friday the SEC officially dismissed its enforcement action against me and against DeSo. This was the last legal issue I had to deal with and I am now completely and totally free to innovate and build again, unhobbled for the first time in years. Three important points: 1) This dismissal was NOT a settlement. It was "without costs or fees" to me or anyone involved (extremely rare) because there was no wrongdoing and no actual aggrieved parties. 2) This dismissal was "with prejudice" (also rare). This means they can't bring any related action back against me or DeSo in the future. 3) In the SEC's own words, it was based on "a reassessment of the evidentiary record," meaning the actual facts regarding my innocence were heavily scrutinized and drove the decision. Simply put: The government made a mistake in bringing this case in the first place. The government accused me of misleading an investor who I knew I had a great relationship with, as in they backed me two separate times and I literally had breakfast with them at their house not long prior to the charge. As a result, soon after the charge I found out that not only were they not upset with me, but they wanted the government to go away as badly as I did. As I understand it, the government compelled the investor to do an interview and then took their neutral testimony and represented it as adversarial. It was an alleged fraud with no actual misrepresentation nor any actual aggrieved parties. My lawyers said they'd never seen anything like this, and I think it speaks to how dogmatically anti-crypto the prior administration's SEC was. In the coming days and weeks, I will be hopping on some podcasts to tell the whole story, and boy is there a story to tell. Stay tuned, and if you know anyone who'd like to have me on as a guest please reach out. I'm also excited to start sharing more about what my team has been working on soon. We haven't been twiddling our thumbs. For now, though, I just want to explain why DeSo is so important to me. DeSo is still the only platform on the internet where you can post content directly to a blockchain without fear of censorship, and where you can monetize your content directly with crypto (including stablecoins). It's really quite shocking how in 2026 we not only have virtually no viable alternative for this clearly-important category, but also other important efforts are actually shutting down. The world needs more people working on decentralizing social media, not less. I'm excited to finally be able to share our vision directly again, and to start bringing more people who care about freedom and censorship into our community. What we have built with DeSo is something people take for granted until they really need it, but hopefully we can convince them sooner than that. Lastly I want to say how grateful I am to everyone around me. My family, my friends, my backers, and everyone in the DeSo community. For me, this experience showed me just how trusting, loyal and caring everyone around me really is, and reaffirmed my belief that always trying to do the right thing really does pay off. We're just getting started.

Finally had a chance to listen through this pod with Sutton, which was interesting and amusing. As background, Sutton's "The Bitter Lesson" has become a bit of biblical text in frontier LLM circles. Researchers routinely talk about and ask whether this or that approach or idea is sufficiently "bitter lesson pilled" (meaning arranged so that it benefits from added computation for free) as a proxy for whether it's going to work or worth even pursuing. The underlying assumption being that LLMs are of course highly "bitter lesson pilled" indeed, just look at LLM scaling laws where if you put compute on the x-axis, number go up and to the right. So it's amusing to see that Sutton, the author of the post, is not so sure that LLMs are "bitter lesson pilled" at all. They are trained on giant datasets of fundamentally human data, which is both 1) human generated and 2) finite. What do you do when you run out? How do you prevent a human bias? So there you have it, bitter lesson pilled LLM researchers taken down by the author of the bitter lesson - rough! In some sense, Dwarkesh (who represents the LLM researchers viewpoint in the pod) and Sutton are slightly speaking past each other because Sutton has a very different architecture in mind and LLMs break a lot of its principles. He calls himself a "classicist" and evokes the original concept of Alan Turing of building a "child machine" - a system capable of learning through experience by dynamically interacting with the world. There's no giant pretraining stage of imitating internet webpages. There's also no supervised finetuning, which he points out is absent in the animal kingdom (it's a subtle point but Sutton is right in the strong sense: animals may of course observe demonstrations, but their actions are not directly forced/"teleoperated" by other animals). Another important note he makes is that even if you just treat pretraining as an initialization of a prior before you finetune with reinforcement learning, Sutton sees the approach as tainted with human bias and fundamentally off course, a bit like when AlphaZero (which has never seen human games of Go) beats AlphaGo (which initializes from them). In Sutton's world view, all there is is an interaction with a world via reinforcement learning, where the reward functions are partially environment specific, but also intrinsically motivated, e.g. "fun", "curiosity", and related to the quality of the prediction in your world model. And the agent is always learning at test time by default, it's not trained once and then deployed thereafter. Overall, Sutton is a lot more interested in what we have common with the animal kingdom instead of what differentiates us. "If we understood a squirrel, we'd be almost done". As for my take... First, I should say that I think Sutton was a great guest for the pod and I like that the AI field maintains entropy of thought and that not everyone is exploiting the next local iteration LLMs. AI has gone through too many discrete transitions of the dominant approach to lose that. And I also think that his criticism of LLMs as not bitter lesson pilled is not inadequate. Frontier LLMs are now highly complex artifacts with a lot of humanness involved at all the stages - the foundation (the pretraining data) is all human text, the finetuning data is human and curated, the reinforcement learning environment mixture is tuned by human engineers. We do not in fact have an actual, single, clean, actually bitter lesson pilled, "turn the crank" algorithm that you could unleash upon the world and see it learn automatically from experience alone. Does such an algorithm even exist? Finding it would of course be a huge AI breakthrough. Two "example proofs" are commonly offered to argue that such a thing is possible. The first example is the success of AlphaZero learning to play Go completely from scratch with no human supervision whatsoever. But the game of Go is clearly such a simple, closed, environment that it's difficult to see the analogous formulation in the messiness of reality. I love Go, but algorithmically and categorically, it is essentially a harder version of tic tac toe. The second example is that of animals, like squirrels. And here, personally, I am also quite hesitant whether it's appropriate because animals arise by a very different computational process and via different constraints than what we have practically available to us in the industry. Animal brains are nowhere near the blank slate they appear to be at birth. First, a lot of what is commonly attributed to "learning" is imo a lot more "maturation". And second, even that which clearly is "learning" and not maturation is a lot more "finetuning" on top of something clearly powerful and preexisting. Example. A baby zebra is born and within a few dozen minutes it can run around the savannah and follow its mother. This is a highly complex sensory-motor task and there is no way in my mind that this is achieved from scratch, tabula rasa. The brains of animals and the billions of parameters within have a powerful initialization encoded in the ATCGs of their DNA, trained via the "outer loop" optimization in the course of evolution. If the baby zebra spasmed its muscles around at random as a reinforcement learning policy would have you do at initialization, it wouldn't get very far at all. Similarly, our AIs now also have neural networks with billions of parameters. These parameters need their own rich, high information density supervision signal. We are not going to re-run evolution. But we do have mountains of internet documents. Yes it is basically supervised learning that is ~absent in the animal kingdom. But it is a way to practically gather enough soft constraints over billions of parameters, to try to get to a point where you're not starting from scratch. TLDR: Pretraining is our crappy evolution. It is one candidate solution to the cold start problem, to be followed later by finetuning on tasks that look more correct, e.g. within the reinforcement learning framework, as state of the art frontier LLM labs now do pervasively. I still think it is worth to be inspired by animals. I think there are multiple powerful ideas that LLM agents are algorithmically missing that can still be adapted from animal intelligence. And I still think the bitter lesson is correct, but I see it more as something platonic to pursue, not necessarily to reach, in our real world and practically speaking. And I say both of these with double digit percent uncertainty and cheer the work of those who disagree, especially those a lot more ambitious bitter lesson wise. So that brings us to where we are. Stated plainly, today's frontier LLM research is not about building animals. It is about summoning ghosts. You can think of ghosts as a fundamentally different kind of point in the space of possible intelligences. They are muddled by humanity. Thoroughly engineered by it. They are these imperfect replicas, a kind of statistical distillation of humanity's documents with some sprinkle on top. They are not platonically bitter lesson pilled, but they are perhaps "practically" bitter lesson pilled, at least compared to a lot of what came before. It seems possibly to me that over time, we can further finetune our ghosts more and more in the direction of animals; That it's not so much a fundamental incompatibility but a matter of initialization in the intelligence space. But it's also quite possible that they diverge even further and end up permanently different, un-animal-like, but still incredibly helpful and properly world-altering. It's possible that ghosts:animals :: planes:birds. Anyway, in summary, overall and actionably, I think this pod is solid "real talk" from Sutton to the frontier LLM researchers, who might be gear shifted a little too much in the exploit mode. Probably we are still not sufficiently bitter lesson pilled and there is a very good chance of more powerful ideas and paradigms, other than exhaustive benchbuilding and benchmaxxing. And animals might be a good source of inspiration. Intrinsic motivation, fun, curiosity, empowerment, multi-agent self-play, culture. Use your imagination.

Today we’re announcing Coinwatch Track - a new way for crypto projects to see what their market makers are doing in real-time, using their API keys.

Openfund is getting critical upgrades that will make trading easier than ever. 1) Real-time infrastructure, via our web-sockets integration, is now live on Openfund. This makes Openfund on-par with UX you’d see on centralized CLOBs. The infrastructure upgrades that enhance these real-time components will also apply to all other DeSo apps that use GraphQL and State-Syncer, so expect to see significant performance improvements on Focus as well. 2) The Openfund Wallet has been upgraded to support multiple quote currencies like DESO, USD & FOCUS, including the ability to cash-out from $FOCUS directly via Openfund. We also included a Stake $DESO button so that more new users entering DeSo can learn about the benefits of staking their DeSo and earning 20% yield. 3) SOL-USDC support for both cash-in and cash-out, which we added recently, is growing at record levels. This proved to be a critical upgrade as it allows users to onramp into DeSo pretty much for free (network fees on average are less than $0.001). 4) UX upgrades we’ve added include the holders table on the trade page so you don’t have to click through between Focus & Openfund. We also added hover cards to view more details about users, similar to Focus. It’s a small detail, but note that we are the only social orderbook in existence where you can see exactly who is making a trade. Imagine looking at Coinbase and knowing who’s buying or dumping your token! 5) TWAP backend has been thoroughly tested and is now ready for frontend implementation. TWAPs will provide traders with a robust set of features that allow for more precise and secure trading. It’s difficult to understate how significant of an upgrade this is. Traders will soon be able to set a flexible schedule for placing bid or ask orders, executed through derived keys, with a number of configuration options like enabling slippage protection and the ability to safeguard against unfavorable trade executions with price limits. 6) Market making incentives are being planned. We’re working on coming up with a program to incentivize more third party liquidity. As a reminder, our Python SDK makes this much easier to do now programmatically. Also, thought I’d give a little shoutout to @consensusFuture who has done nearly ~$8m in trading volume!

Announcing our updated priorities for DeSo and our associated products. focus.xyz/post/e1035239370be…

The rumors are true and I can finally talk about it: The DOJ has dismissed its case against me and my name has been cleared. This is an amazing result for me, for my family, for my team, and for DeSo. There is no limit to what we can achieve from here. At some point, I'll share my full story. But for now, I just want to clarify a few important points since I can now speak openly. 1. My innocence withstood intense scrutiny. The government meticulously combed through my private texts, my private emails, and even private documents I'd written looking for any shred of wrongdoing. They went to people I'd done business with and essentially pressured them to say bad things about me (which nobody did, not even people I'd let go in the past). The process was extremely adversarial. They weren't looking for a reason to clear me, they were looking for a reason to convict me, and any reason would have worked as long as they thought it would convince a jury. After months of searching, using every method and tool at their disposal, including applying pressure to those around me, the government decided to dismiss their charges. It's hard to understate how rare a dismissal like this is. After going through this process myself and seeing what the government is capable of, I believe it's highly unlikely that anyone who has ever done anything wrong, or even anything that "feels" wrong, would ever survive the government's scrutiny without being convicted. I truly believe it only happened in my case because I've always gone above and beyond to do right by everyone I've ever done business with, and because I truly believe in my heart that what we're doing with DeSo is important for the world (and this came out in all my private communications). 2. There was no victim. In their complaint, the government claimed that a conversation they had with "Investor-1" led them to believe that I had defrauded this investor. Many things were incorrect about this claim after it was scrutinized: 1) I never lied about anything. In fact, I was beyond transparent at all times, and I'm confident that Investor-1 would agree 2) Investor-1 was and still is up on their purchase, even after the government's FUD tanked the price by over 70% 3) I am confident that Investor-1 does not consider themselves to be a victim. Not only that, but Investor-1 has never been anything less than an amazing partner to me all throughout my career for almost a decade now. When I saw them mentioned in the complaint I immediately suspected that the government had compelled their testimony, and was either misunderstanding or misrepresenting an innocent conversation to reach the conclusion they wanted to reach. I believe that if you asked Investor-1, the only entity they'd consider themselves a victim of is the US government for wasting so much of their time, and for costing them more in legal fees than the amount allegedly lost to fraud (which to be clear was zero because they are still up on their original purchase of tokens). In summary: I believe the case that was brought against me consisted of a no-loss non-fraud against an alleged victim who doesn't even consider anything negative to have occurred, other than the actions of the government itself. 3) DeSo is fully-decentralized. Perhaps the allegation that hurt the most was the government's claim that BitClout/DeSo, the blockchain that I've been working on for years now, is not fully-decentralized. They did this by pulling a text message I sent out of context. In the message, I said something like "even something that is fake decentralized would probably still not be a security." Right *after* that message I clarified that BitClout/DeSo is *actually* decentralized, and thus has virtually no securities risk as a result. Unfortunately, the government didn't include that context in their complaint, which in my opinion is an act of bad faith on the government's part. For the avoidance of doubt, I will say on the record right here and now that BitClout/DeSo has been fully-decentralized from approximately late 2020. To say I thought anything else would not only be wrong, it would contradict actual hard fact. 4) This was some hard stuff. A lot of things about what I went through were hard. One day I will tell the whole story and I think it will be quite interesting for people to hear-- but not today. I don't want to come off as arrogant or hyperbolic, but I feel I have to give my honest assessment and say that I'm pretty sure something like this would have broken most people. There is something "life or death" about a crisis like this that I feel few working in traditional companies have ever really dealt with, even at the highest levels. At minimum, it would break their team and make it hard to continue to operate normally... This being said, I'm proud to say that our team remained solidly intact, and we even successfully launched two major products through all the noise: Openfund and Focus (which you should try, by the way), as well as a major network upgrade to Proof of Stake. I always knew that I hadn't done anything wrong and that it would all get resolved. But everyone around me did as well, including my team. That belief, combined with the absolutely heroic support of my friends and family, made it manageable without too much stress. And of course, it doesn't hurt that I believe DeSo is one of the most important things I can be doing for the world, and worth fighting to the death for. Lastly, I have to mention that if it weren't for all of the efforts of others in our industry, especially @brian_armstrong and his work with Coinbase, I'm not sure crypto would be where it is today, and I'm not sure we would have gotten such a swift dismissal of my case. === In the short-term, I've got big plans for DeSo, Focus, Openfund, and HeroSwap (my team's core products). Every single one is best-in-class at what it does and a potential billion-dollar business on its own. Now that I'm able to operate at full capacity, free from stifling constraints, and with my reputation and network restored, I'm confident we'll realize that potential. Now, let's get back to work.

How much should a Layer-1 be worth relative to an app that's built on top of it? I believe this is the most important question in crypto right now, and one that will decide the ultimate fates of Ethereum, Solana, and every other blockchain on the market today. Yet I don't see many people talking about it. What should the layer-1 premium be and why? 🧵 Right now, I think it's easy to dismiss layer-1s as a commodity product that is overvalued by the market. The reason is that apps built on blockchains can theoretically "swap out" what chain they're using to minimize costs. For example, an app can theoretically move to whatever the lowest-cost blockchain is in order to power its features. And apps like Focus even build cross-chain features in from day one to minimize friction for users coming from other chains (it even supports BTC). But what determines how commoditized blockchains will be in practice? I think the switching costs boil down to three categories of friction: 1) moving assets, 2) moving liquidity, and 3) moving user data. And these frictions are heavily affected by whether an app allows users to self-custody or not. Let me explain... When you want to move a user from one blockchain to another, the first question is whether your app does self-custody: I.e. whether the user controls their assets or the APP controls their assets. An exchange like Coinbase is custodial (meaning the operators of the exchange control users' assets) while an app like Focus is non-custodial (meaning the user's assets and account are wholly controlled by the user themselves). What are the implications of this? For a custodial app like Coinbase, they are able to achieve virtually "frictionless" switching between chains. If tomorrow they wanted to convert all customers' USDC from Ethereum to Solana, they could do it (and Binance even went so far as to convert all stablecoins from other chains to BUSD on Binance Chain under the hood). Cases like these make it understandable to assume that the layer-1 premium will be small. After all, why would Coinibase nudge users to anything but the lowest-cost blockchain in the long run? But where things get more complex is when non-custodial apps enter the picture. With an app like Focus, all of your assets and content are stored directly on-chain. If you make a post on Focus, the content of that post is stored on the DeSo blockchain. And if you place an order on Focus (or another DeSo app like Openfund), then that order is actually stored directly on the DeSo blockchain as well. The same goes for other self-custody apps like Uniswap or Raydium on Solana. This is where it's important to consider the three categories I mentioned previously: 1) assets, 2) liquidity, and 3) user data. When an app that is self-custody wants to move users to another chain, it requires an onerous opt-in from the user because you literally need them to sign a transaction that moves their money from one place to another. You can't do it "under the hood" like Coinbase can, and that's kind of the point: Your money belongs to YOU, not the app that you're using. But moving a user's assets is actually the easiest thing to do because it's a single-player decision. A user can switch from using USDC-ETH to using USDC-SOL or even USDC-DESO without that much of a hassle. Where things get more complex is actually with liquidity and user data (the other categories). It's one thing to move a user's assets to another chain but another thing entirely to move *liquidity* to another chain. What do I mean by liquidity? Well, consider an on-chain order-book like the DeSo DEX that powers Openfund (but you can also consider liquidity pools like Uniswap or Raydium). If there's millions of dollars in resting orders on a DEX, moving a user to another chain makes it much more onerous to trade because you have to move them BACK to the high-liquidity chain in order to execute all of their future orders. You can do it, but guess what: You're going to have to pay the toll on the high-liquidity chain no matter what (the network transaction fee), meaning that liquidity yields pricing power. What's more, having liquidity on your chain results in repeated transactions, one for every trade the user wants to do over their lifetime, vs a single-transaction single-player switching like moving assets gives you. How powerful is the "moat" provided by liquidity on your chain? It actually depends on what assets we're talking about. For major assets with lots of liquidity across a lot of venues like BTC, ETH, SOL, etc... I think reasonably strong but it's important to note that it's most likely not a "winner take all" dynamic, but rather a "split" dynamic. Chains above a certain threshold of liquidity, enough to get within a few basis points of spread of each other for such assets, will likely be somewhat inter-changable, though still hold pricing power in proportion to their share of the market. Where things get really interesting, though, is on longer-tail assets such as meme tokens or, eventually, wrapped securities on-chain. If a chain is the exclusive originator of a token, it's highly likely that the liquidity dynamics will switch from a split to winner-take-all. For example, if you launch your meme token on Focus or on PumpFun, it's highly unlikely that there will be a lot of liquidity on chains other than the original "issuing" chain. In such cases, the "pricing power" a blockchain has switches more to a winner-take-all dynamic rather than a split in my view. Lastly, I want to talk about a category of friction that hasn't really shown up in crypto yet, which is "user data" such as a user's identity or a user's on-chain content (like on a decentralized social network like DeSo). When it comes to content, I believe blockchains have much more competition from centralized players than they do with regard to liquidity. Blockchains offer permissionless access to assets, as well as the ability to issue your own assets (eg meme tokens), which gives DEX'es a major advantage over centralized exchanges. However, no such advantage exists when comparing a decentralized social network to a centralized one (they are effectively substitutes except for the fact that one is non-custodial and can't censor your content). Nevertheless, it is important to note that if a decentralized social network like DeSo were to gain traction, then I believe it would have the ultimate pricing power via its content. This is because when you make a post on a blockchain like DeSo, it is "fanned out" to all of the apps built on DeSo, meaning your reach is maximized from posting to the chain with the highest content liquidity. You will always post to all chains to maximize reach, but the toll charged by them will likely vary more like a winner take all rather than a split due to the highest coverage of the "dominant" content chain. So where does that put the layer-1 premium? Again, right now layer-1 network fees are really cheap because everyone is in a cutthroat battle to gain market share, and so it's easy to dismiss all layer-1s as having "low revenue potential." But ignoring the underlying power dynamics of the chains leads to the incorrect long-term assumption that they will be worthless, rather than have potentially-significant pricing power due to 1) liquidity aggregation and 2) content aggregation. Similar to saying Amazon is worthless because it's "not profitable" in 2005. As such, I believe focusing on the long-term power dynamics yields a different conclusion, which is that layer-1s with significant liquidity (both asset and content liquidity) should be valued at a significant premium to their current fees, and possibly even in proportion to the revenue made by apps built on top of them. I have a lot more to say here. I'm writing this post stream of consciousness (didn't even reread it because too busy). I will probably just record a long video about it all. But these are my raw thoughts in response to a question I've been getting.

As the creator of the DeSo blockchain, I want to talk about how crypto will fundamentally change social media and solve the bot problem 🤖 When Elon took over X, he made everyone pay for a verified check. There's only one problem with that plan, which is that bots can easily afford it. How do I know? Focus, our next-generation SocialFi app, was ASSAULTED by verified bots when we launched our invite-only "Bounty Hunter," which gives $100 to Verified Twitter users. The problem is that it's equally expensive for a bot to buy a blue check as it is for a legitimate user, and so it doesn't provide enough differentiation. The key is to have something that's effortless for real users but prohibitively expensive for bots. What could that be...? It turns out crypto wallet balance is an *amazing* metric in this regard. Whales naturally keep thousands of dollars in their wallet, but a bot can't afford to keep more than a few hundred at most. And it scales in the sense that the more you have, the less likely you are to be a bot, even if you don't have a long track record of posting. Focus is truly novel in many ways, but perhaps the most interesting is how it tightly integrates assets with one's social identity. And it doesn't stop there. Once on-chain assets are tied to identity, you can build much better ranking algorithms, a better upvote mechanic that factors in wealth, and many other novel innovations that can finally prevent bots from having a disproportionate influence on our discourse. You may never have heard of Focus, but it's the only app in the world that's innovating on this stuff right now, and it's going into public access next week. We've been working on it for years and I'm excited for everyone to try it. I truly believe it's an app that can not only materially enhance the quality of our information on the internet, but also solve a lot of problems related to misinformation as well.

As the creator of BitClout, I want to share some thoughts about the future of meme tokens... 🧵 Today there is a problem with meme tokens, which is that they "end" when the creator decides to start dumping. Extraction by the creator signals the end of the party, and so the game is solely based around buying it before it pumps and dumping it before it's "over." Even worse, once a creator's balance of their own token hits zero, that's it. The creator has no incentive to continue caring about their token. But what if you could design a token without a discrete extraction event? A token that was designed to last forever, and earn the creator income into perpetuity...? I believe a better token model could become a key monetization mechanic for anyone with a following on the internet on the internet. Ideally what you want is for there to be NO discrete extraction event. No point at which a creator signals to the market that the party's over, and no point at which they become unable to monetize. The ideal would be that extraction is just perpetually happening, like the rake at a Poker table. This would not only give meme tokens longevity, it would **permanently align** the creator with the success of the token. This can be achieved with a new model for token monetization, whereby trading fees are paid to the creator on every trade of the token. At first blush, it seems so simple, and yet no platform does it. A fee-based approach eliminates all discrete extraction issues with the "dump on your followers" model used by pumpfun and others. When a creator extracts via trading fees, not only is there no discrete dump signal, but the the creator is also **permanently aligned** to keep the token interesting. In the old model, once a creator has a zero balance of their token, that's it-- they don't get any money from it anymore. In contrast, if extraction happens via fees, then the creator is permanently aligned to continue to care about the token because they continue to earn from it. The last thing I'll say is that, when we were doing BitClout, the single biggest hesitation from creators joining was "I don't want to monetize by dumping on my followers." I spoke with probably over two dozen A-list celebrities, and this was the key sticking point. As such, I believe this new model eliminates the single biggest known concern that real creators have when it comes to embracing crypto and tokenization. You can experiment with this new mechanic, as well as many other innovative concepts, on Focus, our new SocialFi app👇

Social order-books are the future