Judgment Layer
A prototype for AI-era reading that shifts attention from fluent output to evaluation, conceptual structure, and endorsement.
Status: in-development
A prototype for reading and writing in the age of language models
Language models make it easy to generate fluent text. That changes what matters. Fluency is no longer proof of understanding. This prototype explores how interfaces can better support evaluation, not just production.
Problem
Most AI interfaces are optimized for output. They help produce text, but offer very little support for evaluating it. As a result, confident language can pass as credible reasoning.
Hypothesis
As synthesis becomes cheap, interfaces should better support judgment. They should help users inspect assumptions, omissions, relationships, and endorsement before treating fluent text as trustworthy.
What this prototype does
- Reading view with linked concepts
- Concept panel for definitions and relationships
- Synthesis vs judgment comparison
- Map view for structural relationships
Why it matters
This is not just about faster writing. It is about helping users decide what to trust, question, revise, or endorse.
Role
Concept, interaction design, visual design, front-end prototype
Outcome
This prototype proposes a small set of interface primitives for AI-era reading. It focuses less on generating content and more on helping users inspect meaning, criteria, and relationships. The result is a quieter model for AI product design that treats judgment as a first-class interaction problem.