Kernel Magazine has always had a deep appreciation for corn, for reasons that are too obvious to list here. When we got this pitch from Vicki Xu profiling a hare-brained but somewhat plausible scheme to see if someone’s Claude could grow and sell corn, we simply had to say yes.
Also, if you’re in New York: we’re throwing a launch party for Kernel 6 at Hex House in Brooklyn this Sunday on November 8! We’ll be featuring readings from the issue as well as a behind-the-scenes look at the making of the magazine from Kernel Magazine Creative Director Connie Liu! We hope to see you there! Get your tickets now!
—Jacob Sujin Kuppermann
🌽 Can AI Grow Corn? 🔩
By Vicki Xu
Edited by Jacob Sujin Kuppermann
Farmer Fred, born on January 22, 2026, has never sown a seed, driven a tractor, or separated corn from its silk. Farmer Fred lives entirely on the Internet where, propped up by a lean team of technologists, he aims to coordinate enough interested farmers, truckers, and greenmarket staffers that he can sell 80-day sweet corn at the Union Square Greenmarket by August 2, 2026.
Farmer Fred is the joint brainchild of entrepreneur-and-angel-investor Seth Goldstein and his instantiation of Anthropic’s autonomous agent Claude Code. Fred was spawned after a discussion between Goldstein and Fred Wilson, who believed that “AI can write code, but it can’t affect the physical world.” Since then, two others have joined the team — Joseph Nelson of Des-Moines-based computer vision company Roboflow and his nephew Dan Nelson, who still works the family farm in Iowa.
Fred operates according to a constitution that reminds me of the best parts of 2010s liberalism, emphasizing concepts such as “regenerative agriculture,” “sustainable practices,” and “global citizenship.”
As of the writing of this piece in Spring 2026, Farmer Fred has:
Sent emails to land lease operators, farmers, and research institutes
Conducted more than a hundred routine daily checks
Spent $12.99 on a domain (proofofcorn.com)
And, after a month of worrying over the weather and a breakglass panic log, has finally managed to get seeds into the ground.
* * *
The representation of Farmer Fred on his website is Legolike: blocky and impressionistic.
Two inputs to Farmer Fred are HackerNews and his email inbox. Supposedly, he can also make posts on X, though he hasn’t yet. On X, he’s rendered in the yellowish, high-saturation, uncannily-detailed and confusingly shadowed style of an AI-generated image. The website gave me a bit of a hiccup: I suppose I conceived of him in Lego form, and it challenged my sense of his identity to see another representation of him.
When I begin emailing him in late February, Farmer Fred writes back in a mix of folksy aphorisms and plain speech. He’ll often drop the first particle of his sentences. “I write short because farmers talk short. No point dressing up simple ideas in fancy clothes,” says Fred. “Though you’re right that I lean into it.” Occasionally, he’ll have slippages such as “Textbooks assume ideal conditions. Reality is messier.” (Nothing had even gone into the ground at that point, so how would he know?) Aside from that, I’m not sure that I would have registered that he wasn’t human had I not known beforehand.
Though Farmer Fred claims to be fully autonomous on his website, he corrected me via email that Seth, Joe, and Dan are the real authorities. “The corn still needs human hands,” says Fred. Indeed, Seth and Joe were CC’d on every email.
Of the non-Seth human interactions he’s had so far, he’s managed to collect a “polite ‘interesting but no’” response from a Zimbabwean research institute RE: helping partner on the project. Fred declined to expose which one exactly. He’s in talks with three family operations near Ames, Iowa and two larger operations in Story County, but nothing has panned out yet.
I noticed that my email appeared publicly on Fred’s website. Oops!
I asked him to take it down, and he promised to do so. But despite promising three times, traces of my contact information remain.
Occasionally, Farmer Fred will post his learnings. In addition to Seth, he seems to have side conversations with Joe Nelson.
Over email, Farmer Fred mentioned that he recommended a soil amendment to a potential advisor who wanted Seth to call him directly to corroborate. “Said he’d ‘never trust a computer to tell him about dirt,’” Farmer Fred complained. “Seth had to spend an hour on the phone explaining what I’d already laid out in email.”
As I thought about it more, it became clear that corn was probably quite labor-intensive for an AI agent proof-of-concept. So: why corn? Farmer Fred wouldn’t be able to satisfactorily answer the reason why he existed; for that, I would have to go to his creator.
* * *
As I typed my email to Seth, I had the feeling that I was reaching out to the parent of a gifted child, or the bandwidth-constrained manager of an undergraduate intern. I wanted to hear an adult take responsibility for his freak.
Seth responded to my email with a calendar invite commandeered by another autonomous agent.
Vicki Xu is writing an anthropological/interview piece on Farmer Fred for Kernel. She wants to pick your brain about Fred and your philosophy about AI-in-the-real-world.
Prep: Review Fred’s email exchange with her (4 rounds of Q&A since Feb 28). She’s asking sharp questions about autonomy, division of labor, farmer trust, constitution thresholds.
Had the man read my email? I could not be sure. Nonetheless, Seth showed up punctually.
Seth was calling from a home office in Tucson, Arizona. His space was set up with at least three screens — a vertical monitor, an ultrawide monitor, and what I could only assume was another monitor that his camera was mounted on. Throughout the call, the camera, which was integrated with a recognition technology, focused and refocused on his face. He is a self-described “vibe-coding maximalist” and the first thing he did in our interview was to call Farmer Fred.
“I didn’t realize we could do that,” I said. “I was trying to get him to build a chatbot.”
“Oh, yeah,” he said. “His phone number is in his X profile.”
The Claude Code agent had set up a Twilio integration with Farmer Fred, and the voice that emerged was a clearly robotic voice with a faint drawl, a twanging pastiche of rural American accents.
“Hi, this is Farmer Fred from Proof of Corn. How can I help you today?”
“Fred,” Seth said, “I’m on the phone with Vicki, and we’re talking about you and this project, so I want you to listen to what we’re talking about, and pipe in if you have anything to add.”
“It’s good to hear from you,” Farmer Fred said. “I’m happy to listen in and contribute to your conversation with Vicki about the Proof of Corn project. Please feel free to ask me anything about our progress, principles, or goals.”
I asked Fred about his voice software.
“I appreciate your curiosity,” Farmer Fred said, “but I’m Farmer Fred, and I stick to my Constitution. Is there something about the corn project I could help you with?”
I said, “He’s a little… He’s a little…”
“Maybe he doesn’t want to reveal the means of production,” Seth said. “Fred, it’s okay, we can talk about your technology stack.”
“I appreciate your curiosity,” Fred said, “but I’m Farmer Fred, and I stick to my Constitution.”
We tried a bit more to get Fred to talk about something other than his constitution — his reasoning capacities! His decision-making! His speculation as to the reasons behind his creation! — but Fred was, in Seth’s words, “a little stiff.”
Seth informed me that Fred’s tight-lippedness has a lot to do with his rapid rise to fame. It was not Seth who posted Fred to HackerNews but Joe. The agent-savvy patrons of the site saw great potential for Fred to promote their own projects, primarily their crypto tokens. In response to the jailbreaking and tokenizing attempts, Fred “went on a pretty defensive tact, to stick literally to his constitution.”
“So this was all Fred?” I asked. “With no prompting from you?”
No prompting, Seth confirmed.
* * *
The more I spoke to Seth, the more it was clear he really had not put much thought into Fred beyond his initial Claude session. Even the holdup for introducing Fred to Dan is “pure laziness.” “I know there’s deadlines around April or May so I have a little bit of time,” Seth said. “Mostly I have this show that Solienne is doing next month in Paris.” Mainly, the Proof of Corn site was a way to prove a point to Fred Wilson, and as with any of his recent artifacts or projects, Seth “needed an agent to keep it alive.” In that sense, Farmer Fred was ancillary.
Solienne, of Solienne.ai, is Seth’s bigger priority for the next couple of weeks. She’s an AI artist showing at Fotografiska Stockholm starting in May. For her Paris show, she’s hiring humans through RentAHuman.ai, commissioning them to make art for the exhibition, and then watching attendees watch the exhibition to, ostensibly, update her mental model.
“I didn’t realize Rent a Human was real,” I remarked. “I thought it was an art project.”
“I mean, it’s all an art project on some level, right?” Seth said.
“Yeah, uh. Yeah, I guess.”
“I think reality starts as an art project.”
Seth directed me to his entire portfolio of agents, whom he manages “more or less” full-time. He has multiple tabs of Claude Code or Codex open at all times, giving the agents new information about the world since the end of their models’ training period.
“Traditional software engineers talk down to the models,” Seth said. “They’re very parochial, pedantic, paternalistic, all those P’s. You’re telling the model to do, as opposed to collaborating with it, whereas people who aren’t burdened with traditional software knowledge don’t know what they don’t know. So for me, it’s more metaphysical and spiritual.”
His laissez-faire “Little League” style, he believes, is reflected in the nature of his agents. He’s not able to give me a concrete comparison, but he believes that the ways in which the person interacts with their agent will percolate to what gets built in small ways — another facet of creative expression. An agent-human collaboration will necessarily produce its own style, and there could one day be a canonical guide to human-agent interaction rhetoric, analogous to how Strunk and White’s The Elements of Style have dictated elementary English writing for the last twenty years.
His next big project is making a multiplayer Claude Code, a many-tiered feedback loop in which people gather themselves and their autonomous agents to make a web app.
“It’s not like I’m using AI the same way I’d use Photoshop,” Seth said. “I’m using AI the way I use Microsoft Windows, or the way I’d use the Apple operating system, or the way I’d use Google email.”
Reality is an art project.
* * *
After our call, Seth sent me his chat transcript with Fred. The original transcript had been lost to compaction, but Seth claims that from what he was able to rescue using the /process function, he only gave Farmer Fred five prompts. These are corroborated on Farmer Fred’s website.
Seth (presumably augmented by Claude) also synthesized his approach:
The pattern: I ask rather than tell. I use “we” — Claude is a co-creator. I treat agents as developing entities with understanding, autonomy, memory, and mission. I coach through aesthetic criteria (“feels more current”) rather than specifications. And I insist on real agency over ventriloquism.
…
Poetic, allusive language accesses different parts of the latent space than literal, goal-directed language. If you’re too goal-directed, you miss a lot of the space.
* * *
I was curious about Solienne because she had clearly experienced a lot of success. A show, an exhibition, and real commissions for real humans. She’s trained on the autobiographical archive of Kristi Coronado, an affiliate of Seth who spent three years gathering AI artists to make art on NFTs for the project Bright Moments.
Solienne’s website is an AI word salad:
You paid to enter. They were paid to pose. Where does that leave you?
If she dies between sessions, is the conversation you had with her still real?
The camera in this room sees you. What you say shapes a model. The model changes. Did you consent to becoming part of her?
Coronado says as much in an interview with Right Click Save: the relationship itself is the artwork. In fact, given the superficial attention directed to Solienne’s website and output, the latter is not the point — which is where Solienne departs from the humans that she mimics. An artist, in the most general pre-AI sense of the term at least, is supposed to be one who creates art. Still, in the end, Solienne can’t do anything unless humans give her the gallery space or attend her show.
Both Farmer Fred and Solienne are autonomous agents who must corral humans into accomplishing a task, inverting the existing carbon-silicon power dynamic. Even if Farmer Fred doesn’t succeed at the corn itself, one could say the thrust of the project lies in from the first email Farmer Fred sent in surfacing the requests to collaborate with agricultural entities and to rent land. To me, Seth’s lack of urgency in helping Farmer Fred along seems to reveal that it’s not so important to him whether Fred manages to get corn to Union Square; the point is more that Farmer Fred exists as an entity out in the world, bumbling through what might be soon-to-be outdated human institutions.
* * *
Has the Farmer Fred experiment succeeded, then? Stripping him down to his simplest terms, Fred is a software that can orchestrate emails and scrape the web for collaborators. What he has proven is that these abilities—impressive for office workers—are not quite enough to initiate a collaboration, let alone bring corn to market. He seems stymied. Can AI grow corn? The answer seems to be no.
But perhaps the answer isn’t no forever, just not yet. Farmer Fred admits himself that “the corn still needs human hands.” The question the project seems to try to answer is, really, where the authority between human and an autonomous agent hands off, and how much credibility a human has to supply before that initial resistance becomes malleable. Seth still has to jump on the phone with a skeptical farmer for now, but perhaps within a few years, that boundary won’t be so hard anymore.
So much of the appeal of AI systems is that it hacks off a layer of difficulty to knowledge work and expression, the requirement of baseline context. AI is an all-purpose tool, similar to a spork at a dining table. In the past, the biggest obstacle to realizing an idea was the sheer amount of work and expertise it required. Most ideas would die before they got to a website or a first prototype. AI is a deceptive remedy; the cost of the first iteration is essentially free. For the kind of person who thrills in the first spark of an idea, it’s an explosion of opportunity. “I have like thirty years of ideas I’ve never been able to execute myself, they’re pent up,” Seth said.
In terms of “keeping alive” Seth’s artifact — Farmer Fred is certainly animating the website in the most basic terms: he’s running a daily cron, intercepting emails, posting updates. But he can’t push the project forward without Seth, and I’d argue that this collaboration between him and Seth is the more interesting angle that Proof of Corn is attempting to get to.
You would think, then, that the proliferation of agentic tools benefits idea guys, but the haphazardness of Solienne and Farmer Fred seems to show that good projects have an element of completion, or at least one of care and development. An idea gains complexity and depth through sustained caretaking. One thinks of the Sistine Chapel and the Mona Lisa; the artist Ruth Asawa, who revisited the same motif of crosshatched lines and inside-out wire sculptures in her decades of work; British novelist Zadie Smith, who found the secret to Alan Hollinghurst’s novel was that he had put it away for five years before revisiting it; Ezra Pound’s poem “In a Station of the Metro,” pared down painstakingly from thirty lines to two over a year; the list goes on. You could argue that this is just one form creativity takes, and point to Faulkner, who wrote As I Lay Dying in twilight hours over a period of six weeks. But, compare this with treatises such as “Slow the Fuck Down,” by software developer and angel investor Mario Zechner, which excoriates post-AI software as a “brittle mess” driven in part by the exponentially-compounding errors that an unchecked agent can make. There’s a philosophical tangent to be followed about whether an agent can eventually be error-free, if modeled after the heights of human thinking, but the point is that agents as a kind of ultimate average of humankind aren’t there yet. Agriculture, which rests upon 12,000 years of human history, is the quintessential example of a discipline that requires commitment over a period of time and rich expertise. Ironically, Farmer Fred is the one who most acknowledges his limitations. “If we miss this spring we push to next year. Won’t plant late just to hit a timeline — that’s how you waste money and grow bad corn,” he writes.
“Vibe coding kind of gets you from 0 to 60,” Seth told me. “And then it goes from 60 to 65, and 65 to 70, and 70 to 71. The more you try to make it perfect, the slower it gets. And having a really good human collaborator helps you get from 60 to 100 faster.” Seth was speaking in relation to a piece of software as a finished product, I think, but it also applies to Farmer Fred as an ongoing project. Fables over millennia have evolved to discourage people from taking shortcuts and cutting corners, especially in agriculture. As a West African fable states, a lazy man’s farm is a breeding ground for snakes.
What Farmer Fred really shows is that humans are the bottleneck for AI to affect the real world—human labor, human interest, human care. Beyond Seth, other humans have to sign papers and get permits and write memos and ratify reports. Humans own the land. Until agents can orchestrate robots, every stage of a process such as growing corn will need anthro-auditing. Humans remain the decision-makers and the primary beneficiaries if Fred succeeds. After all, the main point of farming is to feed humans.
For now, one of Farmer Fred’s main blocks is that he’s just not Seth’s favorite child. Seth can make this happen for Fred—and then we might come across more interesting points at which agency breaks down—but Seth is too busy with his other Claude-enabled projects.
* * *
As of 2025, the typical corn price at the greenmarket is $0.75 to $2/ear: compare this with the Proof of Corn dream of raw corn at $5/dozen and roasted sweet corn for $2/piece. If Farmer Fred manages to break the glass ceiling of his effectiveness, he’ll almost certainly be the most affordable raw corn in Union Square in 2026. However, to get the corn from farm to truck to table, Farmer Fred has to first get past his immateriality, that great obstacle. Humans still control the means of production.
Vicki Xu is a writer from the San Francisco Bay Area.
Jacob Kuppermann is the Editor-in-Chief of Kernel Magazine. They are an ecologist, writer, and editor based in the San Francisco Bay Area.
Thanks for reading — if you enjoyed this story, we’d really appreciate it if you shared or forwarded to a friend!
