Matthew Brunelle · 2026-04-23 · notable
It's OK to Use Coding Tools to Finish the Projects You Were Never Going to Finish
268 HN points for an essay distinguishing 'wish-list' projects (things you want to exist but won't build) from learning projects. For wish-list projects, AI coding assistance is legitimate and not a deskilling risk.
A viral essay argues AI coding tools are legitimate for 'wish-list' projects — things you wanted to exist but realistically never would have built.
What is it?
A personal blog post by Matthew Brunelle that drew 268 HN points and a broad discussion about when using AI for coding is intellectually honest. The central argument: there is a meaningful difference between 'learning' projects (where deskilling is a real risk) and 'wish-list' projects (things you genuinely want to exist but lack time or specific knowledge to build). Using Claude Code for the latter is not a shortcut — it's practical.
How does it work?
Brunelle used Claude Code (Opus 4.6) with a structured setup — Pydantic models, type annotations, Google-style docstrings — to build a YouTube Music connector for the OpenSubsonic API in an evening. A project he'd wanted for years but never prioritized. The post is concrete: it describes the actual workflow, not just the thesis.
Why does it matter?
The HN thread surfaced a broadly shared anxiety: using AI for code feels like cheating unless it's a learning exercise. Brunelle's framing gives practitioners a useful mental model for when it isn't — and validates a use case many developers recognize but haven't articulated: finally finishing the backlog that was never going to happen otherwise.
Who is it for?
Developers with a pile of untouched side project ideas