# Reputation is not the same as safety *Published 22 July 2026*  Silicon Valley Bank had a reputation, right up to the week it collapsed. In DeFi, a handful of trusted names price the risk for billions in loans. Gauntlet, Chaos Labs, the firms that charge protocols like Aave and Compound millions to answer one question: how risky is this asset, really? We take their word for it, largely because of who they are. Tomy spent six years building in that world, working alongside those firms, and came away thinking the model itself is the weak point. When he came on one of our calls to walk through what he and his co-founder are building on Bittensor, he didn't start with a product. He started with that problem. Worth understanding, whether or not you ever touch what they're building. ## **The problem he kept watching** Tomy's argument isn't that Gauntlet or Chaos Labs are bad at their jobs. It's that the whole arrangement is fragile. You have a few big, reputation-heavy analysts running private models, and everyone downstream trusts the reputation. Then the reputation fails. He reels off the examples: Silicon Valley Bank. 2008. And, as he puts it, in Argentina every single month. Different scales, same shape. A system trusted a few names to understand the risk, and the risk turned up somewhere the names weren't looking. His fix is simple, almost obvious once you hear it. **Stop trusting, start competing.** Instead of paying one famous firm and hoping they're right, you make risk analysts compete in the open and let the best analysis win. "Come and prove it," as he put it to the incumbents. That competition is the actual product. On Bittensor, the miners on his subnet are the risk analysts, scored against each other on how well they price and analyse risk. The network's output becomes **something you can check, rather than someone you have to trust.** ## **Root is Bittensor's risk-free rate** Stake TAO to root and you earn the network's baseline yield. In the same way a US Treasury bill is the risk-free rate the rest of finance prices against, root is Bittensor's version. So anything that asks you to move your TAO has to pay you more than root, or why bother. That's the whole basis for a money market. You price everything as root plus a premium. One caveat, though: root is "risk-free" only in TAO terms. It still carries TAO's price risk. A good mental model, not a literal Treasury bill. Hold that, and the rest falls into place. ## **Two things wearing one name** Here's where it's easy to get confused, and where the design is clever. Endure is two things. Underneath is Forge, a money market: deposit TAO to earn yield, or lock your alpha as collateral and borrow TAO against it. It does what an Aave does, and it's the part you can touch today. On top is the subnet, the competitive risk analysis. Forge is deliberately framed as just its first customer, a live proving ground. A slice of Forge's revenue buys back the subnet's token, which is meant to grow and sharpen the pool of miners doing the analysis. The application funds the intelligence layer that makes the application safer. The reason for keeping them apart is ambition. A lending protocol on Bittensor is capped to the size of the network. Reliable risk analysis is something you could, in theory, sell to the rest of DeFi and one day to banks and brokers. That last part is a vision, not a roadmap, and Tomy is honest that it's a long way off. But it's why the subnet exists at all, rather than just the protocol. ## **The bit that will matter most to alpha holders** If you hold alpha or run a subnet, this is the design to understand. On Forge, you lock your alpha as collateral and borrow TAO against it, and you keep the yield and the upside of that alpha the whole time you're borrowed. Selling caps both. This is a way to fund what you're doing without giving up your position. And they deliberately won't let you do the reverse. You can't borrow alpha against TAO, because that would be a way to short, and they'd rather stand with the subnet owners than hand people a tool to bet against them. Small choice, clear values. ## **What careful building looks like** The thing I'd point to most is how they're building, not what they're claiming. The parameters are conservative on purpose. Loan-to-value capped at 30% at launch, caps on how much can be deposited and borrowed, a liquidation buffer sized to survive a crash rather than assume one won't come. Tomy's stated north star is plain: **no bad debt in the protocol.** On the token side, they'd turned their own emissions off, choosing to wait until the mining and validation mechanics were robust and hard to game rather than switch them on soft. Since then the network has taken the decision out of anyone's hands: the protocol was updated so that a subnet without a live incentive mechanism and active miners earns no emissions at all. It's early. Forge is on testnet with mock tokens, the security audit hadn't started at the time of the call, and the competitive risk engine, the actual differentiator, isn't switched on yet. Right now it's a carefully-built money market with the clever part still to come. And the safety record he's proud of, over a billion in past TVL without an exploit, is his own account. To his credit, he told the room not to take even that on trust: go and check it yourself, he said, I could be saying something that isn't true. That posture, assert it and then tell people to verify you, is rare enough to be worth naming. ## **Why it matters** Credit is one of the oldest pieces of financial infrastructure there is, and it boosts economies. It was a little bit crazy to Tomy that Bittensor still didn't have a money market. Now it's getting one, built by people who seem more worried about bad debt than about being first. Whether the bigger bet pays off, competitive risk analysis good enough to sell beyond Bittensor, is unknown. It's the hard part, and it's barely begun. But the question underneath it is a good one to sit with. We've spent years trusting the biggest names to tell us what's safe. What would it look like to make them prove it instead? --- https://endure.network https://x.com/EndureNet