I Can Build It, but I Need AI to Explain It

I'm not an engineer or a developer, but AI, and Claude Code in particular, has let me explore fields full of opportunities for discovery and improvement. As I move through these fields, long the territory of mathematicians, engineers, and developers, I'm finding that I can bring real value to the table. I built my own GPU flip-graph search engine for bilinear schemes with coefficients in {−1, 0, 1}. It has lowered the best known rank over any ring for 7×11×15, 11×13×15 and 11×14×14, and produced 22 further ternary records. The work has been recognized by leading researchers in the field. I also built an engine that grows logic circuits, which I run on my KV260 FPGA board. With it, I reached #1 on the LatchUp leaderboard within a few hours. And there's more I'm working on that keeps pushing things forward. Underneath all of this is my own harness pipeline, the foundation that makes building these tools and engines possible in the first place. I'd describe myself as a systems thinker. I take the tools available to me and shape them into something useful. I may not know every piece in depth, but I can see how the pieces fit together, and that bigger picture is what turns them into something truly valuable. But this raises a deeper issue. More and more people like me will build things like this, yet struggle to explain the inner workings to engineers and developers without AI's help. For now, the things we build and the results they produce are what speak for our contribution. Increasingly, communities ask people to answer in their own words, without AI. I understand why. Those rules exist to filter out low-effort, AI-generated noise from people who don't understand what they're posting. That concern is fair, and I share it. But there's a difference between AI as a substitute for understanding and AI as an interpreter for real work. My results can be checked: records anyone can verify, hardware that runs, outcomes that can be reproduced. For someone like me, AI isn't a way to avoid thinking; it's how I translate what I understand into language experts can follow. Without it, the ideas are still there, but I lose the means to share them. The question isn't whether someone used AI to explain their work. It's whether the work holds up. what do you think about this ?