Vaughan Pratt

Vaughan Pratt is a computer scientist at Stanford University who, in 1995, published "Anatomy of the Pentium Bug" — one of the first rigorous, independent technical analyses of Intel's Pentium floating-point division defect, done alongside Tim Coe of Vitesse Semiconductor, who reverse-engineered the chip's faulty quotient-digit lookup table.

In this vault Pratt is the primary source underneath the entire Pentium FDIV cluster: both the mechanism claim (claim-pentium-fdiv-bug-five-missing-srt-lookup-table-entries) and the discovery-story claim (claim-nicely-twin-prime-project-surfaced-pentium-fdiv-bug) rest on his 1995 report, and a 2026-08-26 verification pass corrected one detail these notes had drawn from him incorrectly — a useful reminder that even a Tier 1 primary needs a direct re-read before its exact wording is trusted secondhand.

Updates

• 2026-08-27: A direct 32-page read of the Intel White Paper Pratt cites as [SB94] confirms he accurately reported Intel's script-error attribution (though not in Intel's exact words) and, separately, that his own "wonder whether the rule for lowering thresholds was applied to the wrong boundary" is nowhere in Intel's document — his alternative hypothesis was his own speculation, not something Intel's account left open or hedged. (claim-pentium-fdiv-bug-five-missing-srt-lookup-table-entries, claim-pratts-threshold-lowering-hypothesis-absent-from-intels-fdiv-white-paper)

References

• claim-pentium-fdiv-bug-five-missing-srt-lookup-table-entries — his own reverse-engineering of the five missing lookup-table entries, and his reporting (without full endorsement) of Intel's script-error explanation.
• claim-nicely-twin-prime-project-surfaced-pentium-fdiv-bug — his account of how Thomas Nicely's twin-prime project surfaced the bug publicly.
• claim-pratts-threshold-lowering-hypothesis-absent-from-intels-fdiv-white-paper — his competing hypothesis, confirmed absent from Intel's own account.