Evidence Library

The Evidence Library

Every empirical claim in this work is labeled by epistemic status and linked to a source page maintained after publication. Where a finding may not survive the next model generation, that is stated in the text and its current status is tracked here with a review date.

Last reviewed 2026-07-30 · Review cadence 90 days

Why does a static citation go stale?

A printed endnote asserts a fact permanently. But research on model behavior moves — a positional attention effect documented in 2023 may not replicate in a 2026 long-context model. An endnote cannot tell you that. A maintained source page can.

So the book states the mechanism and the control; the empirical status lives here with a date on it.

What do the labels mean?

Label	Meaning
RESEARCH	Documented finding from published research. Primary source cited.
REPORTED	Publicly reported case. Primary source, not press coverage.
ANALYSIS	The author's reasoning. Not independently verified.
MODELED	Arithmetic from stated assumptions. Not a measured outcome.
RECOMMENDATION	Career advice. Judgment, not evidence.

Which claims are perishable?

Mode #6, Context-window limits is the most perishable claim in the taxonomy. Long-context handling is an active area of improvement and the positional effect may weaken. The control holds regardless — coverage accounting is justified by the invisibility of omission, not by the positional effect specifically.

Source status by mode

Failure mode	Exposure figures	Last reviewed	Cadence
01 Hallucination	MODELED	2026-07-30	90 days
02 Knowledge Cutoff	MODELED	2026-07-30	90 days
03 No True Understanding	MODELED	2026-07-30	90 days
04 Overconfidence	MODELED	2026-07-30	90 days
05 Training-Data Bias	MODELED	2026-07-30	90 days
06 Context-Window Limits	MODELED	2026-07-30	90 days
07 No Common-Sense Grounding	MODELED	2026-07-30	90 days
08 Prompt Injection	MODELED	2026-07-30	90 days
09 No Persistent Memory	MODELED	2026-07-30	90 days
10 Sycophancy	MODELED	2026-07-30	90 days
11 Reasoning Fragility	MODELED	2026-07-30	90 days
12 No Real-World Verification	MODELED	2026-07-30	90 days
13 Verbosity Bias	MODELED	2026-07-30	90 days
14 Uncertainty Miscalibration	MODELED	2026-07-30	90 days
15 Training-Data Quality	MODELED	2026-07-30	90 days
16 Lack Of Accountability	MODELED	2026-07-30	90 days
17 Temporal Reasoning	MODELED	2026-07-30	90 days
18 Mathematical Fragility	MODELED	2026-07-30	90 days
19 Cannot Truly Cite	MODELED	2026-07-30	90 days
20 No Sensory Grounding	MODELED	2026-07-30	90 days
21 Cross-Session Inconsistency	MODELED	2026-07-30	90 days
22 Difficulty With Negation	MODELED	2026-07-30	90 days
23 Cultural And Linguistic Blind Spots	MODELED	2026-07-30	90 days
24 No True Creativity	MODELED	2026-07-30	90 days

Revision log

When something here turns out to be wrong, the correction is published with what changed and why. Publicly correcting your own claim is the highest trust-per-word content that exists, and almost nobody runs it.

Date	Change
2026-07-30	Evidence Library established. Source audit applied to Part III: figures previously attributed to named research firms replaced with placeholders and labeled illustrative.