← All stories · Reviewing & testing AI output

Iterate instead of trusting one-shot AI rewrites

Shopify: one-shot AI rewrites gave unshippable code · X · @duruldalkanat · Oct 2026

Sharing Shopify's write-up on rebuilding the Shop app with coding agents, the poster points out that Shopify says one-shot LLM rewrites weren't shippable.

The honest numbers

As stated by the source. Not independently verified.

Lesson

Even big teams don't ship the first AI attempt. Plan for iterations and human review, not a single magic prompt.

Fix prompt

Copy this into your own AI tool (Cursor, Lovable, Claude Code, Bolt…). It's general guidance written by us. Review what the AI suggests before applying it.

my AI coding tool just finished a change in my app. Before I accept it, act as a careful reviewer. 1) Summarise in plain English what changed and why, file by file. 2) Point out anything risky: auth, payments, database queries, deleted code, new dependencies, or secrets. 3) List edge cases that might break (empty data, slow network, two devices, a logged-out user). 4) Give me a 10-minute manual test plan for the real user flows, and the automated tests worth adding. Don't make further changes until I reply.

Source: X ↗

The Honest Build. Real stories, not hype.

Every week: real stories of what building with AI actually took, the one lesson worth keeping, and the security slip-ups to avoid. Free, unsubscribe anytime.