In a research post published on October 16, 2026, Anthropic states that an internal Claude research model traced the seeds of the NIST P-192, P-224, P-256, P-384 and P-521 curves to output of the C library's lrand48 generator seeded with Unix timestamps between May 1997 and April 1998, finding them in under half an hour, and that the post's published code reproduces those seeds exactly.
✓ DELIVERED · settled 2026-10-22
| Claim made in | Claude reveals ECDSA curve origins (anthropic.com) |
| Bank record | Anthropic — see its full delivery record in the bank |
| Market opened | 2026-10-18 |
| Market type | Flash market (1–3 days) |
| Calls lock | 2026-10-20 08:00 Eastern Time (12:00 UTC) — 24 hours before the check, so nobody can call it with the answer already visible |
| Check time | 2026-10-21 08:00 Eastern Time (12:00 UTC) |
| Source of record | anthropic.com |
Background
For twenty-five years, nobody could say where the seed values behind NIST's elliptic curves came from. The NSA supplied them without explanation when ECDSA was standardised in 2000, and the gap fed a durable suspicion: that the curves, P-256 among them, might have been picked to hide a weakness. The account that survived was second-hand. A longtime NSA employee, Jerry Solinas, recalled in 2015 that the seeds were probably the hash of a joke phrase along the lines of “Give Bob and Jerry a raise”. Solinas died in 2023, no archive turned up, and a public bounty for the seeds' origin went unclaimed. Anthropic says an employee put the question to an internal research model in October 2026, without passing on the passphrase tip, and that in under half an hour Claude had the answer. The seeds are not hashes of anything. They are raw output of lrand48, the random-number generator from 1990s C libraries, started from Unix timestamps between May 1997 and April 1998. Claude spotted the generator's fingerprint — the top bit clear in every 32-bit word of four seeds, a pattern with about one chance in 5,810 of appearing by accident — then recovered the hidden state bits, using lattice reduction for the prime curves, and wound each generator back to a clock reading. P-256's seed, for example, was candidate 1,842 in a run started on August 16, 1997. The post publishes code that regenerates fourteen seeds exactly, traces six base points to the generator in Arjen Lenstra's FreeLIP package, and is explicit about what it does not settle: one binary-curve seed, B-163, and the base points of P-224, P-256, P-384 and P-521 remain unexplained, as does the reason some workable candidate curves were passed over. Anthropic's reading is that a seed pulled from a clock leaves little room for a planted weakness, and it stresses that this is not a break of ECDSA. The $36,864 bounty goes to the Internet Security Research Group.
This market exists because launch announcements are easy to make and hard to grade. Press releases describe what a company intends to ship; they rarely come back later to say whether it shipped. AI Rank Bank pulls one checkable sentence out of the coverage — the sentence above — gives it a date, and lets readers price it in public. The launch itself stays with its publisher; what lives here is the claim, the crowd's call, and eventually the receipt.
How this claim is judged
Judging criteria, fixed when the market opened: Delivered if, at the check time, Anthropic's post remains published with that account — lrand48 output seeded from 1997–1998 Unix timestamps, the five named prime curves, the under-half-an-hour finding, and reproduction code whose output matches the published seeds — and the account has not been withdrawn or materially corrected. The post having been retracted, or the seeds shown not to reproduce from the stated timestamps and candidate indexes, does not count.
The source of record is the publisher's own page linked above. A claim about a company's product is judged against that company's public output — its product pages, documentation, changelogs and official posts — not against rumors or third-party summaries. Only yes-or-no outcomes count: partial credit is not a verdict, and a claim that needs a paragraph of excuses has already answered itself.
How readers called it
2 readers called this claim. 50% said it would deliver.
1 Will deliver · 1 Won't ship · final split when the market locked
Points reward being right against the crowd. A correct call pays 100 × (1 − the share of readers who agreed with you when you placed it), with a floor of 10: side with a 90% consensus and a win pays 10, stand alone and a win pays up to 100. Wrong calls pay nothing. The price is locked in when the call is placed, so early readers are paid for judging before the crowd forms, not for copying it.
How readers split
Share of readers saying “will deliver”: opened at 0%, now at 50%. 48 snapshots on record since 2026-10-18 — a snapshot lands whenever a call is placed, and at least hourly while the market is open.
Verdict
✓ DELIVERED — checked against the source of record and settled on 2026-10-22. The evidence is public: https://www.anthropic.com/research/ecdsa-curve-origins. Readers placed 2 calls on this claim; 50% of them said it would deliver. The market paid out 100 points in total. Verdicts are permanent. If the evidence ever shows this stamp was wrong, the record is corrected openly and the correction stays on this page.
About this market
AI Rank Bank is an AI news front page with a memory. Every launch on the homepage carries one claim like this one, with a check date. Readers call claims with points, build a public record, and climb the leaderboard. Companies build a record too: every settled claim lands on the company's bank page, where its delivery rate is simple arithmetic that nobody gets to edit afterwards. There is no money in the market — points buy bragging rights only, which keeps the incentives pointed at being right.