WindBorne Systems

At the end of Elicit's board meeting last week, founders Andreas Stuhlmüller and Jungwon Byun said: "One of our goals for the year is to automate the entire company. If we all go on vacation at the end of the year, the company should, in many ways, keep running without us." More on this, and what others in the Footwork portfolio (e.g. WindBorne Systems w/John Dean, Anything w/Dhruv Amin, Marcus Lowe) are doing as they rethink how their companies operate in the agentic era, in my latest post: https://lnkd.in/gWQZ5Qcg

Bigger building = bigger constellation

Let's talk about what it takes to create weather certainty. Here are three opportunities to chat with us this month!

AMS 2026!! Over 6000 people gathered in Houston, Texas last week for the American Meteorological Society's annual meeting. For me, this is an event that I look forward to every year where I get to gauge what's hot in the weather enterprise and catch up with old friends. This year, a very hot topic was AI-driven weather prediction. While this has been a topic for many years, several things were now obvious - AI driven weather prediction has already improved weather forecasts that consumers see every day on phones, and that improvement is continuing - rapidly! Development in AI weather prediction really expanded this year, with many efforts in data assimilation and high resolution forecasting. What’s amazing about this (r?)evolution is that just a few years ago, very few even thought this would be possible, never mind operational. Observations - As always, there was growth in observation collection. Companies are expanding observational data collection with interests in commercial and government business needs. Front and center was the Global Atmospheric Rivers Reconnaissance Program (GAARP) - an initiative to expand the highly successful Atmospheric Rivers Reconnaissance program globally. This highly coordinated effort is incorporating data collection from aircrafts (with dropsondes, radars, GNSS-RO sensors and more), ocean sensors (buoys and submersible platforms) and novel long-range weather balloons to deliver critical observational data to operational forecast systems and provide a foundation for research efforts. Part of my motivation for attending AMS is to share with the community what my company, WindBorne Systems, is doing to improve weather forecasting and insights in the industry. Through 7 scientific presentations, our WindBorne group shared how we are flying a continuous constellation of hundreds of balloons to collect atmospheric observations and assimilate these, combined with the global network of observations, into both government weather forecasts (NOAA’s GFS) and WindBorne’s highly accurate WeatherMesh forecast model. We are excited to be on a path to support the weather enterprise through highly scalable observation collection, weather forecasting, and ultimately decision making. At AMS, I spent much of my time in areas of observations and forecasting. There’s no question that I missed other hot topics, but that’s the wonderful nature of the AMS annual meeting — there is so so much to see and do that you just can’t do it all.

In 2025, WindBorne solved our hardest problem yet: scaling a planetary-scale balloon network that has never existed before. How do you go from 10 balloons aloft at all times to 10,000? There are thousands of little things we solved and learned, but I'd like to talk about the single most important ingredient most people miss: **Bring product designers into operations.** If you are blazing a fundamentally new trail with no playbook to copy from, like WindBorne, you need to rethink the design of every part of the operational system from first principles. You don't just want to hire "operations" people from other industries who will awkwardly try to pattern-match off the wrong patterns. If you are working in the world of atoms, humans will be a key component. If you just leverage hardware and software engineers to design systems that scale, they will generally insufficiently attend to the humans in it. So you need people great at connecting technology to humans. Product designers are precisely those people. Below is an example of a custom inventory tracking page, with beautiful design. I am not talking about hiring product designers involved in the design of the product itself—that is the norm. I am saying that you need product designers involved in building the machine that builds that machine, along with the other types of engineers. Lastly, the people you need to solve your zero-to-one problems are unlikely to be the same as those who solve your one-to-ten problems. The former requires grit and "get shit done at all costs," while the latter requires discipline and systematic thinking. What do you do with the zero-to-one people as you scale? You let them continue with the problems of that nature—if you're building something truly new, there's always another zero-to-one problem waiting.

Last week my WindBorne Systems colleagues and I presented four talks and three posters at the annual American Meteorological Society meeting in Houston, TX. We've made tremendous progress in a relatively short amount of time on sensors development, AI weather forecasting, and AI data assimilation, and it was exciting to share that hard work with former colleagues and the broader meteorological community. Everyone understandably wants to present the best versions of themselves in front of their peers. However, the thing that has stuck with me is how many attendees approached us to commend not only our successes, but also the transparency in our struggles. WindBorne's WeatherMesh is one of the best weather models in the world, but it's not without weaknesses (blur + extremes). Similarly, our AI DA is tremendously exciting, but we're still learning how to most effectively ingest our balloons observations. Being open about these struggles, iterating with the community, and implementing fixes allows us to continually improve and establishes us as a legitimate leader in the field. Science never progresses in a straight line and openly acknowledging that fact builds trust and leads to better outcomes.

Come help us build our planetary nervous system. RSVP to today's live balloon launch (and make your own loonicon too) - https://luma.com/49zxylq9

Africa operates fewer weather stations than Germany (despite being 50x larger). The gap matters. It means farmers are making planting decisions without reliable rainfall data, and communities get storm warnings hours before impact instead of days ahead. Since September 2024, WindBorne has partnered with the Gates Foundation, the Kenya Meteorological Department, and TomorrowNow to change that. Our weather balloons have collected 37x more atmospheric data than traditional systems, flying continuously to capture what's actually happening in the atmosphere and delivering forecasts communities can act on. Read more about our work building weather infrastructure where it's needed most: https://lnkd.in/geUzZAYk