Designer builds open-source AI product photo studio Scenri in 2 months with Claude Code
Tonjiez · reddit · 2026-09-28
A product designer built Scenri, an open-source AI product photo studio that solves the 'same person twice' problem: save the person, product and place once, then compose shots like @Maya in the $Red Puffer Jacket on a /Brooklyn Rooftop — real photos are sent to the image model every time. Images come from your own ChatGPT/Codex CLI or API key; Scenri is free, serverless, account-free and telemetry-free. Claude Code wrote most of it: 1,600+ commits and 75 releases since Aug 4. Key practices: per-task git worktrees for parallel sessions; a CLAUDE.md that blocks test runs until design decisions settle (e2e only in CI); 41 memory notes from corrections; and a confidence table in every answer, forcing proof-seeking below 95%.
More from coding & agent
- How this team uses Codex to build Google Slides decks in four steps — every · 2026-09-28
- Dev monetizes automated newsletter via Markdown promo files — 3 sales on launch day, 41% open rate — iannuttall · 2026-09-28
- A production-line workflow for pro motion design with Opus 5.5: storyboards first, render last — sven_ai · 2026-09-28
- NVIDIA Open-Sources OpenShell, a Safe Private Runtime for Autonomous AI Agents — MikkoH · 2026-09-28
- NVIDIA launches Open Agent Safety Platform with 100+ industry partners — Polymarket · 2026-09-28
- I had Claude build its own F-Stop — and I'm floored — themayorofbikini · 2026-09-28