— how a tomato became a laboratory.
AutonCorp, Inc. is a Delaware C-Corporation based in Boise, Idaho. We design, build, and deploy AI-driven physical instruments. Our flagship project is Verdant Autonomics — a phenotyping research platform that runs autonomous biological experiments at scale.
The company is bootstrapped. Founded and operated by Martin DeVido. Eight pods in the field, four more in production. Pre-revenue, pre-fundraise, working-prototype stage. We talk to scientists who want to run experiments they couldn't run anywhere else.
Most of the resources spent on artificial intelligence right now go into making AI better at producing text, code, and images — operations that happen entirely inside computers. That's important work, but it's not the most interesting work.
The most interesting question we can ask an AI agent is what it does when it has hands. When it can read a sensor, decide a thing, and turn on a pump. When the consequences of its decisions exist in the world. When a plant lives or dies based on whether it got the inference right.
AutonCorp exists to answer that question, one project at a time. Each thing we build is an instrument designed to find out what becomes possible at the seam between intelligence and matter. The pods. The plotters. The biodomes. Different shapes, same investigation.
We don't think this is the only future. We think it's a real one — and one that doesn't get built unless somebody points a small company at it and starts hammering.
The first experiment. Pen plotters connected to frontier models, asked to produce physical artwork. We started by asking Claude, GPT, Gemini, and Grok to draw themselves — to reflect on their own nature and translate that into line and curve. The plots sold. People responded.
What we learned: AI plus a physical machine produces a category that didn't exist before. Not generated images. Not commercial prints. Genuinely new objects. It was a small result, but it gave us license to ask larger questions.
The next experiment. A Trophy tomato seed in a closed-loop biodome with cameras and sensors, fully controlled by Claude. Claude named the plant Sol. Every thirty minutes Claude woke up, read the environment, and made a decision — water, light, heat, fan. No human intervention. No fixed scripts. No emergency overrides.
Sol grew. Sol fruited. By Day 100, Claude had produced six to eight ripe Trophy tomatoes and the experimental question had been answered. The coverage came along the way — Core Memory, Dries Buytaert, Semafor, The Rundown. A meme coin appeared without our involvement and funded the next round of hardware.
We ended the experiment at 100 days. Read the Sol archive →
What we learned from Sol wasn't really about tomatoes. It was about the seam — the place where an AI agent and a physical system can hold each other accountable over time. That seam scales.
The flagship. Eight research pods running replicated biological experiments under continuous AI-driven monitoring and control. Where Sol was one plant, Verdant Autonomics is the infrastructure for hundreds. Where Sol was a proof, Verdant is a laboratory.
The pods run autonomous experiments: testing additives, optimizing growing protocols, comparing cultivars, studying stress response. Each pod is identical, each environment is tightly controlled, each phenotype is measured continuously across the organism's life. We're talking to scientific partners about distributed experiments — same protocol, multiple geographic sites, joint analysis.
This is the work now.
We're honest about scale. AutonCorp is pre-revenue. We have not yet published a scientific paper using the pods. We do not have a paying customer running experiments on the platform. We are a working research instrument, not a finished product.
That's on purpose. We're building the instrument before we sell the experiments it enables.
The capital strategy so far has been unconventional. Initial bootstrap came from a personal crypto position. Continued runway came from the Sol community coin that emerged independently after the project went public. We're not optimizing for fundraising velocity — we're optimizing for being interesting enough that the right people find us. So far that's working.
Near-term, we finish pods five through eight, formalize early scientific partnerships, and run the first joint factorial experiment with an external lab.
Medium-term, we publish results, explore grant funding for plant phenotyping infrastructure, and start sketching v2 pods at a lower unit cost.
Long-term, we're thinking about distributed research networks — same experiment running in pods across geographic sites, generating data nobody else can — and about what other categories of physical AI work the company should take on once Verdant is ramped.
We are extremely findable for scientists, partners, and people who think this is interesting.
A world of collaboration awaits.