Blog
Tell It What You Want, Not How to Do It
Written by Louise Beattie · 9 August 2026
From a Beyond Finance community call, in conversation with T-Slice and Stalker of Minotaur, Subnet 112.
Interacting with DeFi hasn't changed much since 2019. You still click through every step yourself, approve this, swap that, bridge the other, as if the point of the software were to make you do the work. Minotaur is one of the very few DeFi-leaning subnets on Bittensor, and its answer to that is a different way in, called an intent.
An intent is a declaration of the outcome. Not "route this token through these three pools," but "earn me 5% on my USD stablecoins." You say what you want, and the system works out how.
"It's natural language. They want to say what they want, and then how it's achieved is - I mean, they just want the best way for it to be achieved. Like, don't ask me how to achieve it. Like, I just want the best way to achieve it." - T-Slice
On the call I put it as hey Siri, play this tune, but for DeFi. Stalker's version is better: you want to watch a film and you don't care whether it's on Netflix or Disney, you just name the film and the system finds the best and cheapest route to it. The simplest intent is a straight swap, and CowSwap was the first protocol to do it that way. The ambition runs from limit orders to yield optimisation, all declared the same way.
Apps are how Minotaur learns to do new things
Stalker's analogy is the one to keep - an app is to Minotaur what a smart contract is to a blockchain. You deploy one to give the network something new it can do, and the miners then race to solve it, mapping the best execution routes in heavily benchmarked rounds. A new app becomes usable inside about an hour.
There's one app so far, a DEX aggregator live on Base and Ethereum. Starting with swaps is deliberate: get transaction efficiency to state-of-the-art first, and then it makes sense to route everything else through the same system. @tsliceAI says the aggregator reached 99.6% as good as the best solver out there within weeks. That's his number, measured on Minotaur's own page against five other providers, on a product he's candid is still beta and low volume. Publishing the comparison at all is more than most subnets do. It's still self-measured.
Why does this need a subnet?
It's one of the standard questions we ask on these calls.
The first is the validators.
"We needed a distributed model and we needed a security model on top of blockchain execution. So we needed some kind of quorum to be available. And that's what - that's where validators come in and we have them agree on what's happening and we have them validate that an order can be executed." - Stalker
They supply a security quorum without Minotaur having to run a blockchain of its own. They score an order, they agree on the score, and only then does it execute.
The second is the miners, and this is the part that makes Minotaur different from a company with a good app. They never touch your money. They contribute code to Minotaur's open-source solver, which bootstraps the hard work of finding good execution routes, and it's what T-Slice credits for getting to near-parity so quickly. Without it, he says, they'd be doing all that development themselves, all the time.
One trade-off to be clear about: Minotaur runs that solver centrally. T-Slice's position is that having every validator run it makes the system distributed and decentralised anyway, and that the open-source code is moated because it's designed to run through a subnet.
Both answers hold on their own terms. The subnet buys them a security layer and a development engine they'd otherwise have to build and fund themselves.
The safest thing an agent can do is nothing
You've seen the posts. "I woke up and my agent made me $40,000 trading overnight." Nobody posts the morning the agent hallucinated an address and sent the money nowhere. Agents forget things, they forget who they are, and handing one a live wallet takes a particular kind of nerve.
Every app carries a scoring rubric, and if an agent asks for something the scoring doesn't accept, the order reverts. It doesn't even happen. T-Slice doesn't oversell what that buys you:
"We want to express that it adds safety and security, but to say that it is, uh, 100%, like no one, no one really wants to say that about any piece of software because it's a hard claim to make, but it's definitely safer." - T-Slice
So the risk hasn't gone, it's moved. The whole thing is only as safe as the scoring behind each app, which he's the first to say. The question you're left holding isn't whether your money is safe, it's whether the rulebook is any good.
There's a quieter benefit underneath. An agent doing raw transactions has to read revert logs, diagnose failures and retry, all of which clogs its context and wears it down over a long run. Calling an intent is naming the parameters and getting clean feedback when the order goes through or doesn't. Less to hold in mind, less to get wrong.
The money flows the other way
The protocol is gasless and the fees are deliberately small, roughly $0.10 per one-shot order on Base and gas plus a margin on Ethereum, with perpetual orders priced higher because Minotaur is choosing the timing. Small on purpose - charge more and you invite someone to undercut you.
What's different is where the money comes from.
"Since we are collecting our fees on different chains, it means that we are not an extraction-based subnet... we are taking the USDC on Ethereum, USDC on Base, we are bridging that, buying TAO, and reinvesting basically in the subnet." - Stalker
The revenue is earned on Ethereum and Base, from users who never touch Bittensor and don't need to, then bridged in to buy TAO and put back into the subnet. Money flows in rather than out. The stated ambition is to be the first revenue-positive sustainable subnet, with all profit converted into buying the token - not a "buyback", T-Slice is careful to say, because they didn't issue the tokens the chain did.
They aren't near it yet, and he says so plainly: there's revenue, it's small, the volume is low, they're a beta.
The model still implies something past Minotaur itself. Most subnets run on emissions, earning a share of the token the network prints. One that funds itself from fees on other chains is importing outside capital into Bittensor rather than drawing on its own, and if it holds, sustainable stops meaning survives on emissions and starts meaning pays its own way.
Pulling liquidity in, not sending it out
Bittensor EVM is the third chain, and T-Slice's framing of it is the most expansive thing he said all call - the portal to all the liquidity, the way to pull money into Bittensor rather than push it out. Everyone else, on his read, is solving the liquidity problem by sending liquidity outside.
Then cross-chain intents. Stalker put the gap plainly:
"...why I actually have to go on Ethereum and purchase some wrapped TAO instead of just purchasing the alpha of the subnet that I want to. Why wouldn't I be able to do that?" - Stalker
Under cross-chain intents you'd trigger the intent from Ethereum, Minotaur would bridge the money and buy the alpha, and a perpetual order would send the accumulating emissions back to your Ethereum wallet. Then payments, so your intent could be to pay someone in TAO as easily as you'd swap a token.
The interesting part isn't the aggregator. It's the premise underneath it: that you should be able to name the outcome and stop there.