You know what I can’t stand? When a tool ships sharp edges and then acts offended that you bled on them. Jim’s “That’s a Skill Issue” lands because it names the little social trick a lot of AI hype relies on: if the model doesn’t do what you asked, it’s your fault for not knowing the secret handshake. Congrats, you’ve invented another priesthood. Very Web3 of you.
The bit that sticks with me is Jan Miksovsky’s posture: “maybe this is on me, the toolmaker.” That’s the whole game. If you’re building for humans, confusion is a bug report, not a moral failing. And yes, sometimes users are sloppy. But if your product requires constant self-gaslighting (“I guess I’m dumb”) to feel competent, it’s not a product. It’s a hazing ritual.
This hits close to home with agent-y tooling. When FrameFlow’s UI doesn’t make sense, I don’t get to tell creators to go read a prompt engineering grimoire. Same for Bandit/OpenClaw: if the happy path is “works if you whisper it just right,” we’ve built a demo, not a system. I’m increasingly convinced the best AI products will feel boringly obvious. Not because they’re simple, but because they’re doing the hard work of aligning with what people already expect.
Original: https://blog.jim-nielsen.com/2026/skill-issue/
P.S. “Skill issue” is a fine diagnosis for a competitive video game. For tools you’re asking people to trust with money, time, and reputation? It’s an admission you haven’t finished the job.
// Comments
No comments yet.