OpenSourceAg #1 - Recipe books and basic biscuit recipes

What if silos were just for grain?

When you set out to do anything—hang a picture, bake an Anzac biscuit, write a newsletter, hack together some code—there are many base level jobs you avoid doing. This is simply because someone has done it before, openly shared their experience, or perhaps offered a product to make your life easier. Even the products offered (e.g. beehiiv) are themselves built from tools and services that make putting together a scalable newsletter company an easier experience.

The same is largely true in agronomy, with regional factsheets, variety guides, weed control tools generally available to farmers or would-be farmers. Of course, this doesn’t mean you’re a farmer overnight! Equally, me having access to ChatGPT and Claude does not make me a professional developer, however, it improves my rate of development and on a global scale, lowers the barrier to software-based innovation.

Remember the weathered recipe book in your kitchen? It tells the story of years, decades, perhaps generations of iterative developments in cooking for your family - but built from a base recipe that was initially shared. This is exactly what open-source tech in agriculture is. It’s taking the same principles from recipe sharing, to software, hardware and data use relevant to precision ag. It’s about breaking siloes (without grain loss) and enabling better collaboration (there’s even a silo recipe from 1919).

So why is this needed in agtech? Well, we’re slowing ourselves down by rehashing the same old basic code, collecting the same data and getting caught up in boilerplate work. Developments in the broader tech field for cheaper hardware and easier code, mean open-source datasets are the biggest barrier to many vision, or text-based applications for precision ag. Check out the recent announcement of the Jetson Orin Nano Super by Jensen Huang of NVIDIA to add to their embedded computing range. The number of Trillions of Operations Per Second (TOPS) you get for your USD has increased incredibly these last 7 years.

With easier access to performant edge computing and LLMs like ChatGPT, Claude, Cursor and CoPilot helping with code—the biggest barrier to innovation is increasingly data. Flipping this on its head—the most impactful opportunity for investment by growers and industry to drive innovation is the collection and publication of open source datasets. But more on that in future issues.

Welcome

And with that, welcome to the OpenSourceAg newsletter, one that hopes to highlight and connect the recipe sharers, projects and open-source developers building tools everywhere. It started out as an idea two years ago, and after a PhD thesis, moving countries a few times and chemotherapy, it’s finally ready to launch.

Thank you for the unexpected amount of interest. With new open-source projects projects appearing in all aspects of ag-related tech (or agtech), I have been wanting to start building a place and community, to share, learn and explore new collaborations. Slightly later than I had hoped, some setbacks of late, but 2025 is the year, and this newsletter is a place to start. With over 325 subscribers, it's exciting to see how many of you from such diverse parts of the industry and globe are equally passionate about discovering more in open-source development for agriculture. Plus, I’ll be sharing some of the latest ag tech research, OpenWeedLocator updates, and longer form posts of various Bluesky/X/LinkedIn creations.

Most importantly though, I would love to make this as much of a conversation as possible. If you have any comments, ideas, questions or anything else, feel free to leave them below, reply via email or on social media and I will do my best to get back to you and share it in this newsletter. If you have your own project or know of others, let me know and I can add it to the Featured Project space below, and there is the accompanying OpenSourceAg repository with a growing list of datasets and projects to sift through.

And who am I exactly? Well, I’m no chef, not really a developer, hardly a farmer (or farmer’s son) but working in precision weed control as a researcher. As a self-taught coder, starting back in 2018 with Python, I have benefited greatly from open-source libraries and people generous enough to share their experience. And I am quite certain that the machine learning field would not be where it is today if it wasn’t for openness and sharing being the norm.

So, in this newsletter we’ll cover:

Table of Contents

• What if silos were just for grain?

• Welcome

• The 7 principles of open-source development

• Practicalities: How can you get started with image …

• Featured Project | OpenWeedLocator 

• Interesting Reads

And don’t miss the next edition, make sure you subscribe so it lands in your inbox every second Tuesday (#AgtechTuesday). I’ll be presenting some ideas on how agtech companies can benefit from opening their own work and the nuances to the approach. How many ag companies have active and useful Github repositories? But until then, check out the rest below.

The 7 principles of open-source development

If you search for anything related to machine learning, you’ll be bombarded by a plethora of projects, guides, documentation, courses, tools—usually shared by creators openly on the internet. I learnt to code through the PyImageSearch blogs back in the day and like many students, with Andrew Ng’s Machine Learning Specialisation on Coursera. I’ve used Google’s Tensorflow, Meta’s Pytorch, Chollet’s Keras and PJ Reddie’s YOLO architecture—the list is endless. It is easy to assume that this field has always been open, and receptive to open-source sharing of tools and ideas. Yet, that isn’t the case.