In this post, we share progress on an internal research project at Homeworld using AI to encourage technoeconomic thinking in climate biotech.
This project was a collaboration between Jesse Lou, Kyle Barnes, Mark Hansen, and Dan Goodwin at Homeworld. I (Dan) will introduce the AI-TEAs work through my personal journey into climate biotech. In this essay, you’ll learn about TEAs, see that AI can make rough TEAs for free, and imagine what free TEAs created at massive scale could do for entire innovation ecosystems.
In addition to reading this blog post, you can:
- Read the complete report.
- Use this very simple TEA template with a corn ethanol example.
- Access the code in this github repository.
- View our database of 88 publicly-available peer-reviewed TEAs is available in this google spreadsheet.
Think of these resources as a prototype that we’re using to get feedback from the community. Does this help you? What more do you need to succeed in climate biotech? Reach out at hello@homeworld.bio to tell us more.
—Dan
My First (Terrible) Idea in Climate Biotech
When I began exploring climate biotech, I was a PhD student in the Synthetic Neurobiology Group at MIT, building molecular tools for neuroscience. As I started searching for climate applications for my skill set, I discovered carbon capture. My reasoning went: binding/unbinding to gases is the core technical challenge for direct air capture (DAC) and a primary function of some proteins—so perhaps protein-based sorbents could be a worthwhile avenue.
This led to a conversation with Klaus Lackner, one of the founding figures in the direct air capture field. In a warm but rigorous discussion, Klaus helped me see what commercial DAC firms look for in a sorbent: roughly, they want to pay no more than $100/kg for a sorbent that can survive at least 10,000 cycles of CO₂ capture and release. Even though cellulases and laundry detergents have shown that proteins can be mass-produced cheaply, the cost and durability of proteins are still 2–5 orders of magnitude away from what’s needed. I was moving in a futile direction.
That was the first climate biotech idea I’d killed because of economics—and it wouldn’t be the last.
Without that conversation with Klaus, I might have wasted years on a cool-sounding but ultimately worthless project. As Devon Stork recently wrote in his essay “When to give up,” killing ideas is one of the most critical skills we develop as technologists.
The Importance of Basic Numbers (and When to Run Them)
At Homeworld, we meet weekly with researchers who have big ideas to make an impact. Every conversation that passes the first gate —“Is this a worthwhile problem to solve?”— inevitably faces the next bottleneck: “Well, have you run the basic numbers to show you understand how your idea would actually be deployed?”
I often see people going through the same lengthy journey I did, and I believe Homeworld can build something to help everyone move faster. So we set out to see if new AI tools can generate formalized techno-economic analyses (TEAs) to help biotechnologists explore their ideas from petri dish to pilot plant.
What is a Techno-Economic Analysis (TEA)?
A TEA is the tool that brings a technical idea to life by rooting it in a real-world system. From mining to carbon capture to pollution mitigation, anything with planetary ambitions must have a model. A simple TEA often lives in a spreadsheet but more complex models will be built in expensive software. The TEA captures the major practical aspects of a facility and asks questions as shown in the figure below:
TEAs are the backbone of:
- Project financing decisions
- Venture capital decisions
- Many philanthropic investments
But it can’t be just financing professionals who speak TEA’s language. Scientists also need TEAs to protect their most valuable resource: time.
How TEAs Work
A TEA often starts by drawing the flow diagram of a pilot-scale process and assigning numerical assumptions to each piece of the system. The result is a model that helps you imagine what would need to be true for an idea to be worthwhile.
Pro tip: TEAs can be as simple as a “back of the envelope” or “Fermi calculation” (as Van Gelder 2023 calls it) or as complex as the 147-page NREL gold-standard analyses (Fig. 1).
I was lucky that my carbon capture idea was so far off in terms of cost and durability that even a rough mental TEA was enough to rule it out. But for many ideas, a spreadsheet is the best starting point—and for large-scale processes, you may need highly detailed TEAs, such as the seminal Corn-to-Ethanol 90-page analysis.
TEA Services Are Increasing (and AI Integration is Inevitable)
Most investors won’t invest in deeptech without a basic TEA. And top investment firms like Breakthrough Energy emphasize TEAs. Their Breakthrough Energy Fellows program pairs Business Fellows with Technical Fellows in about a 1:2 ratio. These Business Fellows, often from Harvard Business School, McKinsey, or relevant industries, help the Technical Fellows imagine commercialization strategies from day one.
Jesse Lou is one of these impressive Business Fellows. He has an MBA from Harvard, worked at McKinsey, and has years of experience in product management and heavy industry. We’ve known each other since grad school, and he’s shown me when and how to use TEAs effectively.
Using AI for TEAs: Early Experiments
In summer 2023, Jesse and I started wondering: Can AI build TEAs for us? We asked ChatGPT (specifically GPT-4 with “Code Interpreter” access at the time):
“Give me a gold-standard TEA for the technical approach in this paper.”
The idea originally came from Darren Platt in an online conversation. He shared a recent biomanufacturing paper and got a basic TEA-like analysis out of GPT-4. Our request yielded a simple but functional economic model. Jesse then took it to the next level:
“The point of a TEA isn’t just the final sum; it’s discovering how sensitive the result is to changes in each assumption. Let’s do a sensitivity analysis.”
We asked GPT-4 for a “tornado plot” to see how unit changes in the most important variables would affect the final economics. We ended up getting a volcano plot—plus a surprisingly good 2D sensitivity plot. The results were basic, but it was clear: large language models already possessed substantial TEA capability.
The Bigger Vision: Free, Open-Source TEAs for the Bioeconomy
We started thinking: What if AI could generate TEAs at essentially zero marginal cost for every process in the bioeconomy? What if we could make:
- A generative tool for ideation: Not just the standard reductive step used for financing.
- A dynamic, interactive version of a 20-year-old static figure (Fig. 4), where anyone can tweak parameters and see results in real-time.
We decided to try building a single biomanufacturing TEA from just a paper. Over a few months, Jesse, Mark Hansen (a creative coder (YC ’22) with a deep background in engineering and startups), Kyle Barnes (designer/programmer focusing on data-driven activism), and I worked together:
Soon, Mark had a Python script capable of assembling a TEA automatically. And if we could create one TEA for one process, why not create a dashboard for the entire bioeconomy?
Kyle drafted the prototype interface that let you explore countless parameters, feedstocks, outputs, locations, prices, etc,. all in one place. He crafted a demo which shows the biomanufacturing space condensed into a graph structure in which the links are related to the number of overlapping manufacturing steps.
While this remains a work in progress, we now see both the potential and the challenges. Running GPT-4 for these TEAs cost about $0.25 each, which could add up. Still, the runway for advancing AI-driven TEAs is long.
Interested in helping? Email us at hello@homeworld.bio.
AI TEAs Team – Where are they now?
- Dan Goodwin – Dan presented this work at the Novo Nordisk Automated Scientist Conference. Homeworld also held a workshop to discuss these models with industry practitioners.
- Kyle Barnes – Kyle is a fellow at Schmidt Futures.
- Mark Hansen – Mark is still working on TEAs w/ the Esposito Lab at Columbia.
- Jesse Lou – Jesse is now the CEO/Founder of Conductor Labs and collaborating with Homeworld on open source TEAs for mining. Check out his episode on the Climate Biotech Podcast with Geobiotechnology Lead Jayme Feyhl-Buska!
Closing Thoughts
Techno-economic analyses may sound esoteric, but in climate biotech—and especially for planetary-scale solutions—they’re essential. Whether you’re killing an idea early or scaling a viable technology, TEAs keep teams realistic and guide valuable resource allocation.
With AI tools now increasingly capable of generating these analyses, we may be on the brink of a new era of rapid ideation, faster iteration, and a more democratized approach to assessing climate biotech solutions. I hope this blog and our shared resources help you see the possibilities—and maybe inspire your own projects.
Questions or ideas? Don’t hesitate to reach out at hello@homeworld.bio.
—Dan Goodwin, Homeworld
Lastly, we want to acknowledge Yoel Cortes-Pena, the creator of BioSTEAM, the leading open-source library for high-quality techno-economic analyses. BioSTEAM significantly informed our own explorations and will likely continue to be a cornerstone in future TEA work.