Read and retrieve

Agent access

The explorer is a static collection. Every report has a readable HTML page, a complete Markdown copy, and a JSON record with the full proposition and exact claim text.

Entry points

• data/index.json — report IDs, titles, dates, scope labels, probabilities, and paths to each format.
• llms.txt — a plain-text guide and direct report links.
• Probability method in Markdown — definitions, conditioning, materiality, and scope exceptions.
• Coverage in Markdown — source notes on the included mechanisms and their boundaries.

Reading a record

Resolve all paths in JSON relative to the explorer’s root directory, not the data/ directory. Follow source_markdown for the full source, or structured_data for the extracted record. HTML pages also declare both formats with rel="alternate" links.

claims contains exact source wording in claim and claim_markdown, a concise reading aid in question, the three conditional judgments in percent, and the cumulative base product. The question does not replace the exact claim. causal_proposition_markdown supplies the complete scope and definitions. sections preserves the report’s detailed reasoning, evidence, hypothetical resolution experiments where present, and qualifications.

source_headline_markdown and source_cumulative_display retain the author’s displayed rounding. The numeric probability_percent values are computed from the claim table. The website’s short titles and summaries are navigation aids; use the full proposition for scientific interpretation.

Probability semantics

Each conditional judgment after the first assumes all preceding claims are true in the same mechanism. Multiply the conditional values once, dividing percentages by 100, to obtain joint belief. Do not multiply cumulative values again. “Favorable” is the common field name for the sources’ favorable or optimistic scenario.

The probabilities express subjective belief that the defined causal proposition is true. They are not patient risks, attributable fractions, or treatment success rates. Scenario products are sensitivity ranges, not confidence intervals. Mechanisms overlap, and their probabilities must not be added or treated as independent.

The default proposition concerns material structural loss in human peptidergic C-fibers in early type 2 diabetes. Population, timing, endpoint, and intervention definitions in each report take precedence over the short scope label. Read the complete method for the default thresholds.

Snapshot and citation

Research is ongoing. These are copied files, not live references to a research directory. Each record distinguishes the snapshot date from the review date stated in its source. The research README is excluded. Source Markdown retains its original text; links to internal notes outside this collection are not bundled. No research prompts or goals are included.

Use a report’s page permalink and snapshot date when citing a particular assessment. The citation below credits the collection; cite the original papers for their results.