CASE 008 · SESAME 2025–26 · private beta · native macOS
Sesame collection grid — visual references arranged in a searchable layout
AI / ML · Desktop · Local-first

Sesame.

A local-first visual taste tool that learns your aesthetic from your own references — embedding-native semantic search, style briefs, and 3D cluster mapping. Built from scratch as a native macOS app.

Sesame is a native desktop application that helps designers organise and understand their visual reference collections. Import images from anywhere, search them by description, and let the app surface patterns in your taste you might not have noticed.

Everything runs locally — no accounts, no cloud, no images leave the machine. I built it to solve a problem I kept hitting in my own design work: references scattered across platforms with no way to search or make sense of them.

This project applied the same rapid-prototyping and systems-thinking skills I use in physical product design to a software context — scoping a real problem, building a working solution, and shipping it under tight time constraints.

Type
Personal project · design tool
Year
2025–2026
Role
Solo — design, architecture, build
Stack
Swift (SwiftUI) · CLIP embeddings on-device · local SQLite
Status
Private beta · native macOS

The problem.

01 / PROBLEM

Designers collect visual inspiration across many platforms, but these collections end up scattered, unsearchable, and disconnected from the actual design process. There's no good way to ask "find me images with that warm brutalist feeling" across everything you've saved.

Existing tools treat inspiration as a static archive. Sesame treats it as a living, searchable resource that becomes more useful the more you engage with it.

4daysconcept to working product
7interconnected views
100%local-first / offline

What it does.

02 / FEATURES
● Sole developer

Natural language search

Describe what you're looking for in plain language and the app returns visually relevant results — no manual tagging required. CLIP embeddings are generated on-device when images are imported.

Sesame — natural language search results

Refine view

A full-screen view for rapidly reviewing your collection. The app learns from your responses over time, personalising results to your taste.

Sesame — refine view for rapid taste calibration

Visual mapping

An interactive 3D visualisation that reveals how your references relate to each other in embedding space, making implicit aesthetic patterns visible.

Sesame — 3D cluster map of image embeddings

Style briefs

Generates written design briefs grounded in your actual collection — principles, vocabulary, and direction based on what you've curated, not generic labels.

Sesame — auto-generated style brief

Moodboard builder

Arrange selected images on a canvas and export presentation-ready boards directly from your curated references.

Sesame — moodboard builder canvas

Design approach.

03 / DESIGN

Privacy-first. Designer reference collections are personal and often contain copyrighted material. Everything runs on-device — no accounts, no cloud uploads, no data leaves the machine.

Minimal interface. Dark background, warm accent, system typography — the UI stays out of the way and keeps focus on the images. A consistent design system ensures coherence across all views.

Sesame — collection grid Sesame — settings panel

Skills applied.

04 / SKILLS

This project applied the same methodology I use in physical product design — just in a different medium:

  • Rapid prototyping under constraints. Scoping, building, and shipping a working product on a tight self-imposed deadline — the same discipline used in fabrication sprints and client deliverables.
  • Systems thinking. Designing interconnected components where each part informs the next — the same approach used in physical product systems.
  • User-centred problem solving. Built around a real workflow problem, shaped by direct experience as a designer.
  • End-to-end delivery. Owning every layer from backend to interface to packaging — the same full-ownership approach I bring to physical product work.
05 / REFLECTION

Building Sesame reinforced something I've found across all my work: the best outcomes come from treating constraints as design drivers. A tight timeline forced ruthless prioritisation. Technical constraints forced cleaner decisions. The result is a focused tool that does what it needs to and nothing more.

This project confirmed that the skills I've developed through physical product design — scoping problems, prototyping fast, and shipping complete deliverables — apply directly in a software context. The medium changes; the methodology doesn't.

← Back to work