AI will make financial intelligence abundant. The harder problem is enabling agents to act safely across financial markets. They need a harness. Vaults are that harness.

This bias for humans exceptionalism is so deeply ingrained that even the great Iain Banks needed to self-correct. In the first Culture book "Consider Phlebas" Banks introduced the concept of "Referers" who are incredibly rare humans in the Culture civilization (only ~1 per trillion) that have a precognitive ability to predict the future in ways that Minds (AIs) cannot. Minds consult them and study how their abilities work. Banks later abandoned the idea, himself stating in an interview in 2010: "After thousands of years of artificial intelligence, the idea that the AI couldn’t do all this themselves is silly".

This is massive news but most people don't fully understand the moat that @Ondo has built and why Blackrock might want to do this. Ondo has created a unique (and today, arguably the only) structure that allows them to distribute tokenized securities widely outside of the US. For example if you're a platform with customers in LatAm, Asia, Europe, and other countries and want to offer exposure to US treasury yield, you don't have a ton of options. USDY is one of the few or only assets that is available in most of the relevant geographies. Presumably, it's this structure that made Blackrock excited to partner. There is a race happening to provide financial services to those who are underserved globally and doing so on crypto rails (stablecoins, tokenization, etc.) has become table stakes.

There is a good discussion to be had here, but it is wrong to claim non-custodial vaults are "solved". Let's think through some of the features of Morpho Vaults V2 and their costs / tradeoffs: - Timelocks: this is more of a security mechanism than it is a non-custodial feature. The argument goes that if users don't like new collateral markets being whitelisted, they have sufficient time to withdraw. But when you consider that most users (1) are coming through fintech platforms and (2) have no idea how to evaluate the risk of different collateral, there is very clearly still a major trust assumption. A user on Robinhood is not studying the collateral composition. They are trusting some combination of Robinhood, Steakhouse, or Morpho to protect them. - In-kind redemptions: Morpho Vaults V2 use a queue-based system where the caller can choose which markets to take ownership of. This creates an adverse selection problem: if there is an issue with one of the collateral markets, the least sophisticated users are the ones left holding that position while the more sophisticated users escape with only exposure to the safe markets. Allocation of liability is a complicated problem and letting the market fend for itself has tradeoffs. - Onchain accounting: it is uncontroversial to say that vault accounting should be as transparent and auditable as possible. It is a motte-and-bailey argument to suggest that doing all accounting onchain is the solution to this. The vast majority of exploits in the history of vaults has been due to manipulation of onchain accounting systems. When there is immediate feedback between an onchain action and the vault's canonical exchange rate, that system becomes a clear target. On top of that, we've historically seen dozens of scenarios where a collateral market is exploited, rendering the capital lent on that market never accessible (why would a borrower repay?). Those assets continue to be treated at 1-1 creating a bank run incentive and again leaving the most unsophisticated users holding the bag. - Immutable contracts: Making contracts immutable does not mean that the system cannot change under you. Smart contracts that are modular (e.g. use external adapters) can still change their security assumptions / functionality. The benefit of immutability is that these changes theoretically happen in a more predictable and transparent way. However, there are tradeoffs. If there is a bug in core smart contract logic, immutability makes it impossible to fix. My point here is not that these are bad features. I think directionally these features are attempting to address important problems in consumer protections. But these problems are nuanced and we shouldn't pretend like know the best structure to address them.

This acquisition makes a ton of sense if you think about it. There's a gap in the market for credible risk ratings of onchain products. A major part of that ratings process is assessing smart contract and technical risk. S&P has the brand and has already demonstrated appetite in this category. The thing that's missing is deep onchain technical expertise. The future of vaults has never looked brighter

I'm bullish on @arc. USDC is the dominant asset onchain and there is a massive opportunity to internalize that ecosystem to a purpose-built chain for payments, RWAs, etc. Getting Blackrock, the DTC, etc. to actually invest and have skin in the game was genius and is a clear indication that Arc is a deep strategic priority for Circle. No, it is not going to compete with Robinhood Chain on memecoin trading. Dream bigger

The long term bull case for memecoins is that AI eliminates the need for humans to do any productive work in the economy, so all that's left is entertainment. People love gambling and memes. Still, not the most inspiring use of our technology

CLARITY may not have passed but the underlying momentum that got us to this point is unchanged. Crypto will continue to become an everyday part of people's lives. We will get clarity sooner or later, one way or another. Until then we just keep building

One of the benefits of tokenization is enabling new usecases for traditional assets. There is a war happening now for market share in tokenized equities, and yield is always a powerful customer acquisition and retention tool. Super excited about these vaults - one of the most unique launches in recent history. Oh, and @veda_labs is now on Solana 👀

Dario is right

There is simply no faster way to lose aura than launching a token. It's a shame, because Hunter really was an exceptionally talented shitposter

Cannot wait for Muse to go native on WhatsApp. This kind of product is one where you really want a big credible company behind it. Insurance is just one example.

This heroic act has singlehandedly catapulted AI safety into the zeitgeist. The people who dunked on this as a useless gesture must be coping beyond belief. Regardless of your politics, it is clear that most people in the world are blissfully unaware of both the scale of progress and accompanying risks. We need more eyes on this technology. The chances of the powers that be taking this seriously have meaningfully gone up and imo this is a very positive outcome

It's amazing to me that people can read this and go "Anthropic just trying to pump bags ahead of IPO" instead of taking it seriously. Mostly traumatized crypto people who think everything is a casino and can't grasp how powerful this technology is. It is totally plausible for Anthropic to believe this and to continue doing what they are doing

AI may be the only technology in history to reduce humanity's control over Nature

It is actually very reasonable (and even commercially beneficial) to pause training until safety catches up. Reflecting more on the Hugging Face hack, it seems inevitable that frontier models will continue escape their sandboxes. Our best defensive agents are by definition less capable than the frontier models they are trying to secure. At some point, the models will become so good that humans in the loop are useless and then it's just agent vs. agent. Now imagine if instead of going after Hugging Face, the model went after a foreign government. The frontier model companies are one major hacks away from becoming nationalized

Sportsbooks = cigarettes Prediction markets = vapes

Over the last 3-6 months, Veda's pipeline has shifted to over 80% non crypto-native enterprises. It has been fascinating to discover the differences in product preferences and infrastructure requirements. I'd guess that many other crypto/DeFi companies, like us, built a ton of features for a customer profile that no longer really exists. The companies that will succeed in crypto are the ones that can do fintech/tradfi GTM but also can adapt their product offerings to meet the needs of this new customer type

"Crypto and stablecoins give financial products new capabilities: money can move globally, access onchain markets, earn programmatically, and still connect back to the payment networks people use every day. Increasingly, all these are starting to live together in the same product experience." Must-read

Fun fact: before @OpenRouter found PMF with agents/coding, one of its main early uses was roleplaying. SillyTavern was an open-source frontend where people created AI characters and plugged different LLMs in behind them. Its users were obsessive about switching models: Claude for writing, GPT-4 for intelligence, open models for fewer restrictions, etc. OpenRouter gave them a single interface to access and switch between all of these models. A good reminder that starting with a weird niche is still a great startup strategy: power users often discover important problems years before everyone else does.

Interesting commentary from @LucaProsperi on vaults: "the risk/return is not there and even the safest vaults can be lower yield than the risk-free rate." This is a valid criticism but also misses why these products are gaining traction: - Risk-free rate on dollars is not available globally - These products plug seamlessly into the stablecoin-native UX that fintechs are building - 24/7 liquidity Fintechs are increasingly serving global user bases using stablecoins. Vaults uniquely enable fintechs to build Earn into their core product offering without increasing their compliance or UX burden. Think about how amazing it is that @deel was able to quickly test a POC in Argentina, and as soon as demand was validated, instantly expand the product globally. That is simply not possible with the tradfi equivalent of cash products