Source Verification for Journalists With the SPRAPP Suite

Verification Is the Job

A newsroom lives or dies on getting it right. AI can accelerate verification, but it can also confidently invent. The SPRAPP suite is structured to use multiple models against each other rather than trusting one. It is a tool; the journalist verifies and the editor decides.

Screening Inbound Tips

Tip lines receive content from strangers, some of it adversarial. Sprappy Filter scores incoming tips across its 25 categories, catching injection attempts and manipulation aimed at any model that later processes the tip. This keeps the verification pipeline from being steered by a malicious source.

Cross-Checking Claims With Panel

When a claim needs a sanity check, SPRAPP Panel can review it once and report where models converge or disagree.

• Convergence is a weak signal worth following up with primary sources.
• Divergence is a loud signal that the claim is contested or ambiguous.

Crucially, a panel that disagrees is doing its job — it tells the reporter not to trust the easy answer. No model output substitutes for confirming with real sources.

Offline Drafting in the Field

Reporters work from places with no signal. TinyLM runs on-device, so a reporter can draft a story outline or notes in the field with no upload and no connection. Sensitive source material stays on the device.

Keeping Tip Metadata Small

A tip line accumulates short strings — tip IDs, status tags, routing labels — across years. smoltext compresses these short strings efficiently, where gzip would waste space on per-record overhead, keeping the tip archive lean.

The Newsroom Stays in Charge

Filter guards the intake, Panel stress-tests claims, TinyLM drafts offline, and smoltext keeps records compact. Verification remains a human craft.

Where to Start

Put Sprappy Filter on the tip line first, then introduce Panel claim-checks as a research aid, never as a publishing gate.