One of Bitcoin’s most prolific application-layer builders just walked away from Lightning. In a second sit-down with the XXIM podcast, Fiatjaf said the quiet part out loud: Lightning doesn’t work in practice, doesn’t work in theory, and eight years of work on it was a waste that should stop.
This is not a random critic. Fiatjaf built LNURL, the LN txbot, LNbits, wallet webhooks, and contributed to the CLN codebase. He spent 2019 to 2023 inside Lightning’s internals, including a stint forking the SBW wallet under ZEBEDEE sponsorship. His exit matters because the people who built the thing are now the ones saying it failed.
The theoretical breaking point
Fiatjaf’s deepest critique is one most Lightning users have never heard. Every Lightning payment depends on the preimage, the proof of payment that the receiver generates and the sender must receive back. But when fees are high, small HTLCs are not actually created on chain. The protocol fakes them. It pretends the HTLC exists, because creating it would cost more than the payment is worth.
“You’re not going to get the pre-image because there’s no actual HTLC published through the chain,”
He says. And without that preimage, which he calls
“the most important thing in the protocol,”
you can lose your channel trying to recover a payment that was never really secured.
This collides with Bitcoin’s own design.
“Everybody assumed that fees would be high and the fees being high, that means that you cannot send small payments.”
Fees must stay high to secure Bitcoin with small blocks, and high fees make small Lightning payments economically meaningless. The entire system only functions if everyone in the route behaves honestly. At that point, you have trusted intermediaries with extra steps.
Even the Lightning Network inventors Joseph Poon and Tadge Dryja pointed out limitations and openly discussed its limits: it’s not a silver bullet, doesn’t solve everything, can be overhyped, and with small blocks it may mainly serve richer users/companies.
Dryja stepped back from active development amid disagreements, and Poon moved on to other work. They always treated it as an incomplete piece of the puzzle with real trade-offs. But they did not see it as a failure
Structural decay?
The numbers tell the rest. Node count fell from roughly 20,700 in 2022 to about 6,000 today. Channels collapsed to roughly a quarter of their peak. The hobbyist node-runners, the uncle-Jim operators Fiatjaf once championed with LNbits, quit. Even the most tech-savvy users decided running a node and managing channels was a burden bigger than they could carry.
What remains is consolidation.
“Lightning was just centralizing more and more in a few big hubs, like a few big hubs of nodes that have a bunch of channels,”
He says. The volume growth Lightning’s defenders cite comes mostly from exchanges, and the data itself is estimated from big-node telemetry:
“they estimated based on some of the big nodes of the network, so it could actually indicate that the network is centralized on this big node so they see more of the total activity.”
If your scaling solution only works for custodial corporations, it did not scale Bitcoin.
Then there is the basic math nobody in Lightning ever refuted. Roger Ver made the argument years ago, and Fiatjaf endorses it: if every Bitcoin block were used exclusively for channel opens, and every person on earth opened exactly one channel in their lifetime, the chain still could not onboard the world. The original mission was numerically impossible from day one.
Is Lightning Innovation Dead?
Perhaps the most damning section of the interview is the list of things that never shipped.
“They had all these plans of implementing taproot channels, which I think LND has implemented, but there’s no spec and no one has implemented.”
BOLT12, which was supposed to kill LNURL:
“after many, many years of specing, they have decided on a spec and people have implemented, but no one is using it as far as I know. It’s not a good idea. I never liked it.”
Eltoo, renamed Lightning Symmetry, was implemented twice and means nothing without a soft fork that still hasn’t happened, the same sighash/noinput proposal from the original 2015 Lightning paper.
In 2022, Reddit threads were already calling Lightning ossified. Fiatjaf agrees. The protocol bogged down in edge cases and theoretical attacks like channel jamming and flood-and-loot that nobody ever executed.
“The solutions to these attacks, they just over-complicate the protocols, very super complicated… It’s, it’s incredible.”
Complexity became a barrier to entry that killed the small implementations, leaving only the well-funded ones.
The wrong analogies in Lightning
Lightning’s defenders now call it Bitcoin’s Swift, a settlement layer. Fiatjaf rejects both halves. Swift is a messaging layer, not a settlement layer; settlement happens separately, netted out overnight or weekly.
“lightning is a settlement layer, but I don’t think that argument holds either because there’s no settlement layer that has to settle on every transaction.”
Send one satoshi and you risk paying thousands of satoshis in fees to safely close the channel that carries it.
“Just close everything”
“There are so many bright minds out there wasting their brains in lightning. Or maybe they all died now after wasting the entire brain’s trying to fix lighting that is unfixable.”
The host’s final question: if the entire Bitcoin and Lightning ecosystem came to you and asked you to fix Lightning, what would you do?
“Well, no, fix lightning. No, no, just close everything.”
His prescription is drivechains.
“There are so many bright minds out there wasting their brains in lightning. Or maybe they all died now after wasting the entire brain’s trying to fix lighting that is unfixable.”
Applied to drivechains, that talent could produce a family of scaling solutions with better properties than Lightning ever had. He points to the glass coins design (formerly shielded CSV) as an example of something that could scale far better as a drivechain layer.
He is not alone in drifting away.
“Rusty Russell announced on Nostr that he stopped working on CLN. Anton, the Ukrainian guy, he left in complete disgust and he hates lightning right now.”
What survives and what can evolve
Fiatjaf is not nihilistic about the work.
“The invoice is very good.”
BOLT11 solved real UX problems, the clipboard-replacement attacks and exact-amount errors that plagued on-chain payments. He still likes the elegance of channels that never close, and the noble effort of getting multiple implementations to interoperate through real specs. But elegance is not a scaling strategy.
His closing line is worth sitting with.
“There was never an original goal. There was never original mission.”
Thaddeus Dryja wanted payment channels between exchanges and was skeptical of multi-hop. Joseph Poon oversold it as scaling for everyone and left for Ethereum. The mission everyone thought they were building was never written down, and the numbers never added up anyway.
Bitcoin still needs to scale. Fiatjaf’s answer:
“the wrong design and it distracted people for years and years, how many years, eight years, nine years,”
And the sooner builders accept that, the sooner they can build the one that works.
Source: XXIM podcast, “I Fell Out Of Love With Bitcoin Lightning,” published Sep 24, 2026. https://youtu.be/2PD9UsU_jY4
