What to build next
AI will build anything. It won't tell you what's worth building.
The hardest question in software right now isn't how. It's what next. A model can scaffold a feature, a service, a whole app in an afternoon. What it can't do is tell you which one is worth your time.
That decision is the compass, and it's the one part of the job you can't hand to a tool.
A lot of developers are now asking: "What skills should I develop? Which framework, which language, which area do I go deep on?" It's a moving target, and the list just keeps growing.
Flip it: what's worth building? Answer that, and the skills, areas, and expertise you need follow from it. What you build tells you what to learn, not the other way around.
So that's where I'd start. Not with a framework checklist, but with one question:
What's the most valuable thing you're waiting for permission to build?
Two ways in:
Already have an idea? Scope it into a one-page MVP brief in 30-60 minutes, free, no signup.
Still figuring out what's worth building? Brainstorm it with me and I'll reply with a direction, even if it's fuzzy. No pitch.
How I've used this to ship apps
Ask the Canon
One idea, scoped first. From downloading Gutenberg book texts to building a semantic search without vector db.
checkpulse
Privacy-first analytics in a single binary. Scoped to MVP, built in Rust, deployed on Fly.io.
Pybites Platform
A full Python learning platform, in production for years. The MVP was tightly scoped around exercise solving.
mdweaver
Turn one or many Markdown files into polished PDFs and EPUBs. A small tool with a focused scope: do one job well.