Every claim on this platform maps to a source URL.
Beyond Valencia is built on a verified requirement graph. We publish counts, not adjectives. This page explains how the graph is built, how AI is used, and what standard every claim on the site has to meet.
The state of the graph, updated in real time.
Data source: GET /api/catalog/verification-summary · Generated 8/12/2026, 10:09:48 AM
These figures are aggregated live from our verified requirement graph and do not include placeholder, draft, or awaiting-review records.
No claim ships without a source.
Every prerequisite, GPA threshold, credit floor, articulation rule, and destination-program requirement on the platform is backed by at least one verification record. A verification record contains a source URL (public catalog page, Common Prerequisites Manual entry, articulation agreement, or tracking sheet), a source title, a note, and a last-verified date.
Unverified data is not exposed on the public site. It sits in a draft state in the admin catalog until an editor attaches a defensible source and re-runs verification.
Institution → program → requirement → source.
The graph starts with institutions (community colleges and destination universities), branches into programs (majors and specializations at each destination), and terminates in requirements (prerequisites, GPA floors, credit thresholds). Every edge is attached to at least one source URL.
When a student runs the Mishap Corrector diagnostic or a paid audit, the underlying evaluation walks this graph — matching Valencia (or any supported community-college) coursework to the destination program's requirement set, using the same sources that appear on the public pair pages.
Language models rewrite; they do not invent.
AI is used in two narrow places: rewriting deterministic templates on programmatic pair pages, and drafting per-student summary language in the paid diagnostic. In both cases the AI is constrained by an injection-schema prompt, its output is validated server-side against an allow-list of course codes and banned adjectives, and any hallucinated code triggers a fallback to the raw template.
AI does not decide admissions probabilities, invent GPA thresholds, or select target schools. Humans do.
Publish, then re-verify on a schedule.
Verified records carry a last-verified date. Records older than the site's review interval are surfaced in the admin verification queue and re-checked against the underlying source. When the source has changed, the record is either updated with a new note and last-verified date, or moved back to draft.
Every AI-generated pair-page intro is regenerable and cached with its model version so we can invalidate entire batches when catalog rules change.
No fabricated statistics. No inflated outcomes.
We do not publish acceptance rates that we did not measure. We do not invent testimonials. We do not round unverified figures. We do not describe our footprint using numbers that we cannot defend from audited records.
The live counts above are the numbers you should hold us accountable to. If a page on this site cites a figure and you cannot find it on this page or the underlying source URL, tell us — see the About page for contact routes.