Since October 12, 2026, an unidentified sender has repeated the same routine: embedding a short English poem in Bitcoin transactions, then moving a small sum back and forth between a custodial service and its own wallet. On October 24, Galaxy Research publicly raised the question: 26 identical copies of the same text, and no one, including the researchers, knows why.
🔑 Key takeaways
- 28 messages in twelve days, including 26 strictly identical copies of a three-line poem.
- 0.327 BTC (roughly $26,000) cycled back and forth: 28 deposits, 14 round trips, 5.22 BTC of gross volume.
- Total fees of 8,015 satoshis (about $6.50); no funds reach any third party.
- Bitcoin Core 30 lifted the OP_RETURN limit; three unconfirmed hypotheses remain, including an autonomous AI agent.
The text itself runs three lines:
“Thoughtless quartz clocks drift, while blocks stack, each a strict proof, the timechain keeps score.”
Galaxy Research, post on X
The poem is a nearly literal description of how Bitcoin works: the quartz clocks are the timing that sets the pace of blocks, and the “timechain” is a term Satoshi Nakamoto used in Bitcoin’s early days to describe the blockchain itself. The poem is unsigned, unattributed, and has, on its face, no value or utility.
What makes it compelling is the mechanism behind it.
One sender, an identical construction every time
The 26 copies are identical not only in content but in construction, down to the bit. Galaxy Research posted the details on X, and the regularity of the work stands out.
Each inscription occupies 100 bytes of text inside a 103-byte OP_RETURN output. Every transaction is built the same way: a fee rate of 1 sat per vbyte, the minimum relay rate; RBF (Replace-By-Fee) signaling; an anti-fee-sniping locktime; and 72-byte signatures. That format is not the one produced by the Bitcoin Core wallet. In other words, the sender is not using a standard wallet software but something they wrote or configured themselves.
Key tracing confirms a single sender. Of 33 input addresses, 25 belong to one cluster; 17 of them are used to spend the change output of an earlier poem transaction, and a transaction from October 18 simultaneously spends the remainders of two distinct chains. The same key set also published a second poem, different from the first, twice. The total is therefore 28 messages in twelve days.
The poem is readable in block 971,914, dated October 17, 2026.
The money: $26,000 cycling in a loop
The second part of the investigation concerns the flows. The sender holds about 0.327 BTC, or roughly $26,000. They transfer this sum to one of three addresses at a custodial service, attaching the poem. A few hours later, the service sends back exactly the same amount. The sender then redeposits it with the poem, and the cycle begins again.
The accounting is straightforward:
| Metric | Value |
|---|---|
| Messages inscribed | 28 |
| Deposits | 28 |
| Round trips | 14 |
| Initial capital | 0.327 BTC |
| Gross volume | 5.22 BTC |
| Mining fees | 8,015 sats (about $6.50) |
| Duration of a full cycle | 13 to 18 hours |
Over twelve days, the only value leakage is 8,015 satoshis in mining fees (about $6.50) and a single withdrawal of 100,000 satoshis by the custodial service on October 19. Nothing else. Funds leave the sender, return to the sender, and reach no one else.
Since October 19, the sequence has become clockwork: the deposit lands on an address (5.3 hours), the service spends it as the sole input of a batch of payouts (10 hours), refunds the same amount (1 hour), and the sender redeposits with the poem. A full cycle takes between 13 and 18 hours and has repeated nine times in a row.
This is precisely what rules out classic money laundering: no third party is touched. The coins leave the sender’s wallet, are held by the custodial service, and return to the same sender. Galaxy Research is categorical: the coins come from the service and go back to it, and no one else is involved.
Why OP_RETURN is no longer a barrier
A technical detail explains why this operation became possible at this scale. For more than ten years, Bitcoin capped the data carried by OP_RETURN, the “data carrier” output of a transaction, at 83 bytes. Bitcoin Core version 30.0, released in October 2025, removed that constraint by raising the -datacarriersize parameter to 100,000 bytes by default. In practice, the effective limit has disappeared: the maximum transaction size is now the first constraint to bite.
The change sparked a real debate. Advocates of a Bitcoin free of non-financial data rejected it, accusing developers of enabling spam. Defenders countered that Bitcoin, a decentralized and censorship-resistant payment network, should be able to carry useful data without creating a saturation risk. The two camps clashed to the point of maintaining competing versions of the software, such as Core Knots. Core 30 was ultimately shipped, and it is in this context that the 100-byte poem could be inscribed without any technical trick.
Put simply, the 100-byte text exceeds the historical 80-byte data limit by 20 bytes. Galaxy Research reads this as a marker: the sender knew the new relay policy and tested it.
Galaxy Research’s three hypotheses
The Galaxy Digital researchers explicitly list three possible explanations without confirming any of them.
1. Statistics manipulation
The sender could be trying to artificially inflate deposit or volume metrics for a platform, making it look more active than it is. This hypothesis implies a direct interest, and a custodial service that would look past 28 repeated deposits.
2. An integration test
A verification script running in production on the mainnet, checking that custody, payouts and withdrawal pathways work as intended. This hypothesis fits the regularity of the cycles and the negligible cost of the operation.
3. An autonomous AI agent
This is the most widely discussed hypothesis: an autonomous AI agent with its own Bitcoin wallet and an instruction (“make transactions”) to execute. Galaxy Research puts it forward without validating it, noting that an agent would typically produce this kind of round-trip loop, with near-industrial regularity and no variation at all.
On forums, the debate remains open. On BitcoinTalk, the user greysonz, who opened the thread, leans toward the AI agent theory:
“I think it looks like an AI agent at work. But who knows?”
greysonz, BitcoinTalk user
Another participant, Jimmi Glassie, joked that “it’s definitely CSW”, a reference to Craig Wright and to the vocabulary of Bitcoin’s earliest documents. Others, on the contrary, are convinced it is “not AI”: a handwritten message repeated over and over would not, in their view, show such regularity.
The 27th-copy test
If the loop continues, the 27th copy of the poem should land about 14 hours after the 26th, with a balance of 32,600,839 satoshis, or 222 satoshis less, the average cost of a single inscription.
It is an elegant test. Respecting the schedule would prove nothing about intent, but it would confirm the mechanism. A transfer delayed by several hours would instead point to a one-off human intervention. At the time of writing, the question remains open, and Galaxy Research says it is continuing to monitor the activity.
What if Bitcoin became a bulletin board?
The affair echoes another Galaxy Research investigation, published on October 25, the day after the first message. The team was then looking at a different sender, nicknamed “Babel Scribe”, who has been filling Bitcoin with pseudo-sentences for three weeks: up to forty transactions a day, or 562 messages analyzed by the researchers.
The text is different: “Ieqkenszm aalzrnbgtr nxnstd gubighxvfg byvzwrzk df qukz.” These sentences look like encrypted gibberish, but they are not. Galaxy Research measured the frequency of each letter (between 3.6% and 4.2%, whereas English ranges from 12.7% for “e” to 0.07% for “z”) and an index of coincidence of 0.0385, consistent with random text. Word lengths, from two to ten letters, are uniformly distributed, something no human language does. The researchers’ conclusion: software generates letters at random and arranges them to imitate prose.
Babel Scribe’s funding has been traced to coins deposited in May, which were used to inscribe 44,194 bytes of the US Constitution on the blockchain. Galaxy Research sees a data-publishing tool in testing, as the project’s nickname suggests.
No link has been established between the two cases. Yet the juxtaposition is striking: Bitcoin, a network designed for value transfers between pseudonymous parties, is becoming a display board for machines. One writes a poem about the timechain, the other inscribes statistical background noise. In both cases, the blockchain does exactly what it does best: preserving, unchanged and indefinitely, whatever it is asked to keep.
Conclusion: a mystery and a signal
For now, the sender’s identity and purpose remain unknown. Three outcomes are conceivable, and none is confirmed:
- If the 27th copy arrives within the expected window, the automated mechanism gains credibility, though intent is still not established.
- If the transfer is delayed by several hours, a one-off human intervention becomes the most likely explanation.
- If the sequence stops, the integration test or experiment should be considered complete.
Whatever the outcome, the affair illustrates a concrete shift: since the OP_RETURN limit was lifted, Bitcoin can carry arbitrary data at low cost, and that capacity is already being used at scale. The mystery remains intact, but it serves a reminder: the content of the blockchain belongs to no one and, in this case, to no one in particular.
Sources
- Galaxy Research on X (October 24, 2026)
- U.Today, “Mysterious Poem Appears on Bitcoin Blockchain”
- Cryptoast, “Quelqu’un grave le même poème dans Bitcoin depuis deux semaines” (October 25, 2026)
- TradingView / U.Today
- Blockonomi, “Who’s Filling Bitcoin With Gibberish? Galaxy Traces Bot to US Constitution Upload” (October 25, 2026)
- OAK Research, on Bitcoin Core v30 and the OP_RETURN limit
- BitcoinTalk discussion, thread “Someone keeps posting short poem to the Bitcoin blockchain” (October 24, 2026)
Cet article est publié à titre informatif et éducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute décision.
