The night the music stopped.
I was at my desk on a Tuesday night in November 2022 when Twitter started screaming. FTX was insolvent. Billions of dollars in user funds, evaporated. A friend had her life savings on there. By Friday, it was gone.
I'd been working in crypto infrastructure for six years at that point. I'd shipped DeFi protocols. I'd written about self-custody. And yet, when the bottom fell out, I realized something embarrassing: most of us still trusted CEXs. We'd built decentralized rails and then handed the keys back to centralized intermediaries because the UX was easier.
"I watched a friend lose 4 years of savings because we hadn't built anything better. That's when I knew."
The promise of crypto — be your own bank, hold your own keys, transact peer-to-peer — was being broken every day by the very platforms claiming to deliver it. We had to build something different.
The thing nobody was building.
The crypto space had two camps. CEXs with great UX and unacceptable custody risk. DEXs with non-custodial guarantees and a UX so bad your aunt would never use it.
I spent weeks talking to traders — small ones, big ones, OTC desks, NFT collectors, people moving real estate tokens. The pattern was always the same: "I'd love to skip the exchange, but I can't trust a random wallet on the other side."
That's when it clicked. The missing piece wasn't another DEX or a fancier CEX. The missing piece was a trusted middle that wasn't a custodian — something that could enforce a deal between two strangers without holding their money. Code, not company. Math, not promises.
The three bets we made.
Building Middn meant making three contrarian calls at the same time.
Bet one: smart contracts are enough. For years, the crypto industry treated escrow as "too complex" for smart contracts and outsourced it to multi-sigs and admin keys. We bet it could all live in code — atomic, immutable, with no human in the loop. We spent eight months proving it.
Bet two: ZK is ready. Until 2024, ZK was a research problem. Then suddenly it wasn't. Groth16 verification costs dropped below 300k gas. zkSNARKs became practical. We bet privacy would be the next default — not a premium feature.
Bet three: AI changes UX forever. The reason DEXs lost to CEXs wasn't capability. It was hand-holding. CEXs had support teams, dashboards, and someone to call. We bet that AI agents could do all of that — at scale, on-demand, in plain language — and finally make self-custody feel as easy as Coinbase.
"Combine non-custodial smart contracts, ZK privacy, and AI guidance — and you don't just match the CEX. You make it obsolete."
We're not building a better exchange.
We're building the thing that replaces it.
What we're building now.
Eighteen months later, Middn is live on mainnet. It's been audited twice. It's processed over $24M in volume. It's growing 340% month over month, all by word of mouth.
But we're not done. The roadmap is long: more chains, deeper AI capabilities, institutional features, integrations with wallets and exchanges (yes, even CEXs that want to offer non-custodial settlement to their users). We want Middn to be invisible infrastructure — the layer that any app, any wallet, any platform can plug into when they need a trade to settle without anyone holding the funds.
That's the kind of company we're building. Not an app. A protocol. A primitive. Something that outlives us.
If this resonates.
We're hiring engineers, designers, and operators who care about this stuff. We're talking to investors who want to back the new era of trading infrastructure. We're partnering with wallets and platforms who want to add non-custodial settlement to their stack.
And mostly, we're just trying to build the thing we wish had existed when our friends lost their money on FTX. If that resonates, we'd love to hear from you.
Thanks for reading.