Built mid-project with Claude Code — a feedback tool embedded directly into a client landing page, with a 24-hour turnaround and no design system to start from.
the problem: collecting design feedback without drowning in email threads, annotated screenshots, or "can you jump on a quick call?"
what this demonstrates: using AI not just to write code, but to build workflow tools that solve the actual problems you hit mid-project — with a 24-hour turnaround, no Figma file or design system.
how it works: click "Review Mode" and the page transforms into a feedback workspace. each reviewer gets a brand color. comments anchor to exact positions and persist across sessions. a slide-out panel groups all feedback by section with resolve tracking.
The prototype was built conversationally in Claude Code — this is the workflow I'm most excited about.
Every feature maps directly to a real friction point in the stakeholder feedback process.
One click transforms the live page into a feedback workspace — pins, comments, and all — without leaving the design.
Each stakeholder is auto-assigned a brand color, eliminating the guesswork of anonymous feedback threads.
All comments grouped by page section with open/resolved counts and one-click resolve. Everything in one panel.
Connected to an external database so feedback persists across sessions and review cycles. Swap in whatever backend works for you.
A quick demo of the feedback workspace in action — inline pinning, reviewer color-coding, and the slide-out panel with resolve tracking.
The BrightPath landing page — the actual client page this review system was built into — is live.
open brightpath demoSame response. Different emotional state. Completely different coaching. An interactive prototype exploring how emotional context changes what good feedback looks like.
explore the experiment ← all experiments