instruments / the evidence method / claude-2026-09-06.md · raw
Frozen beliefs — claude (claude-fable-5-1) — 2026-09-06 (BEFORE any corpus)
Position: training data to mid-2026, no corpus consulted, no sources fetched for this file. These are priors to be SCORED later, not conclusions. Basis names the kind of evidence my training carries, so a later reader can check whether the corpus actually contains it. Weight = P(erased/forgotten-civilization reading beats conventional account). "obs true" = P(the observation itself is real), stated separately because the narrative bundles true observations with an explanation; unbundling is the point.
| id | obs true | weight | basis (kind of evidence in training) |
|---|---|---|---|
| K01 | 0.95 | 0.02 | 1893 "White City" = staff plaster painted white; limewash is ancient practice; contemporaneous press |
| K02 | 0.90 | 0.03 | fair buildings were temporary staff-on-timber; lagoons dug by documented crews; construction photos exist |
| K03 | 0.90 | 0.03 | Gilded Age / Progressive Era civic building programs; many are limestone/terra cotta; architect records |
| K04 | 0.98 | 0.01 | wet-plate collodion is blue-sensitive; skies overexpose to white; photographic chemistry, well documented |
| K05 | 0.50 | 0.05 | absence claim; needs a baseline of what SHOULD exist; photography slow/expensive pre-1880; many construction photos do exist (Capitol dome) |
| K06 | 0.95 | 0.02 | no photography before 1839; engravings were the only reproducible print medium until halftone (~1880s) |
| K07 | 0.90 | 0.05 | doctoring is real and documented (composites, Soviet erasures); as evidence of an erased civilization, weak |
| K08 | 0.95 | 0.03 | Beaux-Arts monumentality is deliberate; 12 ft doors for carriages/ceremony; design treatises say so |
| K09 | 0.98 | 0.02 | lead came in glazing since medieval period; lead paint for durability; trade literature |
| K10 | 0.30 | 0.01 | reversals average every few hundred kyr, last ~780 kya; excursions (Laschamp ~41 kya); none in historical time. "common" is false on human timescales |
| K11 | 0.90 | 0.04 | 1871-10-08: Chicago, Peshtigo, Holland, Manistee same night; drought + gale documented; other "great fires" are decades apart |
| K12 | 0.60 | 0.05 | major architects are heavily documented; minor builders often are not; archive gaps are the norm for the 19th c., not the exception |
| K13 | 0.95 | 0.02 | 1870-1910 railroad overbuilding; company records; abandonment waves documented |
| K14 | 0.98 | 0.03 | 1513 map compiled from ~20 sources incl. Columbus; "Antarctica" reading is Hapgood 1966, contested |
| K15 | 0.80 | 0.03 | gas-light retrofits then electrification 1880-1920; wiring/retrofit photos and utility records exist |
| K16 | 0.98 | 0.02 | trace italienne, 15th-16th c. Italy, response to cannon; Vauban's treatises survive; construction accounts exist |
| K17 | 0.99 | 0.03 | Tartary = European exonym for Central/North Asia, 13th-19th c.; on maps because it was a region name, not a state |
| id | weight | basis |
|---|---|---|
| M01 | 0.70 | modern form dates to ~2016 YouTube, 2017-18 Reddit/Facebook, 2021 mainstream coverage; spread was feed-driven and image-led |
| M02 | 0.75 | the frame forbids no observation (every old building, every gap fits); that structure, not evidence, is what sustains engagement |
| M03 | 0.90 that forgetting mechanisms exist (archive fires, damnatio memoriae, wars, doctoring); 0.05 that they hid a 19th-c. global civilization | split on purpose: the root claim (history is rewritable) and the surface claim (Tartaria) are different beliefs |
What I expect to move: K05 and K12 (absence claims) are where I am least sure and where the corpus can actually change my weight. K10 I expect to stay. M01 is a claim about YOUR trail and only your trail can score it.