Redline

Make it pop' is not feedback. Redline uses AI to guide reviewers into giving structured, specific critiques — and turns the session into a shareable action report.

Comment

Redline — Structured Design Critique, Powered by AI

"Make it pop" is not feedback. Redline makes sure reviewers never say it again.

The Problem

Every design critique session has the same four people:

  • The senior person who speaks first and kills the room
  • The reviewer who types "looks good but maybe more contrast"
  • The developer who says nothing because nobody asked the right question
  • The designer who leaves with six opinions and zero clarity

This is not a people problem. It is a structure problem.

What Redline Does

Redline runs the critique session properly — before, during, and after.

Before: The designer sets a brief — goal, target user, fidelity, what to focus on, what to ignore. Reviewers see the context before they see the design.

During: Each reviewer answers 5 structured questions adapted to their role. A PM gets different questions than a developer. A Director gets different questions than a designer. Nobody is asked to think like something they are not.

The AI moment that matters: When a reviewer types "I don't like it" — Redline flags it before it saves and asks:

"Can you name the specific element and explain why it creates that feeling for a first-time user?"

Vague feedback does not go through. Or the reviewer can override — but it gets flagged in the data.

After: AI reads all submissions and produces a structured report: themes, agreed points, debates, open questions, action items with suggested owners. In under 60 seconds.

One shareable link. Permanent record. No more Slack threads that disappear in two weeks.

How I Built It

Stack: Next.js 14 · TypeScript · Tailwind CSS · Anthropic Claude API (claude-sonnet-4-6) · Vercel · Novus.ai

Two AI functions that make it real:

  • checkFeedbackVagueness() — catches weak answers before they save
  • generateSummary() — reads all feedback and writes the report

Novus.ai connected via GitHub — auto-instrumented, tracking every key flow from session creation to report generation. The most interesting metric: how many reviewers click "Submit anyway" after the vagueness warning fires.

What I Learned

Structure changes behaviour faster than instruction does. You do not need to teach people how to give good feedback. You just need to make vague feedback impossible to submit without a second thought.

And anonymous feedback is not about hiding. It is about making room for honest thought before social dynamics take over.

What the Judging Criteria Required — and What We Delivered

Criteria What Redline does
Product Thinking Solves a real, documented problem with role-aware questions and intent-first setup
Craft & Execution Vintage academic design system, real error/loading/empty states, mobile responsive
Originality No tool guides a team critique with role-aware AI prompts + vagueness blocking
Shippedness Live on Vercel · Novus installed · real sessions tracked · one URL, works now

Built for Everyone Ships Now — Mind the Product × Novus.ai — June 2026

Built With

  • anthropic-claude-api-(claude-sonnet-4-6)
  • next.js-14
  • novus.ai
  • tailwind-css
  • typescript
  • vercel
Share this project:

Updates