• October 2026
  This site

  Redesigned this site.

• October 2026
  New projects

  Designed three new projects: The sportsbook on campus, Tuition in yuan and Lost in translation. More.

• October 2026
  The $150,000 cliff

  Found that two SBA programs capped loans at the cutoff ($150,000), which can fake part of the result. Excluding their 127,841 loans made the migration test far cleaner. More.

• October 2026
  The time machine problem

  Made the earnings-per-share check allow for cent rounding: failures fell from 26.6% to 5.2%. The rest look like a units problem. More.

• October 2026
  The time machine problem

  First results: 4.9 million point-in-time SEC numbers for 315 companies, an identity audit, and a memory test on eight dated models.

• October 2026
  The $150,000 cliff

  First results on 1.69 million SBA 7(a) loans: the pile at the cutoff, a replication of a published estimate, and no measurable difference in risk at the line.

• October 2026
  LedgerGym

  First results: simulator, fifteen planted problems and scorer built; a rules baseline handles 96.5% of reconciliation unattended under 1% error.

Next

• LedgerGym. Run AI agents, and an AI-plus-rules hybrid, through all 80 simulated months, and test whether an instruction hidden in an invoice can make an agent change a vendor's bank account.
• The time machine problem. Ask a commercial model to forecast earnings with the company named and masked, before and after its training cutoff.
• The time machine problem. Trace the 952 earnings-per-share failures that survive the rounding allowance back to the filings.
• The $150,000 cliff. Estimate how strictly each lender screens, and run the stress test on every active loan.
• New projects. Start The sportsbook on campus with a sourced calendar of state launch dates.