Link claims to passages
Let readers trace a proposition to a precise passage, see whose position it expresses, and judge whether the source supports the interpretation.
About DerridAI
It began with a corpus
When an interpretation hinges on a few lines, readers should be able to find them. That principle shaped DerridAI from its beginnings as a multilingual study of Jacques Derrida.
Project history
The starting point
A corpus for close reading
The early corpus preserved details such as edition, page, and translator so researchers could locate a passage. It also tracked quotation and stance to help them interpret it in context.
Beverly and Cambridge · 0.80.x
Reviewable metadata
The project moved to a common representation for field assertions. That made it possible to record who or what proposed a value, which evidence supports it, and how a reviewer evaluated it.
September 2026 · cELF 1.0
A specification others can implement
cELF describes how to preserve source locations, record revisions, assertions, and evidence across research workflows. DerridAI is its originating reference implementation.
Learn about cELFFall River and Gloucester · 0.81–0.82
A broader research environment
Later releases brought together evidence-aware review, configurable pipelines, and portable research-site publication. Researchers could inspect more of the work between source ingestion and a cited answer.
Founder
Aaron John Schlosser, PhD
Researcher, software engineer, and author of cELF
Schlosser studied comparative literature at Dartmouth College and earned a doctorate in French language and literature at Yale University. His software career has included engineering in healthcare analytics and technical leadership on voting systems.
Schlosser founded DerridAI and wrote cELF, the specification for recording source passages, metadata assertions, and review decisions.
Research goals
Test local models
Compare local models on comparable questions. Measure response time alongside citation accuracy, attribution, coverage, and interpretive value.
Keep data and inference local
Keep models and data under the researcher’s control. Use reproducible checks for source identity and citations, and make model interpretations available for review.