Bindwell Is Using AI to Design Pesticides and Paul Graham Thinks They'll Do Alright

The Macro: The Pesticide Industry Hasn’t Had a New Idea in Decades

Global agriculture loses between 20 and 40 percent of crop production to pests every year. That’s not a rounding error. It’s hundreds of billions of dollars in food that gets eaten, infected, or destroyed before it reaches a human mouth. Pesticides are the primary line of defense, and the industry that makes them has been running on the same basic approach for a very long time.

The way pesticides get discovered is slow, expensive, and increasingly ineffective. A major agrochemical company like Syngenta, BASF, or Corteva will screen hundreds of thousands of chemical compounds against target organisms, looking for molecules that kill pests without killing everything else. The process takes 10 to 12 years from initial discovery to commercial product. It costs over $300 million on average. And the hit rate is terrible. For every compound that makes it to market, tens of thousands are tested and discarded.

Making this worse is resistance. Pests evolve. The pesticides that worked twenty years ago are losing effectiveness as target populations develop genetic resistance. The fall armyworm, Spodoptera frugiperda, is a textbook example. It’s an invasive species that has spread from the Americas to Africa and Asia over the past decade, devastating maize, rice, and sorghum crops. It’s developed resistance to multiple classes of existing pesticides. Farmers in sub-Saharan Africa and Southeast Asia are losing significant portions of their harvests to an insect that current chemistry can’t reliably control.

The big agrochemical companies know this is a problem. They’re investing in biological controls, RNA interference technology, and other next-generation approaches. But the traditional chemical pipeline is still dominant, and it’s still slow.

AI-driven drug discovery has shown that machine learning can dramatically compress the timeline for finding molecules with specific binding properties. AlphaFold proved that protein structure prediction is solvable. Recursion Pharmaceuticals, Insilico Medicine, and others are applying similar approaches to human therapeutics. The question is whether the same logic applies to agrochemicals.

It does. And Bindwell is the company making that bet.

The Micro: Wolfram Research Alumni Meet Caltech Bioinformatics

Tyler Rose and Navvye Anand met at a Wolfram Summer Research Program in 2023. Tyler is the CEO. His background is in ML engineering and computational biology research, with prior work at Wolfram Research. Navvye’s background is in bioinformatics and AI, specifically the intersection of biology and natural language processing. He’s a former Caltech student and World Science Scholar.

Both were named to Forbes 30 Under 30 in science. That’s not usually something I mention because awards are noise, but in this case it signals something useful: the scientific community has vetted these two as legitimate researchers, not just startup founders who read a biology textbook.

They came through Y Combinator’s W25 batch and have raised $6 million from a roster that reads like a who’s-who of early-stage investing: General Catalyst, SV Angel, A.Capital, Character VC, and Paul Graham personally. Graham’s quote about them is characteristically understated: “They will probably do alright. They’re smart and have a good idea.”

The company has four people and is based in San Francisco. They use custom AI models combined with high-throughput laboratory assays to accelerate pesticide discovery. The key phrase is “in-house.” They’re not licensing molecules or building a platform for other companies to use. They’re developing their own pesticides. That’s a different business model than most AI-for-science startups, which tend to sell picks and shovels rather than dig the mine themselves.

Their first target is the fall armyworm. That’s a smart choice for several reasons. The pest is economically devastating, current solutions are failing due to resistance, and there’s urgent demand from farmers across multiple continents. If Bindwell can develop an effective compound against Spodoptera frugiperda, the market is immediate and global.

The website is polished and includes both English and Spanish versions, which suggests they’re already thinking about Latin American and global markets. They’re contactable at [email protected], and their GitHub organization exists, though I’d be surprised if the core IP is open source.

The Verdict

Bindwell is the kind of company that makes me want to pay attention for years, not weeks.

The AI-for-molecule-discovery thesis is proven in pharmaceuticals. Applying it to agrochemicals is logical and underexplored. The pesticide industry’s R&D pipeline is genuinely broken. Resistance is a growing crisis. The incumbents are enormous but slow. A small team with better computational tools and a willingness to develop compounds in-house could build something very valuable.

At 30 days, I’d want to know about their wet lab setup. AI can propose candidate molecules, but validation requires actual chemistry. Do they have their own lab, or are they partnering with a CRO (contract research organization)? The speed of the feedback loop between computational prediction and experimental validation is everything.

At 60 days, the question is regulatory pathway. Pesticide registration through the EPA (in the US) or equivalent agencies globally takes time and requires extensive toxicology and environmental impact data. Has Bindwell started those conversations? Do they have regulatory expertise on the team or board?

At 90 days, I’d be watching for the first proof-of-concept data against the fall armyworm. If they can show a novel compound with efficacy against resistant strains, that’s not just a startup milestone. That’s a scientific result that would draw attention from the entire agrochemical industry.

The risk is execution timeline. Even with AI acceleration, developing a commercial pesticide is a multi-year process. The $6 million in funding is a strong start, but they’ll almost certainly need more capital before they have a product on the market. The investor roster suggests they’ll be able to raise it.

I think this is one of the more genuinely important companies to come out of the W25 batch. Not the flashiest. Not the fastest to revenue. But potentially one of the most consequential.