The last twenty percent.
AI can supply a convincing structural explanation. The harder question is whether I know enough to trust it.
The hardest part of working with AI on furniture hasn’t been getting ideas out of it. It’s been figuring out which parts of its answers I can actually use.
In this part of my chair experiment, I started looking at anti-racking mechanisms: the structure that would keep the chair from distorting sideways under load. AI had plenty to offer. Steel stretchers, threaded inserts, plywood backers. It could describe the need for reinforcement and produce diagrams that made the suggestions look quite convincing.
It sounded right. The diagrams looked right. I had moved beyond an attractive image into an explanation of how the chair might work.
But I don’t have direct furniture design experience. I couldn’t tell whether the proposed bolt spacing made sense, whether the suggested 3 mm steel was enough, or whether those threaded inserts would be properly seated for the loads they would see. The answer had become more specific without my ability to judge it becoming any stronger.
That was the uncomfortable part. A vague answer is easy to recognise as unfinished. A detailed answer gives you much more to work with, but it can also make the remaining uncertainty harder to see. Familiar technical language starts to feel like evidence.
The diagrams were still useful. They pointed me toward questions about load paths, internal reinforcement and which elements would actually do the structural work. They helped me see what needed attention. That is a meaningful step forward, even if none of the proposed details can yet be treated as resolved.
AI can make it feel as though the work is eighty percent there. But that last twenty percent can include the connection that decides whether the chair holds together. It looks small beside everything AI has supplied. Its consequences aren’t.
I hadn’t appreciated how large the gap could be between receiving a technical proposal and knowing whether I could trust it.
I have the proposal. I’m still learning how to make the judgment. Until that gap closes, these are directions to investigate, not details I can confidently build from.