Separate "more matches" from "better matches"
The market is full of tools optimizing for volume. The gap was a tool that optimizes for truthful signal instead.
Case study · Honest dating-profile review
HingeReview helps you show up honestly.
A React mobile-web prototype that helps improve Hinge profiles without becoming fake, generic, or manipulative: clearer photos, stronger prompts, and better dating-intention alignment.
STATUSFunctional prototype · cloud review path next
ROLEProduct, scoring design, and provider integration
STACKReact 19 · Vite · Cloudflare Workers
HINGEREVIEW · SNAPSHOT
6CORE SURFACES
2AI PROVIDERS WIRED
PDF+ MARKDOWN EXPORT
- Audits photos, prompts & intention signal without pushing anything fake
- Computed profile audit, action plan, and a prompt rewrite flow
- Provider seam for Claude + Gemini vision review, cost-aware by design
- Local draft persistence: nothing leaves your browser until you export
Problem
Most dating-profile advice pushes you toward generic or fake.
Templates and "growth hack" prompt libraries optimize for matches, not for honest, sustainable connection. The result is a wave of profiles that all sound the same.
HingeReview instead asks what would make this specific person's real self read clearly: sharper photo choices, prompts that reveal something true, and dating-intention signals that build trust instead of volume.
PRODUCT LOOP
Audit, then rewrite with more truth.
The loop turns vague self-consciousness about a profile into a specific, actionable plan.
01
Input
Paste in your current prompts and photo choices, or use demo mode to see the tool work first.
02
Audit
A computed profile audit scores clarity, prompt strength, and intention alignment.
03
Plan
An action plan and prompt rewrite flow turn the audit into concrete next steps.
04
Export
Export the plan as Markdown or PDF, or keep iterating with local draft persistence.
Process
Built with the same operating discipline as my BA work.
Keep scoring specific to intention
A generic photo score is not useful. The audit needed to reason about what a given dating intention actually needs to communicate.
Build the deterministic engine before the AI
Ship a real local review engine with demo mode first, so the async Claude/Gemini review path has a reliable baseline to compare against.
Current Build
- Input loop, demo mode, and local draft persistence.
- Computed profile audit, action plan, and prompt rewrite flow.
- Markdown/PDF export of the full action plan.
- Cloudflare Worker backend seam for a future vision-model review.
Next Bets
- Wire the live UI to the async cloud review path.
- Validate Claude/Gemini requests with real keys.
- Expand the action-plan rewrite flow.
- Usage-based cost guardrails on the provider seam.
What I learned
Structured output matters more than model choice.
The hardest part was not picking Claude vs. Gemini: it was designing a JSON contract precise enough that either provider's output could be trusted and rendered the same way.
That structure is also what keeps the tool honest: a reviewer that can only answer inside a fixed rubric cannot quietly invent flattering nonsense.
Functional prototype today, cloud review next.
HingeReview works end to end locally: input, audit, action plan, and export all run today. Next: wiring the live UI to the async cloud review path and validating real Claude/Gemini keys.