S15 E25: Nicholas Gregory & Tom Trevethan on MercuryLayer, Blinded Statechains & Lightning

Nicholas Gregory keeps coming back to Bitcoin Takeover with a new version of the same obsession: how do you move whole Bitcoin UTXOs around without begging the base layer for a receipt every time? In Season 11 he walked me through MercuryWallet and statechains. In Season 13 he returned for mobile plans and Lightning experiments. Season 15 Episode 25 is the rewrite. This time CommerceBlock CTO Tom Trevethan sits next to him, and the product is no longer marketed as a rival CoinJoin wallet. It is MercuryLayer: a blinded Bitcoin multiparty-computation Layer Two that treats statechains as infrastructure other people can build on.

The short pitch is almost too clean. Ruben Somsen’s original statechains idea needed something like ANYPREVOUT to transfer ownership of a UTXO off-chain. CommerceBlock spent years shipping a version that worked without that soft fork. Then Taproot and Schnorr arrived, they deleted a mountain of wallet code, and they rebuilt the service so the operator is a blind co-signer. Instant transfers. Zero marginal fee per hop. Whole coins only. Unilateral exit if the server vanishes. That is the episode.

From MercuryWallet to MercuryLayer

Nicholas is honest about the first act. MercuryWallet was supposed to be an alternative privacy wallet next to Wasabi and Samourai: statechain CoinSwaps, desktop app, real volumes, real traction, not enough mindshare. “If we’re honest,” he says, “it didn’t really take off.”

What changed was not the marketing deck. It was cryptography. In the old design the swaps were trustless between users, but the server still saw Bitcoin addresses. Addresses can be censored. With Schnorr and a blinded signing protocol, Tom explains, the server does not see addresses at all. It does not even know it is dealing with Bitcoin. That forced a rewrite: smaller codebase, generic Layer Two, open API, example clients in Rust and JavaScript instead of one monolithic Electron monster where UI logic and protocol logic had married each other and refused counseling.

They will still ship a desktop and mobile example wallet. They do not want that wallet to be the product. They want other wallets and protocols to call the library. Testnet was already up when we recorded. Mainnet was “a few weeks” away on their March 2024 timeline. Builders were invited to complain early, before ossification.

What Counts as Layer Two (According to CommerceBlock)

I open the usual can of worms. Is Liquid a Layer Two? Is an ETF? Is Coinbase?

Nicholas’s test is practical: can you retrieve your funds without coordination if the other party disappears? Statechains give you a backup transaction. Lightning does too. Federated peg-outs do not. So in his framing, Liquid and Rootstock are useful tools, not Layer Twos in the unilateral-exit sense. He is not anti-sidechain. CommerceBlock used to build sidechains. Spanners, hammers, screwdrivers. Different jobs.

Tom sharpens the contrast with Lightning. People get religious about the L2 label, but the useful property is unilateral exit. Lightning and MercuryLayer both have it (Ark would too, if it ships). Lightning’s hard problem is liquidity: channels, routes, inbound capacity. Statechains skip that entirely. You send a whole UTXO to anyone. No channel graph. The trade-off is the opposite pain: you cannot pay an arbitrary coffee amount. You move whole coins. Fixed denominations. That is why the first Mercury product leaned into privacy swaps, and why the new stack still loves equal-amount CoinSwaps even while refusing to run a coordinator business.

Sidechains Store Ledgers. Statechains Move Coins.

Tom’s sidechain critique is blunt. Federated sidechains like Liquid or Rootstock are custodial multisigs wearing a ledger costume. The pegged bitcoin sits behind federation keys. Mercury has a trust assumption (the operator must delete old key shares and follow the protocol), but users keep a unilateral on-chain exit, and blinding makes the operator a much thinner legal and surveillance target than a federation that can see every peg.

Scalability is the other knife. A sidechain is another blockchain. Everyone verifies everything. Popular Liquid becomes Bitcoin’s fee market with a different logo. On a statechain, you only verify the history of the UTXO you care about. Client-side validation. Linear coin history. That is the Ruben Somsen pitch, still alive after the rewrite.

Drivechains come up because they always do. Tom is not existential about them. Another blockchain secured by a miner peg is a different animal from a blinded co-signer with no chain state. He does not see Drivechains as an extinction event for Bitcoin. He also does not expect the soft fork politics to be kind. BitVM workarounds get a polite good luck. Nicholas’s broader line: stop ossifying out of fear. Let Paul Sztorc build. Let the market decide. (Yes, Paul sent a LayerTwo Labs bit during the recording. The man never sleeps.)

Blind Signing, Access Tokens, and Who Pays the Bills

Privacy is still in the DNA, even if they refuse to market MercuryLayer as a privacy product. In MercuryWallet, the server knew which coins swapped, so you needed many rounds to get anonymity from the coordinator. In MercuryLayer, one CoinSwap against a blind server is already private from the operator’s point of view. They want Nostr-style discovery for swaps. They will not run the coordinator. Legal clarity in the UK matters to them. Laws can change. They know that. They still prefer building something they can operate in London over romantic offshore theater.

About seventy percent of inbound interest, Nicholas says on air, is Ordinals people. He is not a JPEG maximalist. He also will not filter them. Ordinals are individual UTXOs that hate paying mainchain fees to move. Blind statechain transfers and atomic swaps let them hop without a miner invoice every time. If that volume funds research into Lightning bridges and Fedimint experiments, CommerceBlock will take the money. High-fee environments make strange bedfellows.

The business model shifted with the blinding. Old MercuryWallet took roughly 0.5% on withdrawal via an extra output. A blind server cannot enforce that. The new plan: an access token for a blinded signature session, on the order of $50 or €50 for a year, paid over Lightning. Each use of the server burns eight hours of that session window. Deposit, transfer, exit. The operator remains a co-signer with almost no stored state. Two or three database tables. Signing checks it cannot read.

Lightning Latch, Fedimints, and the Gold Bar Metaphor

Chat asks about Lightning Latch. Tom lights up. It is a brand-new (as of the recording) atomic swap between a Mercury statecoin and a Lightning payment, built around hold invoices and a preimage that only releases when both sides have verified. Privacy purchase of a coin. Ordinal sale for an arbitrary Lightning amount. Non-custodial both ways, or neither side completes. They coined “latch” because it sounds like a door lock and because submarine swaps already took the cool nautical name.

Fedimints get the Uncle Jim treatment. Nicholas likes small quorums that share a UTXO or a Lightning node, maybe optimized with MercuryLayer. He does not want ten giant mints waiting to get nuked by regulators. Army of mosquitoes, not three whales. Tom agrees custody is what scales payments today, and ecash is the private flavor of that custody, but he is nervous about KYC gravity at size. Scale self-sovereignty enough to keep the system honest. Do not fantasize that eight billion people will run their own nodes.

Nicholas’s meetup metaphor for normies: virtual OpenDimes. Gold bars. Chunks of value that move off-chain with no settlement fee and no liquidity tax. Not coffee. Settlement. Bridges between order books, federations, and services that need to shove serious weight around when mainchain fees go stupid.

Compared with Ark, he piles on: no soft fork required, live now, no liquidity locked in the server. CTV would help and remove constraints. Soft forks may never come. Build anyway.

Wasabi, Tor Trauma, and Covenant Religion

Wasabi sponsors the episode, so we do the friendship dance. They are not competitors. Nothing Much reviewed early Mercury work and helped with Tor pain. Nicholas’s product lesson from MercuryWallet: people who have never shipped a Tor-using wallet do not understand how cursed Tor UX is. Wasabi hid that complexity well. He wishes more privacy wallets also spoke i2p, because one anonymity network is a DoS waiting to happen. He even floats, half serious, that Max Hillebrand might one day run a CoinSwap coordinator beside CoinJoin. High fees make strange roadmaps.

On covenants, Tom is tired of the holy war. OP_CAT and CTV both look analyzed enough. Activating useful opcodes used to be boring. Now it is religious. He prefers simple base opcodes (CAT’s generality appeals) and does not mind CTV either. Nicholas wishes Bitcoin had something like Simplicity so we stop holding Code War every two years. He disliked how Jeremy Rubin was treated over CTV. Smart people need incentives to build. Then he undercuts the magic-bullet crowd: even if CTV lands this year, usable software takes years. Multisig UX is still bad. CTV UX will be worse before it is better. Lightning was sold as the Final Solution. It was a wedge. Covenants will be another wedge.

Where Builders Should Look

I share the MercuryLayer site on stream. It looks like engineers wrote it, because engineers wrote it. Protocol steps, math, no “number go up” hero banner. Nicholas built it with a Rust templating engine for fun. Cute explorers can come later. For now: docs, Telegram for DMs, short explainer clips on Twitter, and a builder culture they openly envy from Nostr (fiatjaf is a fan; CivKit is in the same CommerceBlock neighborhood).

Five-year vision, if it works: settlement rails and surprising use cases under fee pressure, not a retail coffee network. Ordinals as the weird retail edge case. Bridges between Layer Twos that do not all speak Lightning fluently. Instant, zero-marginal-cost transfer of a trunk of bitcoin to anyone, with a backup transaction in your pocket.

I close the same way I opened. They took a cool wallet that the market shrugged at, tore out the parts that made it a closed app, and turned the statechain into a blinded Layer Two other people can plug into. That is the CommerceBlock move in 2024: less shrine, more socket.

Listen to the audio on Bitcoin Takeover, or watch the YouTube upload. For more on why statechain CoinSwaps still matter next to CoinJoins, see my later essay on the limitations of Bitcoin privacy.