First print settles (DECISIONS.md #7). No historical value is ever restated.
Every deviation between vintages, and every methodology error discovered
after publication, gets an entry here and is folded into the NEXT epoch.
Entries are append-only; editing or deleting an entry is forbidden.
Format:
## C-NNNN | <iso-date> | <scope>
- What was wrong, with the wrong value as printed
- What is right, with evidence (source URL + hash, or derivation)
- Forward treatment: which epoch absorbs it, and how
---
C-0001 | 2026-08-16 | pre-genesis (no prints affected)
The early napkin estimate of organic issuance was g = +2.9%/yr. It omitted the aging spend term and misused GBD YLL.
Corrected by open recomputation to g = +0.7197%/yr from the identity dS/dt = B·e(0) − N + N·dĒ/dt (mint +9.6606B, spend −8.0917B, drift +1.0394B). Recorded in DECISIONS.md (Key numbers).
Forward treatment: none needed — predates all prints. Logged for lineage because the openness rule caught it, and that is the point of the rule.
The ORIGINAL tly_v0_calc.py computed births through a float intermediate (line 144 per A-16 ruling D3): full-precision births 132,113,756.251091 x e0 73.123374469 -> mint 9,660,603,670.8547 (9.6606B). The reconstruction, working from 4-dp printed values and possibly revised OWID data, landed at 9.6603B — the famous -0.0026% residual is input-precision/vintage noise, not method.
Forward treatment: recorded as a v0 defect; the v1 engine's float quarantine (test_no_float_in_published_path, adopted per D3c) makes the defect class unrepresentable. No published value restated.