AI’s Biggest Opportunities May Be in Drugs, Crops, and Homes

AI can make it feel as though anyone can build anything—and that every new product might be obsolete tomorrow. Shaan Puri starts from that tension: the rush to build with AI looks crowded, but sitting down to decide what to make can be paralyzing. His answer is to look beyond digital products and ask where AI could change the production of physical things.

Puri thinks AI-generated content and software tools risk becoming “a race to the bottom.” That is not a claim that software no longer matters; it is a reason to look for opportunities where AI could change what is possible in biology, agriculture, manufacturing, or robotics. He calls this “AI in the world of things.”

The distinction is also a bet on timing. Puri says that five years earlier, promising places to build included model companies and products for generating video, images, and text. The harder question now is what will look obvious five years from today. His answer is not a single winning company, but industries where physical work has been expensive, slow, or difficult to scale—and where AI and engineering might alter those constraints.

The strongest examples he offers are drug development, crop breeding, and homebuilding. Each illustrates his central proposition: the opportunity is not simply to add a chatbot to an existing business, but to change how a consequential physical product is designed or made.

Puri’s first example is medicine. He describes an X post by Jeff Tang listing people trying to treat cancers affecting themselves or someone close to them. One entry concerned a GitLab founder who, Puri says, used technology to sequence his tumor, investigate possible drugs, and pursue personalized treatment. Puri says the founder cured his own cancer. The post also listed others searching for individualized treatments for themselves or family members.

Sam Parr adds a personal connection: a friend with cancer had signed up for a service related to the work and was doing well, he says. The hosts also discuss former producer Ben Wilson, who had posted a cancer update. Parr says AI researchers and others reached out after seeing Wilson’s post and connected him with people who could help. Puri says Wilson was still dealing with cancer, while describing a dramatic improvement after treatment.

Puri’s business case rests partly on the time, cost, and risk involved in developing drugs. He describes a typical drug as taking about ten years and two billion dollars to make, with an 80–90% failure rate. He argues that those economics have historically limited who can pursue the work. If AI changes the research process, new kinds of drug and health companies may become viable.

He points to Eli Lilly as a large company investing heavily in AI, comparing its investment in computing capacity to financial firms building infrastructure for their own work. He also cites Isomorphic Labs, a Google company associated with Demis Hassabis, whose displayed mission is to use frontier AI to advance drug discovery. Puri sees an opening to build companies around new scientific capabilities.

For someone choosing where to work or build, medicine is one of Puri’s clearest examples of a technology window opening around a difficult problem with unusually high stakes. The personal stories give the opportunity its urgency; the business case is that AI may change the expensive process behind developing treatments.

Puri’s next example is food production. He points to Nitroduck, a company working on crop editing, and frames the broader opportunity as treating food as technology: deliberately developing crops for qualities such as yield, flavor, disease resistance, or nutrition.

A post displayed during the discussion, attributed to jamexnik, describes a bottleneck in crop breeding. It says developing a new specialty-crop variety can take eight to sixteen years, while consumer preferences, climate trends, and pathogens can change more quickly. The post also describes a challenge with CRISPR: editing often requires putting the tool into a single stem cell and regrowing that cell into a whole plant, a process that must be worked out for each variety. Nitroduck’s stated approach is to deliver CRISPR through pollen, so an edited seed can emerge through a process closer to how breeders already work.

The proposed change is specific. Rather than relying only on random genetic combinations and then selecting useful results, the post says genome sequencing and AI can help breeders decide what to aim for, while CRISPR can make chosen changes. Puri describes the aim as developing fruit that grows better, tastes better, or resists disease.

Parr compares the time involved with the Honeycrisp apple, which he says took 20 years to develop. His example underlines the gap between the everyday availability of food and the long process behind creating a new variety. Puri extends the argument beyond seeds: he also sees opportunities in field equipment, including tractors using computer vision, as well as in developing new crops. He names David Friedberg as an entrepreneur working on improved versions of crops such as potatoes and corn.

Homebuilding is Puri’s most detailed example of applying automation to physical production. He starts with housing costs: for most people, he says, a home is their largest expense, so making homes cheaper could materially improve quality of life. He gives rough figures for his own area—about $600–$800 per square foot for a nice home—and says some companies are building high-end prefab homes at a fraction of conventional costs. He predicts that the cost of building a home could look radically different within ten years.

His main example is ZURU Tech, founded by Nick Mowbray and his brother. Puri describes the company as applying its experience in automated manufacturing to housing in China. Parr cautions against taking “printing homes” too literally: he says it is better understood as a more efficient way to produce and assemble concrete components.

Puri compares the approach to factory production of phones. A digital design informs the production of precise parts, which can then be assembled on site in a modular, Lego-like way. The company’s website, shown during the discussion, displays software for designing homes alongside images of factory equipment and modular construction. The comparison is not that a house is an iPhone, but that a factory can bring repeatable production and automation to work usually done on a building site.

Puri says the company tested a large home in an earthquake simulator, running it through major earthquakes and refining designs through repeated trials. He says workers live in homes outside the factory while the company iterates on prototypes. He also says the Mowbray brothers put about a billion dollars of their own capital into the effort and took no outside funding.

That history matters to Puri because he sees the project as a manufacturing undertaking, not just a software idea. The brothers’ earlier experience was in toys. Puri argues that the precision required to manufacture toys at scale can transfer to other products, invoking Lego as an example of parts that must fit together consistently. He also recounts the logic of simplifying a car component into one molded piece rather than assembling many parts: a manufacturing process that looks difficult at first can change production if the machinery and design are made to work.

ZURU Tech is his clearest illustration of the physical production he has in mind. The claim that this approach could sharply reduce homebuilding costs remains Puri’s prediction; the example he gives is a factory process built around digital design, modular assembly, and repeated physical testing.

Puri’s other examples range from specific applications to early intuitions. In warfare, he imagines a shift toward small, inexpensive, autonomous systems that can be made in volume and used against more costly equipment. He names Anduril and mentions a company working on underwater systems. He says he does not know exactly what is happening on current battlefronts, but suspects that warfare is moving toward drones fighting drones.

Parr brings up a story he recalls reading about a raid in Venezuela. He says reports described guards being affected by an unknown force, characterized as sonic or directed energy, and recalls that Donald Trump called it a “discombobulator.” The account remains uncertain in the exchange; Parr presents it as something he had read, not as a confirmed explanation of what happened.

Puri also sees scope to extend self-driving systems beyond cars. He says his Tesla drives most of his miles, with his interventions largely in parking lots, and points to forklifts in warehouses and machines for cleaning skyscraper windows as further applications. He imagines autonomous equipment maintaining ships and bridges—work that takes substantial human effort today or, in some cases, is not being done.

Weather and oceans receive only a brief mention. Puri recalls a discussion of ocean technology with Will O’Brien and compares the area’s prospects to the rise of space technology. He does not name a particular ocean application, but sees ocean technology as approaching a period of increased activity.

On new materials, Puri is candid that the subject is beyond his expertise. He says people in San Francisco are trying to invent materials or produce substances in ways that sound to him like alchemy, including making gold from another material. He finds material science interesting, while acknowledging he would need to spend time with people who understand it.

His final category is “human body hardware”: devices that connect machinery to the body. He names Neuralink and describes a friend working on Atom Limbs. Puri contrasts the stiff, basic prosthetics he says many people receive after losing a limb with the possibility of a robotic arm controlled by the user’s mind. He sees an engineering problem in bringing the cost down so that people returning from war or living with limb loss can access a more capable prosthetic.

Parr takes the list of potential industries as a reason to be optimistic about the United States. He connects earlier American growth to a combination of resources, infrastructure, food production, and principles such as private property and capitalism. He points to the country’s mineral resources, Great Lakes shipping, farmland, and a large internal market, and sees a possible parallel in current investment in American data centers.

Puri raises the doubts he hears from others: hostility toward companies, billionaires, and AI; uncertainty about whether the United States can rebuild manufacturing capacity after years of outsourcing; and the possibility that AI may either displace workers or prove overhyped. Parr calls this uncertainty and fear rather than simply pessimism. He argues that leaders should be able to support capitalism and technological progress while also taking problems such as expensive housing seriously. Puri wonders whether media incentives reward polarizing voices over common-sense arguments.

The exchange qualifies the optimism behind Puri’s business thesis. Resources and entrepreneurial energy are not enough if the country cannot turn them into production. Puri describes national success as a combination of factors—resources, culture, incentives, competition, and the ability to attract people—rather than the result of any single advantage. Parr remains bullish on the United States as a place where these industries could grow, provided the country does not get in its own way.

Puri’s advice is aimed at people deciding what to do next, especially those who have sold a company or become free agents. He says they often begin with a narrow set of questions: What can I do with people I already know? Which jobs are nearby? What am I good at, and what is the next step in my current industry? His alternative is to think less like a worker choosing among familiar roles and more like an investor looking for the right market.

He borrows the “right body of water” idea from Mark Pincus, who used it to describe the value of being in social media during its early growth. Puri’s exercise for his cousin, who had recently sold a company, was to imagine being given $10 million to invest in five companies and keep the profits. Which five would he choose? Puri says the exercise helped his cousin notice industries he had not been considering and start investigating them.

The point is not that a person needs to be an expert before exploring a new field. Puri says he had not deeply researched every area in his presentation; he had seen enough to recognize potential and then look for a specific angle. People may need to widen their view before they can make a good choice about what to build.

Parr says that many of the projects sound exciting even if he would rather watch someone else undertake them. Puri notes that both men once moved to San Francisco without knowing exactly what they would do there, but believing it was the right place to find out. That example connects geographic mobility with the larger career argument: move toward the environments and industries where the action may be, rather than assuming that past experience should determine the next step.

For Puri, the practical exercise is to widen the map, identify fields that may have a strong tailwind, and investigate until a particular opportunity comes into focus.