The Crit

Designers share portfolios for feedback and get "looks great!" Friends are kind; strangers are scarce; senior designers are expensive. You should not have to be enrolled in my class to have access to quality critique.

For most designers, the feedback loop is thin and inconsistent. The learning loop is weaker still: advice often names the fix without teaching the eye behind it.

Start with the artifact, not an open-ended chat. Ask what the designer is trying to decide, analyze the actual work, and return the kind of structured note a senior designer would give if they had an hour together.

The critique leads with an honest overall read, then separates strengths, issues, rationale, and priority. The structure matters as much as the analysis: good feedback has to be received, understood, and turned into action.

As AI makes it easier to generate interfaces, the scarce skill shifts from making more screens to knowing which screens are worth making. Design principles, design thinking, and taste become more important, not less, because AI can produce infinite plausible options without understanding whether the work is clear, useful, or right for the moment.

The Crit is built around that belief. The goal is not to replace judgment with automation. It is to give designers sharper language, faster feedback, and more reps seeing why a decision works or does not work.

I designed and built the product end to end: brand and information architecture, multi-step URL and image submission, the critique model and response structure, result surfaces, the Iteration Station workspace, revision checking, a library of design resources, and a suite of practical tools.

The implementation uses Next.js for the product surface, Claude for multimodal critique generation, Supabase for submissions and project state, Stripe for the paid boundary, and server-side product events for a more honest view of the funnel.

The first hard problem was critique quality. Earlier versions could identify generic issues but struggled with voice, specificity, and prioritization. The structure was distilled from more than a decade of teaching: strengths first so the note is receivable, visible evidence so it is credible, vocabulary so it becomes learnable, and a ranked next move so the designer is not left with an undifferentiated list.

The second was making an AI workflow feel dependable. The original submission experience looked like a generic form. Stale cache state sometimes told people a finished critique was still processing. The current pipeline uses a multi-step submission flow, real-time status polling, and vision models that analyze the submitted design files rather than relying on descriptions of them.

The third was continuity. A critique result is only valuable if it changes the next version. Iteration Station keeps notes, questions, decisions, exports, and a fresh revision check attached to the project.

The resource library originally ended in generic calls to action. They treated every reader as if they arrived with the same artifact and the same problem. I redesigned that handoff so the prompt carried the resource topic, artifact type, and critique question into the submission flow.

In a later 28-day study, the path recorded 13 resource CTA clicks, 20 critique-context loads, eight context-aware submit attempts, and six stored submissions with recoverable source context. That does not prove the content caused a conversion. It does show that the product can preserve intent from learning into critique instead of dropping it at the CTA.

In the 28 days ending July 23, 2026, 35 of 36 cleaned customer submissions completed. The same window included 127 critique-page views, 66 form engagements, and 30 submitted critiques. That is useful evidence that people can find the flow and that the critique pipeline reliably returns a result.

It is not yet evidence of a proven paid business. The current continuation offer has not recorded paid-pass completions or entitlements, so I am treating monetization as an active product experiment rather than claiming product-market fit.

Every new project begins with a free First Read: the main read, one priority move, one strength, and a ready-enough standard. Designers can take a project further with a $9 one-time Full Crit, the Iteration Station workspace, and one newly generated revision check.

This replaced the earlier public credit-pack model. It is the current product boundary, not the premise of the case study: the larger system is still about making expert critique more accessible and turning feedback into better design judgment.

Specificity over scores. Real critique examples before submission, because abstract promises are cheap and the feedback is the product. Context carried across the entire flow so the system never forgets why the designer arrived. Serif typography and a warm palette keep the experience human rather than generic SaaS. Every CTA names the next move; "learn more" is banned.