A World Bank aid paper estimates a bitcoin leak. The viral line is stronger than the estimate.

What is verified

• The post is from @cremieuxrecueil on 1 October 2026. It attributes a 2-to-6-cent figure to researchers. It does not name the paper in the text that was retrieved.
• NBER working paper 35655, “Crypto Capture of Foreign Aid”, studies World Bank disbursements from 2018 to 2024. The text describes an implied leakage of about 2 to 6 cents per aid dollar at a two-to-five hop forensic convention, and about $1.7 to $4.4 billion across the tranche arrivals in the estimate.
• The Financial Times reported the same range on 24 August 2026 and noted that the estimate is bitcoin-only. Stablecoins are outside it.

What the replies added

The thread’s most useful move is the author’s reply, which supplies the NBER link the original post left out. The most common error in the other replies is the unit. Cents became percent. A 2 percent leak and a 2 cent leak are not the same finding.

• @cremieuxrecueil · 202 likes

  The author’s own reply names the source: https://www.nber.org/papers/w35655

  Reply on X

• @nic_carter · 422 likes

  Anyone in development finance already knows the problem is getting funds past diversion by corrupt officials.

  Reply on X

• @loganb · 7 likes

  Asks whether some of the local activity is just people transacting in crypto, or whether the authors have evidence of fraud.

  Reply on X

• @m_goes_distance · 15 likes

  Calls it a sequel to a 2020 offshore-banking paper. The reply says 7.5 percent. The figure reported with this paper was about 7.5 cents.

  Reply on X

• @0xtechnexus · 13 likes

  Rejects siphoning and says people in high-risk places want control of their own money.

  Reply on X

The sentence that travelled

On 1 October 2026 @cremieuxrecueil posted that researchers had found a spike in bitcoin activity when foreign aid goes out, and that “somewhere between 2 and 6 cents of every dollar of World Bank foreign aid disbursements got siphoned off into crypto wallets.” On this snapshot the post had 1,199 reposts, 142 quotes, 129 replies, 10,070 likes and about 269,000 views. The chart image is the argument for most readers. The paper is not linked in the text that was retrieved.

The range is real. The verb is the poster’s. “Siphoned off” is courtroom language. The research is an estimate built from timing and from chains of hops. Those are not the same object, and a repost does not promote the second into the first.

What the working paper says

NBER working paper 35655 looks at World Bank aid from 2018 through 2024. The authors describe $238 billion of disbursements across 93 countries in the estimation sample. They line up on-chain bitcoin, the creation of new wallets, off-chain exchange records and IP-linked web traffic with the month a tranche arrives. The pattern they report is a short surge: anonymous volume on tax-haven exchanges jumps, new anonymous wallets appear, and the move fades within a month or two. Identified wallets, the ones tied to a known owner, move much less.

The 2-to-6-cent figure is an implied leakage, not a wire they watched leave a project account. In the paper’s own wording, exchange ledgers do not reveal a turnover multiplier, so they map hops to a net figure. At the two-to-five hops that forensic practice often treats as typical, they get roughly 2 to 6 cents per aid dollar. On the tranche arrivals used for that ratio, they put the dollar total at about $1.7 to $4.4 billion. Change the hop count and the cents change. That sentence is the one a viral post leaves out.

What the estimate cannot see

The Financial Times, writing on 24 August 2026, put the same range next to an earlier World Bank figure of about 7.5 cents per aid dollar from offshore banking data. The bitcoin estimate came in a little lower. The authors’ caution, which the FT repeated, is the one that matters for anyone building in this market: the study is bitcoin. Stablecoins are a growing laundering rail and are not in the numerator. A reader who concludes “the leak is 2 to 6 cents, full stop” has added a precision the method refused.

There is a second limit. A disbursement-timed surge is evidence of coincidence with a mechanism, not a named official and a signature. The sectors where the paper says the pattern shows up — transport, water and sanitation, social protection, governance — still receive later funding. The paper treats that as part of the result: capture, in their data, did not cut the next cheque. That is an argument about incentives in aid, not a blockchain feature request.

Why this belongs on a builder’s desk

The useful technical point is narrower than the headline. Public chains make a class of flow visible that correspondent banking hid. They do not make the person at the end of the flow visible. Mixers, fresh wallets and tax-haven venues are the layering step the paper describes, and they are the same shapes a compliance team already has language for: placement, layering, integration. Bitcoin did not invent that sequence. It gave a research design a timestamp.

If you are designing a payout, a donor report, or a public-sector wallet, the lesson is not “bitcoin is dirty” and it is not “the chain will prove the aid arrived.” A wallet address is not a beneficiary. An on-chain spike is not a court finding. A system that pays people in the open still needs an identity, a purpose for the payment, and a way to explain a hop. The Canon’s point stands: a ledger records what it was told to record.

How to quote it without lying

A careful sentence is: a 2026 NBER working paper estimates that, under a stated hop convention, World Bank disbursements from 2018 to 2024 line up with bitcoin activity consistent with an implied diversion of about 2 to 6 cents per aid dollar. A careless sentence is: researchers proved that six cents of every aid dollar was stolen into bitcoin. The post is closer to the second. The paper is the first.

This page is not a comment on any government, and it is not advice to hold or sell bitcoin. The post is here: https://x.com/cremieuxrecueil/status/2105707121073328552. The working paper is NBER 35655.

Related on this site

• Blockchain Canon
• Digital asset stack
• Pulse disclosures

Other Pulse articles

• Pi’s Open USD note is a partnership claim. The stablecoin itself was launched by someone else.
• The SEC put a crypto custody proposal on the table. It is not a permission slip.
• A crypto tax bill is real. The “trillions” line on the viral post is not part of it.
• Coinbase and Citi are wiring stablecoins to a bank checkout. The merchant does not have to hold the token.

Context

• Aug 2026. NBER working paper 35655 circulates. The Financial Times writes it up on 24 August.
• 1 Oct 2026. The X post restates the range and draws 1,199 reposts in this snapshot.