Thirty million impressions on a résumé line.
Evan Hubinger now runs alignment stress-testing at Anthropic. Before that, he was a software engineer at Ripple, writing code into rippled — the server software that runs the XRP Ledger — and into Interledger, a payment-layer protocol built to move value between arbitrary ledgers. Then he pivoted to the problem of whether machine intelligence ends the species. His own number: better than a 10% chance inside a decade. That figure traveled thirty million times.
Crypto did what crypto does with an uncomfortable fact. It made a bit. Galaxy Research's Alex Thorn cracked the obvious joke — the AI doomer used to work at Ripple — and the XRP community laughed along, because the alternative was to look at the chart.
So let's look at the chart.
XRP has never reclaimed its January 2018 high of roughly $3.40. Not in the 2021 cycle. Not in the 2024 ETF cycle. Not now, in a sideways tape where the marginal dollar rotates between memecoins and Treasuries and nothing in between gets bid. Meanwhile Ripple — the company, not the ledger — is reportedly seeking a raise at a valuation of up to $50 billion, and that number is mostly collateralized by the XRP sitting on Ripple's own balance sheet.
That is the datapoint. Not the résumé line. And before anyone trades on it, note that the original item carries a timestamp reading 2026 — either a forward-dated test artifact or a typo. A source with a broken clock doesn't get to make time-sensitive claims, so treat everything below as structural observation, not as a catalyst.
Context: what he actually built, and what he actually said.
rippled is the reference implementation of the XRP Ledger, in production for over a decade — older than most of the people currently posting about it. Interledger is the more interesting artifact. ILP is an open routing standard for moving value across heterogeneous ledgers and currencies, on the premise that you don't need one settlement layer, you need a routing protocol plus connectors holding liquidity on both legs.
Hubinger's own phrasing described it as a "trustless system for cross-currency transactions between arbitrary agents." That's a simplification, and I want to flag it as one. I've spent enough years in cross-chain architecture to know the gap between a protocol spec and a deployment. ILP's trustlessness is inherited from the connectors, not guaranteed by the standard — a connector that custodies both legs is a bridge with a different label on it. Positioned against Cosmos IBC or Polkadot's XCMP, ILP was an early explorer of the interoperability thesis. It did not become the industry standard, and that's a fact worth sitting with before anyone credits a résumé line as technical validation.
There's also a governance migration the source material skips: Interledger is now stewarded by the independent Interledger Foundation, not by Ripple. That matters more than any career footnote, because it's the difference between a protocol with a single corporate owner and one with a distributed maintainer set.
— Root: Auditing the DAO and Ethereum
Core: the coupling nobody prices.
Here is the structural problem with a $50 billion ask.
A company whose equity value is predominantly backed by a token it issues has built a reflexive loop.
XRP price falls → balance-sheet collateral falls → fundraising capacity falls → ecosystem spend falls → developer and liquidity outflow → XRP price falls.
That isn't a risk factor appended to the model. That is the load-bearing wall of the model.
I've watched this architecture fail before. In May 2022 I restructured into stables and BTC weeks ahead of the Terra unwind, because the peg mechanism was legible as a death spiral from the mint function alone. You didn't need to know when. You needed to know that the reserve math had no floor. Ripple's loop is slower and it has a real product underneath, which is precisely why nobody wants to model it. Slow reflexive loops are the ones that grind retail into powder over years instead of days, and they never show up in a 30-day drawdown chart.
What the source doesn't disclose is the part that would let anyone price this honestly: how much XRP Ripple actually holds, whether those holdings are locked, and what the escrow release schedule looks like. Historically Ripple released a billion XRP per month from escrow and returned a large share of it to the next tranche — a cadence that functions as a standing supply ceiling. Without those three numbers, any valuation model of Ripple equity is a guess wearing a spreadsheet.
Now layer the talent on top. Three names surface in the material, and all three have left operational posts. Hubinger, engineer to Anthropic alignment lead. David Schwartz, CTO emeritus, now a public commenter whose stated position is that AI safety rules could threaten free speech — a libertarian read entirely consistent with the community he came from. Emi Yoshikawa, former vice president, gone from the org chart.
One career path proves nothing about an engineering bench. Three senior departures from a project older than most Layer 1s is a pattern, not an anecdote. It could be healthy metabolism — companies mature, people rotate, the org chart settles. It could be the leading edge of something worse. What I can say is that the direction of travel matches a broader migration: capital and talent moving out of crypto infrastructure and toward AI, where funding is unconstrained and the narrative hasn't been farmed to death yet.
There's a second-order read here that the community missed entirely. Hubinger participating in rippled's build process as a student suggests a period when Ripple's core client was open enough for low-friction contribution. Low barriers to contribution are a gift in year two and a governance liability in year ten. Protocols don't die from bad code. They die from the slow drift of who is still willing to maintain the code.
— Root: Auditing the DAO and Ethereum
Contrarian: the joke is the print.
Everyone read the Thorn quip as trivia. I read it as sentiment data.
When a principal at a research firm spends social capital making a joke about your ecosystem's seven-year price ceiling, you are not being attacked. You are being archived. Narrative decay shows up as comedy before it shows up in price — the asset has stopped being a threat and started being a reference point. I've watched that tonal shift precede the slow bleed in several 2017-era infrastructure tokens. Nobody yells at a dead narrative. They needle it.
And here's the trap in this whole cycle: the "AI × Crypto crossover" framing is being manufactured by media, not by protocol teams. A crypto outlet covering an alignment researcher because he once touched rippled is attention arbitrage, not a product signal. I've been on the wrong side of this shape before. In 2020 the entire DeFi sector convinced itself that "liquidity fragmentation" was the industry's central technical problem. It was never a technical problem. It was a fundraising hook with a whitepaper stapled to it, designed to let VCs underwrite the next round of the same product. Every cycle has one. Liquidity fragmentation. Modularity. Restaking. Now AI alignment.
We farmed the yields until the protocol farmed us.
Takeaway.
This is a cultural item, not a signal. Tagging it "blockchain" is a taxonomy error, and anyone folding it into an investment research pipeline is doing themselves harm.
What I'm tracking instead, in priority order: the actual documentation on Ripple's raise, because if it lands at $50 billion with XRP as the collateral, that's a balance-sheet story and not a token story. The escrow release cadence, because that is the only honest supply number in this whole discussion. And whether "crypto → AI" hiring becomes a pipeline or a fad — if it's a pipeline, the next sector narrative is already being staffed while everyone else is still arguing about a joke.
$3.40 is an anchor, not a target. The real question isn't whether XRP gets back there. It's whether Ripple's valuation can ever structurally re-couple to the token's market cap — or whether the loop only runs one direction.
The AI doomer left. The token stayed. Which one is priced correctly?
— Root: Auditing the DAO and Ethereum
