[searls]

In early September, Mike Judge's post challenging the productivity potential for coding agents by asking where the supposed AI shovelware is. It's been stuck in my craw for the past couple weeks, because I keep running into programmers of all stripes who refuse to engage with any of this newfangled AI tooling on the grounds that it's useless.

I realize people are all processing the AI boom differently, and there are many valid reasons to be really mad about all of it, but—as someone who's seen his productivity skyrocket with Cursor, then Claude Code, and now Codex CLI—it seems a little ridiculous to claim that the tools simply aren't capable.

Of course, if we never identify which projects were facilitated thanks to coding agents, we'll never have clear social proof of what they're capable of. That's why I put up this /shovelware page on my blog and started hosting a GitHub badge.