Receipts
Every claim, with a prior and an outcome.
The opposite of plausible deniability. Brier score lands once enough predictions resolve; sealed time-capsule claims unseal in 2031.
Brier score
—
No resolved predictions yet
Brier = mean (prior − outcome)². 0.25 = always-50% (clueless); <0.20 = providing real signal beyond chance.
Calibration
predicted prob → realized rate
- 0–10%n=0
- 10–25%n=0
- 25–50%n=0
- 50–75%n=0
- 75–100%n=0
Calibration plot lights up once predictions resolve. Zero is the honest starting point.
Pending
3US software-engineering employment will fall ≥10% YoY in any quarter between 2026-Q4 and 2027-Q4.
Made May 6, 2026Resolves Dec 31, 2027Prior 35%PendingThresholdBLS QCEW reports a YoY drop ≥10% in NAICS 5415 in any quarter ending 2027-Q4 or earlier.
By end of 2026, ≥3 frontier LLMs will offer ≥1M-token context windows in production (not preview).
Made May 6, 2026Resolves Dec 31, 2026Prior 70%PendingThresholdPublic pricing pages show 1M+ context for at least 3 of {OpenAI, Anthropic, Google, Meta, xAI, DeepSeek}.
Resolved
None yet.
Sealed time-capsule
10 predictions sealed until May 6, 2031. Claims, priors, and thresholds were committed in May 2026 and will be unsealed and graded on the date above.