A modular rotational grip trainer, conceived through conversation with three different AI systems over a single afternoon. This is a build journal — honest about what's real, what's rendered, and what's still in my head.
I have a Spartan race in September. If you've never done one: grip is the thing that fails first. Hercules carries, monkey bars, multi-rig, rope climbs, sandbag hauls — every obstacle is filtered through your forearms. You can have the cardio of a triathlete and still get burpee'd because you couldn't hold on for fifteen more seconds.
Over the years I've collected the usual grip arsenal — Captains of Crush grippers, a twister bar, Fat Gripz, a wrist roller. None of them are bad. But each only trains one dimension, they're scattered around the house, and if I'm honest, I don't reach for any of them often enough to actually move the needle before September.
So I started exploring. Not building, not yet. Just probing the idea with the tools I have access to — three different AI systems, each suited to a different part of the loop.
I described the device in plain English — ab-wheel form factor, central disc housing some kind of resistance, two bars sticking out, swappable grip sleeves. I asked: does this exist? Then: what mechanism would actually work?
The conversation surfaced four candidate resistance mechanisms (torsion spring, friction disc, magnetic eddy current, viscous fluid damper) ranked by manufacturability vs. feel. It also surfaced the closest existing competitor — The Burn Machine — and confirmed nobody had combined the three modalities into a single compact device.
The LLM produced four image prompts — a hero shot, an exploded view, an engineering blueprint, and a use-case triptych. Each was tuned for a different downstream purpose: marketing, engineering communication, manufacturer briefing, narrative.
The blueprint is what surprised me most. It's not a real engineering drawing — it's a render that looks like one. Dimensions are internally consistent. A manufacturer could use it as a starting point for actual CAD, even though no real CAD exists yet.
Before getting any further attached, I sanity-checked the opportunity. Global grip strengthener market is sitting around $812M with a 5.4% CAGR. North America is roughly $153M. Online retail accounts for ~40% of distribution.
The numbers say the niche is real but small — not a unicorn, not a wasteland. The kind of category where a well-positioned product could carve out a defensible slice without needing to be a generational hit. Good enough to keep going.
I researched what's currently best for AI-assisted CAD. Adam CAD (adam.new) generates parametric models from text and exports print-ready STLs. Zoo Design Studio offers a more powerful conversational CAD agent with a steeper learning curve.
The plan: model each part of the device individually via text prompts (disc shell halves, grip sleeves, axle reference), 3D-print them, source the spring and bearings from McMaster-Carr, assemble for under $35 in raw materials. The CAD-AI tooling collapses what used to be the highest-friction step in this loop.
The same conversation that produced the renders also produced a 16-slide Kickstarter pitch deck — problem framing, solution, market data, unit economics, competitive landscape, manufacturing plan, use of funds. Real estimates throughout, with assumptions flagged.
I'm not launching a Kickstarter tomorrow. But having the deck exists at this stage matters: it forces the unit economics conversation early, exposes the cost gaps that need real quotes, and produces an artifact I can share with a manufacturer or potential collaborator without spending another weekend on it.
This project is on my portfolio not because it's done — it isn't — but because it's an accurate sample of how I think about building things in 2026. The interesting question isn't "did the AIs build the product?" They didn't. The interesting question is what I, as the builder, was freed up to do once the friction of ideation, visualization, market research, and document production dropped to near zero.
The answer, for me, was: more thinking, less typing. More projects entered, fewer abandoned at the "I should write a brief" stage. And a slightly higher bar for which ones get to graduate from concept to physical prototype — because the cost of staying in concept is now low enough that the only good reason to print plastic is genuine conviction.