U.S. banks hold $11T+ in undrawn commitments. Promise looping is live on LCC, bringing fractional reserves to capital commitments in crypto Earn promise fees on up to 6.25x your net worth: +7.8% APY on top of your DeFi assets deployed elsewhere, funded via @Morpho debt if called

One of the biggest misconceptions about the middle market for compute is that there simply isn’t enough capital. There’s a ton of capital! It’s more the shape of it and the covenants required that’s sticky. The more interesting problem is the per-loan cost of getting capital there. Every loan has a stack of costs: Origination. Underwriting. Technical diligence. Legal structuring. Warehousing. Settlement. Distribution. And each layer has a real dollar cost. Talk to private credit funds and they’ll tell you that standing up a transaction can easily mean $3-5M of fees and internal/external costs before you even get to the cost of capital itself. On a >$500M transaction, that can work. On a $20M or $30M transaction, the unit economics start to look absurd. Those costs do not disappear. They surface somewhere: higher interest rates, larger minimum deal sizes, tighter covenants, more equity required, or simply a decision not to do the loan at all. That is why the middle market can look capital-starved even when there is plenty of capital interested in the underlying exposure. The financing machinery was built for much larger transactions. The interesting question, then, is whether the cost of creating a loan can itself follow a technology cost curve. Do enough transactions and you stop treating each loan as a handcrafted project. You start building infrastructure. And once you can systematically reduce the cost of every hop, the minimum economically viable loan size falls, borrower pricing improves, and much larger pools of capital become accessible to the middle market. That is where USD.AI has been focused. We are one of the few operators actually doing this at scale. We have roughly $500M of loans on our books by end of the next month and are on track to cross $1B by year-end- not one big billion loan but a lot of smaller <100m deals But that $1B is not one giant transaction. It is many dozens of loans. And that matters. Every loan is another training example. Every underwriting process teaches you what can be automated. Every origination teaches you where information gets lost. Every legal structure teaches you what can be standardized. Every settlement teaches you where latency can be removed. Over time, more of that process moves from artisanal execution into software, agents, standardized workflows, and tech-enabled origination. That is the moat. We found product-market fit around the demand for compute credit. The next question is how you turn execution into infrastructure, and infrastructure into a platform. You only get there by actually doing the loans, and having the reinforcement training runs on these extremely detailed processes Getting your hands dirty, seeing where the process breaks, encoding those lessons into the system, and then doing the next loan better than the last. Our first loan took 120 days, the latest took 21. That is how the cost curve comes down. And that is how a fragmented middle market becomes scalable. We have a few major product updates in pretty excited to get out there that lowers the per unit cost of loan economics for not just us, but for our capital partners as well.

3/ Live with RWAs issued by @paretocredit, @3f_xyz, @MidasRWA with more issuers to follow. Every market carries its own risk-adjusted exposure cap, so no single issuer can dominate the vault. The vault also targets ≥50% withdrawable liquidity at all times. docs.kpk.io/vaults/vaults/mo…