Inspiration
Campus Evidence Lab began because as a high-school student, I have faced antisemitism on an almost daily frequency, I started the project because accountability in these incidents is something that is lacking and for real change, holding the facilitators and actors accountable is key.
What it does
CEL is public evidence infrastructure for civil-rights accountability in higher education. It transforms fragmented public information into structured, source-linked records that journalists, researchers, advocates, institutions, and affected communities can inspect directly. Users can: Search campus civil-rights records by institution, issue, source, and community. Open institution-specific evidence pages. Trace claims back to primary sources and precise document locations. Generate citation-ready reporting packets. Download structured JSON and CSV datasets. Inspect limitations, institutional responses, corrections, and review status. Receive timely CEL Signals connecting current developments with relevant public evidence. CEL does not rank universities, calculate safety scores, or turn allegations into conclusions. Its purpose is to make the underlying public record easier to find, verify, and responsibly use.
How we built it
I built CEL with Codex as a development partner, translating the project’s evidence standards into working software at a pace that would not otherwise have been possible. We began by defining strict rules for what a canonical record must contain: a resolved institution, a supported claim, a public source, a precise locator, explicit limitations, and language that does not exceed the evidence. From there, we built pipelines that: Normalize records from inconsistent government and institutional sources. Resolve institution names, aliases, campuses, and duplicate entries. Preserve exact provenance, including document sections and workbook cells. Separate aggregate statistics, allegations, official findings, and institutional responses. Reject records with ambiguous identities, unsupported claims, or privacy risks. Generate searchable pages, institution dossiers, APIs, feeds, datasets, and reporting packets. Create reproducible snapshots and ProofGraphs so records can be independently checked. Propagate corrections and preserve a public history of changes. Codex helped turn methodological decisions into data models, validation rules, adversarial tests, public interfaces, and production deployments. The result is not simply a database—it is a reproducible evidence system with claim boundaries built into its architecture.
Challenges we ran into
The source material was often inconsistent, duplicated, poorly formatted, or locked inside large government workbooks. Institution names collided, campuses changed names, and seemingly simple statistics required exact scope and year boundaries. Scaling introduced another challenge: increasing the archive from thousands to 10,000 records without weakening its standards. Every new record needed reproducible provenance, safe language, identity resolution, deduplication, and downstream verification.
Accomplishments that we're proud of
CEL received an Emergent Ventures grant, supporting its development as ambitious public-interest infrastructure. At the time of the grant, I committed to expanding CEL into the tens of thousands of records within two months. We reached 10,000 canonical public records in approximately ten days, substantially ahead of that commitment. The platform now includes: 10,000 source-linked canonical public records. 150,000 accepted import-wave quality-assurance candidates. 5,470 generated institution pages. 10,000 verifiable ProofGraphs. Exact source-cell verification for 6,000 newly promoted government records, with zero mismatches. Public correction, review, methodology, and responsible-use systems. Structured feeds and reporting tools designed for journalists and researchers. We have also incorporated documented outside feedback into CEL’s methodology and interfaces. Reviewers have pushed us to clarify what the archive proves, distinguish documentation from prevalence, improve citation workflows, and make institutional responses and limitations more visible. One of the accomplishments I value most is that CEL has grown dramatically without weakening its standards. The archive became much larger while its claims became more precise, more reproducible, and easier to challenge.
What we learned
We learned that the hardest part of accountability work is not collecting information. It is preventing information from becoming overstated. A public statistic is not necessarily evidence of prevalence. An allegation is not a finding. A record count is not a measure of institutional quality. A source can be official and still require careful interpretation. We also learned that trust should come from inspectability, not branding. CEL should never require someone to simply believe us. Every important claim should lead back to evidence that can be opened, reproduced, questioned, or corrected. Finally, we learned how dramatically capable tools like Codex can expand what a small public-interest project can accomplish. CEL was built by a young founder with limited time and resources, but the combination of clear standards, public data, and AI-assisted engineering made it possible to build infrastructure that would traditionally require a much larger organization.
What's next for Campus Evidence Lab
We are actively working on making civil-rights accountability something that does not depend on who has enough time, money, or technical ability to navigate fragmented public systems. CEL’s goal is to make verifiable evidence accessible when journalists, advocates, researchers, institutions, and affected communities need it most. The next step is getting real partnership with similar-minded institutions.
Log in or sign up for Devpost to join the conversation.