# Democritus causal claims from language

- **Registry ID:** `catagi-democritus`
- **Book:** Categories for AGI
- **Documentation status:** `public-code-linked`
- **Experimental status:** document extraction and benchmark comparisons
- **Canonical documentation URL:** https://categorical-ai.sridharmahadevan.com/experiments/catagi-democritus

## Purpose

Causal-claim extraction, Geometric Transformer manifold rendering, and UniCausal comparison.

## Book location

- Chapter 16: Causal Claims from Language
- Chapter 22: Homotopy in Language and Causal Inference

## Documented studies

- Four-domain causal-claim extraction audit
- Geometric Transformer versus UMAP manifold ablation
- UniCausal extractor comparison
- Homotopy-aware claim normalization audit

## Associated code packages

- `democritus-openai` — [Democritus causal-claim extraction](https://github.com/sridharmahadevan/Democritus_OpenAI/tree/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9); **public**; not specified. Document-to-causal-claim pipeline, including the Geometric Transformer manifold branch used by the current Democritus account.

## Start here

These concrete files are selected from the complete resolved source surface. They orient the reader; they are not a claim that every family is independently reproducible.

- **implementation · public-pinned** — [`democritus-openai:pipelines/pipeline_llm.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/pipelines/pipeline_llm.py). Open the pinned public source file.
- **implementation · public-pinned** — [`democritus-openai:scripts/manifold_builder.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/scripts/manifold_builder.py). Open the pinned public source file.
- **implementation · public-pinned** — [`democritus-openai:scripts/visualize_manifold.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/scripts/visualize_manifold.py). Open the pinned public source file.

## Entry points

- `pipeline_llmdriver.py`

## Result artifacts

- `saved extraction outputs and GT/UMAP ablations`

## Resolved code surfaces (3)

- **implementation** — [`democritus-openai:pipelines/pipeline_llm.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/pipelines/pipeline_llm.py)
- **implementation** — [`democritus-openai:scripts/manifold_builder.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/scripts/manifold_builder.py)
- **implementation** — [`democritus-openai:scripts/visualize_manifold.py`](https://github.com/sridharmahadevan/Democritus_OpenAI/blob/0fb1f8ba2f67f2906de5a3435c464751fe20c6d9/scripts/visualize_manifold.py)

## Frozen run records (0)

- No frozen run-level record is registered. Consult the result-artifact list and evidence boundary above for the exact surviving evidence; absence of a run record is not evidence that no experiment was run.

## Evidence boundary

Extracted claims and rendered manifolds do not certify causal truth or identification.

A code or artifact link establishes traceability. It does not by itself establish independent reproduction, statistical adequacy, correctness, or support for a claim beyond this boundary.