Prototyping
Also called Throwaway prototype · Spike
Building a quick, throwaway version to learn something, not to keep. With agents it's cheap enough to do before deciding anything.
Prototyping is building the fast, disposable version of something to find out whether the idea works, what it feels like, or where the hard part is. The deliverable is the answer, not the code. Agents change the economics: a prototype that used to cost a day now costs a prompt and twenty minutes, which means you can afford to prototype before you commit to a design concept, not after.
The danger is the prototype that doesn't get thrown away. A vibe-coded demo works in the happy path, someone likes it, and it quietly becomes the product, with no tests, guessed data shapes and a folder called temp2. Because agent output looks finished, the line between 'prototype' and 'v1' blurs faster than it did when a prototype looked rough.
Decide before you start which one you're building. For a prototype, drop the ceremony: no spec, full vibe coding mode, minimal human review, a fresh worktree or directory you can delete. Write down what you learned when it's done. Then, if it earns a real version, start that version clean, with the prototype as a reference in the context rather than as the codebase you extend.
- Claude Code and CodexA throwaway directory plus an auto-accept permission mode gives the fastest prototyping loop, since nothing needs approval.
- ChatGPT Work and Claude CoworkGood for prototyping documents, decks and dashboards before anyone builds software.
“Should we design the onboarding flow first or just build it?”
“Prototype three versions this afternoon, click through them, then write the spec for the one that felt right.”
Course 01 puts every one of these terms to work: you install Claude Code, run the loop, and ship a real project — permission modes, compaction, hooks and all.
Start Course 01 →