About Prizm

Prizm is media analysis infrastructure. It performs deep dives into persuasion techniques used in podcasts and news articles — showing you exactly what rhetorical patterns are present, where they appear, and how they work together.

Every episode and article gets a full breakdown: individual technique detections with source quotes, chunk-level heatmaps showing where persuasion clusters, and severity assessments across the entire piece.

If you have seen OrgnIQ, think of it as the simplified nutrition label — a single score and a quick summary of what is in the media you consume.

Prizm is the full lab report. Every technique, every quote, chunk-level detail. It is built for people who want to understand the mechanics of persuasion, not just get a number.

All analysis is performed by XrÆ, a 70+ billion parameter AI model that runs locally on dedicated hardware. No content is sent to cloud AI services for analysis. XrÆ was trained through a teacher-student distillation process specifically for rhetorical technique detection.

XrÆ identifies 24 distinct persuasion techniques across 5 families:

• Emotional — fear appeals, emotional language, false dilemmas designed to provoke reaction
• Faulty Logic — straw men, slippery slopes, cherry-picked evidence, false equivalences
• Loaded Language — framing through word choice, euphemisms, dysphemisms, repetition
• Trust Manipulation — appeals to authority, bandwagon effects, ad hominem attacks, testimonial misuse
• Framing — agenda setting, narrative framing, selective emphasis, anchoring

Prizm is built by one independent developer — no corporate backers, no investors, no editorial board. The entire stack, from model training to the site you are reading, is a solo operation.

Prizm is part of TruthSyrum, a broader media literacy effort focused on making persuasion techniques visible and understandable.

Questions, feedback, or suggestions: media@truthsyrum.io