Issue #19 — Carbon Robotics

Alright, let's get back out in Left Field.

Two issues ago, we featured a robot strolling through California strawberry fields at 2am, zapping mold with UV light — TRIC Robotics. Well. We found the company operating the day shift.

Meet Carbon Robotics. They built a machine that drives through a field, looks at every single plant, decides which ones are weeds, and shoots them with a laser. A real, 20-foot-wide robot vaporizing over 100,000 weeds an hour — one plant at a time.

No, not that weed. The one your college neighbor… never mind. I'm talking about the weeds out in your yard.

You've pulled them as a kid. You've pulled them as an adult. And you'll keep pulling them until the day you die. Death. Taxes. Weeds.

And you know the drill. Saturday morning, you spot one, bend down, grab it, pull — and you either get the cleanest pull of your life or it snaps at the soil line and you know, deep down, it'll be back like the plague. Either way you feel briefly like Thanos, eliminating unwanted things. But you look up and there are 40 more waiting. Breaktime!

The alternative is chemicals. Roundup. Whatever's been in your neighbor's garage since 1998. It works — right up until you remember your kid does cartwheels on that grass and your dog licks its paws after walking through it. Suddenly you don't want the stuff anywhere near the yard.

So here's the all-important question: if we won't spray these kinds of chemicals in our own yards, why are chemically intensive weed-control methods so common in growing our food?

Turns out it's a fair thing to ask. Weeds have been outsmarting our chemicals for decades — 250+ weed species have now evolved resistance to more than 160 different herbicides across 66 countries. Same way bacteria shrug off antibiotics. The go-to weapon of the last 60 years is losing, and the chemical industry doesn't have a great "what's next."

Farmers are stuck in the middle of it. Go chemical-free and the weeds eat the crop. Hand-pull instead and you need an army of workers that the labor shortage means you flat-out can't hire. So what do you do when a 60-year-old solution starts failing?

You get a laser. Welcome to the table, Carbon Robotics.

Here's the simple version.

The LaserWeeder hitches to the back of a tractor and spans about 20 feet. On board: dozens of high-res cameras, an onboard supercomputer, and a bank of industrial CO2 lasers.

As it rolls, the cameras livestream the whole field to Carbon AI, the brain running the show (someone please get this on a Twitch stream). The brain looks at every plant and makes a call in milliseconds. Carrot? Leave it. Weed? Zap it.

And to be clear about the pace here: the flagship unit fires at 10,000 weeds per minute. It just killed hundreds of weeds in the time it took you to read this sentence.

And the brain keeps getting smarter. In early 2026 Carbon launched its Large Plant Model — basically ChatGPT but for plants — trained on over 150 million labeled ones. It can spot a weed it's never seen before and know to kill it, no retraining required.

And weeds can't evolve their way out of this. You can outsmart a chemical. You cannot outsmart physics. There's no loophole for getting hit with a laser.

Paul Mikesell is not a farmer. He's a tech guy — the kind who keeps building companies and selling them for absurd amounts of money.

He co-founded Isilon Systems, a data storage company, back in 2001. Took it public in 2006, then watched EMC buy the whole thing for $2.25 billion four years later. Then a database startup, Clustrix, that got scooped up by MariaDB. Then he ran infrastructure engineering at Uber — deep in the computer-vision, machine-learning weeds (pun intended) — with a pit stop at Facebook's Oculus lab along the way.

So how does a guy whose entire résumé screams "San Francisco" end up building a laser tractor?

He asked himself a simple question: computers can finally see now — actually look at something and understand what it is. What's the most important real-world problem that unlocks? Weeds, of course. A massive, stubborn, everyone's-dealt-with-it problem that nobody had actually solved.

So he started building a prototype in his Seattle backyard in 2018. A man welding together a laser device in his backyard. Some real Iron-Man-in-a-cave energy.

A backyard science project to a machine now zapping weeds in fields across 15 countries. Not bad.

Here's the thing about farmers: they are the single hardest customer on planet Earth to sell to.

And it makes sense. Their entire livelihood rides on the crop coming in. One bad bet on some shiny new gadget and that's the season — the mortgage, the payroll, the whole year. So they don't buy million-dollar robots because the demo looked cool at a trade show. They buy them when the math is so obvious it practically slaps them in the face or when the guy one farm over is already running one and won't shut the hell up about it.

And they're buying.

Hungenberg Produce, a Colorado veggie operation, put a hard number on it: their labor bill dropped from around $700k to about $300k.

How about Grimmway Farms — one of the biggest carrot growers on the planet? They're running six of these things, around the clock, on their organic fields.

These aren't numbers off Carbon's pitch deck. This is farmers telling other farmers it works — which, in agriculture, is the only advertising that's ever mattered.

Now, it's not all sunshine and roses out in Left Field.

The gating item for a lot of farms: this machine is not cheap. It's a serious capital purchase, not something you impulse-buy off an Instagram ad — cough cough Matthew.

Though let's be honest, not all impulse buys are bad. I know someone who was asked to go get milk and came home with a Corvette. Milk was also purchased. Win-win if you ask me.

Carbon's whole answer to that upfront cost is the payback period — Paul's goal is under three years, and he says most farmers are landing between one and three. On the right farm, the math closes fast.

And "the right farm" is the actual catch. Carbon's sweet spot is high-value specialty crops — lettuce, carrots, onions, organic greens, think your local farmers market. Weeding is expensive on those crops, and the margins can absorb the machine. Commodity crops at corn-belt scale? Way harder. So this isn't every farm on Earth tomorrow — it's the high-value fields first, and everything else once the tech gets cheaper.

Then there's the competition, and it's not garage startups. John Deere’s Blue River Technology uses the same see-the-weed AI, just paired with a targeted spray instead of a laser — and it's got the biggest name in farming behind it. Add the entire multi-billion-dollar herbicide industry, whose whole business model is you not buying this, and Carbon's got real weight pushing back.

None of that is fatal. It's just the stuff that stands between a cool machine and a farm that runs on them.

Farming is brutal, physical work — endless variables, none of them in your control. So there's something pretty incredible about smart, ambitious teams dragging it into the modern age.

Carbon's Autonomous Tractor Kit is already running ground prep with nobody in the cab — basically the self-driving tractors from Interstellar. When the machines never clock out, the farmer stops being consumed by the operation and starts running it.

And weeds were just the opening move. That plant-recognizing brain can be pointed at almost anything — thinning, scouting, harvesting. Carbon's already torched over 40 billion weeds across 15 countries. The laser is just the first thing they aimed it at.

The endgame is a farm that handles the 100,000 decisions an hour that don't need a human — so the human can go handle the ones that do.

That's a Left Field we want to be in.

Carbon's careers page puts it plainly: "We build technology that works in actual fields, under real conditions — solving genuine problems for farmers and the food supply."

Real fields, real lasers, real problems. Beats staring at an Excel sheet all day.

Open roles → carbonrobotics.com/job-openings

Left Field goes out every week. One company. No filler. Know a company that belongs out here? Hit the comment button and let us know.

If someone sent this to you, subscribe here: