Podcast Extra: The Future of Agriculture Podcast - AI For Regenerative Agriculture With John Kempf

Regenerative agriculture is complicated, despite what some in the media want to tell you. It's not just a matter of cutting out chemicals and tillage and living happily ever after. To really make things work requires decisions made with complex biological interactions in mind. Nobody that I'm aware of has more experience or more data on regenerative agronomy than John Kempf. He's now leveraged all of that into a new artificial intelligence regenerative agronomist called Fieldlark AI.

What technology can do and what it should be doing, in my opinion, is guiding humans and amplifying their strengths. If their strengths are intuition and observation, then there should be scouting apps that guide the farmer or the crop scout or the agronomist to ask better questions and to be more observant.

John is the founder of Advancing Eco Agriculture, and he returns to the show today to talk about why he believes this Field Lark AI tool is an important step toward expanding regenerative agriculture.

We rely and use SAP analysis very intensely. It used to be this consumed tremendous amounts of time and energy and mental bandwidth, and now we've shifted where the recommendations are no longer human generated, they are now human reviewed.

John Kemp returns on today's Future of Agriculture podcast. Hello fellow ag nerd. Thank you so much for joining me for another episode of the Future of Agriculture. My name is Tim Hamerich and every week you and I get to hear from the farmers, founders, innovators and investors shaping the future of the ag industry. Today we're joined by John Kempf, founder of Advancing Eco Agriculture. John first joined me on the podcast about two years ago for episode 386, which was titled Challenging Assumptions about Regenerative Agriculture. That episode was pretty widely shared and provided a ton of feedback for me that I found interesting. So I was eager to bring John back on, especially to discuss this new AI regenerative agronomist tool they've just launched this past year, which is called Field Lark AI. As I hoped would happen, John and I also get somewhat philosophical at times about things like technology use and ethics, and we talk about a separate but similar project he's embarked on of building a digital clone of his knowledge and expertise. We're going to talk about what that process looked like for him and how he's using digital Clone John in his daily life. And for you agronomy nerds out there, you definitely want to stick around to the end because we'll get into some fascinating insights about nitrogen that John and the AEA team have been helping their customers take advantage of. First though, I do want to read John's background here from their website because number one, it's just fascinating but also it really informs his perspective and his focus with aea. John Kemp is an entrepreneur, speaker, podcast host, leading crop health consultant, and designer of innovative soil and plant management systems. He founded Advancing eco Agriculture in 2006 and serves as Chief Vision Officer and Executive Board Chairman. In the early 2000s, Amish farmer John Kempf was on a quest to save his family's fruit and vegetable farm in Northeast Ohio. After years of intensive pesticide use, pest and disease pressure was unmanageable and unaffected by continued chemical applications. After years of extensive agronomic research, first hand growing experience and close mentorship, John found new crop management methods. The practices he discovered and the products he created were the beginnings of what is today advancing eco agriculture. So very interesting stuff here and I think you're really going to enjoy this conversation with John Kempf. I'll drop you into the conversation where he's explaining some of his philosophy that guides their focus at advancing eco agriculture and in turn this fieldlark AI tool that we're going to talk about. Here's Jon Kempf.

I realized early on that as we were taking this very kind of different framework or different approach to thinking about plant nutrition and agronomy and soil and plant interactions and let's just describe what that is when I say different. Mainstream agronomy, historically this has shifted now in the last decade or two, but historically mainstream agronomy focused on balancing plant nutrition to achieve the highest yields possible. That was almost a singular goal with primarily a chemistry oriented approach. And there were very good reasons for that. The fundamental reason is simply that chemistry was the one thing that we were able to measure readily. We could measure soil mineral balance profiles in laboratories and plant nutritional balance and so forth and we didn't have the ready capability of measuring biology. So where agronomy is shifting and evolving now is getting into the space of managing biology. And even to a degree, I think with our work with changing internal plant environments, also changing biophysics. So we're measuring and managing biology and biophysics in addition to chemistry. Not discarding chemistry, but just saying. Actually the picture is more complex than just measuring chemistry alone. And the challenge with this approach that when we were measuring and managing chemistry alone, it was very easy, was comparatively easy. And so we now, as we start looking at incorporating DNA sequencing and measuring all these different microbiome parameters of what's happening within a plant and around a plant, we get this Influx of data and all these different factors that we need to take into consideration when making agronomic recommendations. And in a sense, we're moving away from the easy button. It's becoming more complex, it's been more complicated, there's more factors involved. It's just the decision making is much more nuanced now with the additional information. And the challenge to do that well is to be a really good agronomist. Around with that body of knowledge just requires a depth of information and breadth of information and knowledge and training. And the other aspect is, in today's world of rapidly evolving research, it's virtually impossible to keep up. I enjoy reading. I read and study a lot. I usually read the equivalent of roughly a book a day. It's impossible for me to keep up with all the new innovations that are coming. So you look at that and you say, okay, we need some way of democratizing access to information and making this more readily accessible because it is more complicated. So how do we solve for that? Because if we don't solve for that, this approach to agronomy is never going to scale. It's impossible to scale because you can't train enough people fast enough to really have the impact that I believe is necessary. So roughly 2012 or 13, somewhere in that time period, we started using plant SAP analysis in 2011. And that already was pointing me to the idea that we can have this level of data and information, that we can predict disease and insect susceptibility into the future. But I realized that our limitation to scaling effectively was the people that we could train and hire. And about that time I read Ray Kurzweil's work on the approach of singularity. I read some of Peter Diamandis work, particularly his book on abundance, as well as several of his other books. And I was persuaded of the accuracy of their predictions, that we would be approaching serious capability AI in a matter of a decade. So in the winter of 2013 and 2014, I sat down and sketched out my vision for what at that time I was calling an AI agronomist. What were the capabilities that an AI agronomist needed to have? What was the type of data and information that it needed in order to be accurate and to be correct? And in our work at AEA over the last decade, the technology wasn't there yet. Ten years ago, the technology wasn't there from a compute power perspective. It wasn't there from a sensor capability perspective. But we're seeing this rapid iteration where at that point I had envisioned that we would have in field sensors capable of measuring what was happening inside plants with a high degree of accuracy and being able to measure what's happening in soil with a high degree of accuracy. And we are there, that technology is here or is imminent. It's rapidly being developed. And you combine that data now with rapidly developing AI, and all of a sudden we have access to information and decision making capacities and capabilities that far exceed what a human can do on its own.

And how did you get from that vision to actually, you know, fieldmark being a real thing? What, what does that process look like? Was it years of making sure that you were capturing and storing the data in the right way, or was it more of just like, okay, now that we have all these other LLMs out there, we have the tools available to, to put this together.

Well, you still have the idiom of garbage in, garbage out. And I realized that with this framework and this approach that we were taking. So I would define our approach to agronomy. This is kind of a bit of an overly simplistic explanation, but essentially the kernel of it is that disease and insect resistance is a function of nutritional integrity and microbiome integrity. It's those two things. And of course, when you have plant health that is at a level of disease and insect resistance, you also get elevated yields, because that's just. It comes with the territory. You don't really have one without having the other in conjunction with it. But in order to accomplish those outcomes, we were measuring different data points. And I realized that if we want to get to a point a decade in the future, this is thinking in 2013, 2014, where we have high fidelity data to do machine learning or deep learning and to develop AIs on. It's not just there's two questions. The one is what is the quality of your data and the thoroughness of it? Do you have a large enough database that has enough strength and robustness? But also what exactly are you measuring? What are the important parameters to measure? Mainstream agronomy doesn't measure selenium content and molybdenum content or cobalt content. And yet those elements, along with a handful of others, are fundamental for effective functioning of a plant's immune system. So it's not just the quantity of data or even the quality of data. It is what defines the data. What are the parameters that you're measuring. So the benefit that I had of thinking about this a decade ago was just clearly defining what exactly? With our current understanding, what exactly is it that we need to measure? So that at that future point we will have the data to be able to make good, reasoned, reliable recommendations. And obviously knowledge and information is constantly evolving. What we were capable of measuring 10 years ago is very different from what we'll be capable of measuring in 10 years from now. But I think that was kind of foundational to the success of Fieldlark is we were able to capture not just high volumes of high quality data, but measuring the right thing.

And as you're approaching rolling this out and making this public, which you did this year, what were some of the lessons or pivots you needed to make along the way? Meaning, you know, when you came out the early version of Fieldmark, like, okay, we're not there yet because we need to solve whatever the case may be.

You know, Tim, that's really good question. But we, we iterated. The tech space is so different from the agricultural space in that with a crop you have an entire season that you're committed to a crop and certainly the situation is evolving and you're adapting constantly. But the tech space, I'm struggling to answer your question because we made a thousand micro pivots. It's just the iteration and the evolution. There is such rapid iteration from week to week to week, every week. And this is particularly true of the technology space in general and particularly true of the AI space. So we are constantly evaluating the different engines that we're using. So the way Fieldlark is built is you have, for simplicity's sake, I'll call it two macro information pools. You have the one information pool which is just the general Internet and everything that is out there, which has varying degrees of quality. And then you have this private pool of information that we have put together from all the data that we've collected and all the sources of information that we've used and so forth. So that is curated high fidelity information. Then you have an engine that can access both of those pools of information. And we're able to keep our tool proprietary at least until AI evolves to the point of being able to hack everything. And the example that I wanted to point out is that the engines themselves and their strength and capability is iterating so rapidly. One week GROK exposed to those two databases is far superior to everything else. Two weeks later, another engine outperforms grok. Three weeks later, it's just this revolving door of top competitors because all of those engines are iterating so rapidly. And in the same way, we are also needing to, to the degree that's possible now, use those increasingly robust Engines to improve the quality of our private information pool, to improve the synthesis and to improve the. What's the word that I'm looking for? You can synthesize just raw data and come to different conclusions about that data. And as we're able to analyze it ever more deeply and ever more deeply, we have new insights that are emerging week by week and month by month. And then you incorporate those new insights, once they're validated and verified, you incorporate those back into the knowledge set. So it's just this continuous iteration of learning and improvement.

Yeah. In that process, how do you keep the, the biased information out and make sure it's more objective information? Obviously I understand on your own proprietary side, but on the more publicly available side.

That's a fascinating question, Tim, and it makes me laugh because of an interaction that I had on social media. So we trained Fieldlark and we gave Field ARC a, a fairly broad ranging and very thorough set of instructions on ethical principles. What are the ethical principles that needs to be captured, the precautionary principle and so forth. And then we gave it very deliberate and carefully constructed instructions to remove all bias and to consider externalities. So there's all types of bias that can occur. There can be just the bias of having 60 peer reviewed papers saying one thing and then three or four other papers saying something that is in direct conflict or in direct contrast to those. And so you have a bias of just sheer volume of information. You also have a bias of recency versus historically. So synthesizing all that is you have to think through fairly carefully how to evaluate those. But it was given the direction of, of removing all biases and then considering externalities, considering second and third order consequences of not just what happens today, but what just happened, what happens as a result of making a recommendation several years into the future or with preceding crops or in the macro environment in general. And what made me chuckle is when we first released Fieldlark as a pilot. Actually no, this was post pilot stage. We put it out there publicly and we had a farmer from the UK literally within the first 24 hours. One of the comments that popped up was that oh yeah, this thing is biased. And I found that fascinating because what happened? So first of all, FieldArc is trained primarily similar to our work at AEA. And what our domain of expertise is, it is not really trained on making pesticide recommendations, herbicides, fungicides and so forth. It's not really its core strength. It can do that, but it's not what it really focuses on. It focuses More on the nutritional and microbiome, agronomic management pieces and genetics and so forth. And this user had a certain expectation of fertilizer recommendations that. Because what happens is when you take, when externalities are taken into consideration and when you remove the bias, then all of a sudden there are some fertilizers that just do not come up to the top of the list. Fieldlark doesn't make recommendations for anhydrous ammonia, for example, or potassium chloride, not because it's been instructed not to, but because it's looking at the broad array of evidence and looking at the additional externalities and saying there's much smarter ways and much better ways of applying nitrogen than using anhydrous ammonia. So we haven't given it any instructions, for example, to take a biological agriculture approach or to take it. Give it a region of agriculture approach, because it needs to be bigger than that. The training needs to have a truly macro and global perspective and not a bias perspective. But in the process of removing bias, some users perceived the presence of bias, which is fascinating. A reflection of our own biases that we have when we're using is fascinating.

Yeah, it makes me think, you know, on one hand we talk about removing bias as if it's always a good thing and it's always what everybody wants, but when you have users on the other side, as in this example, it makes you wonder like, okay, to what degree do we need to factor in what bias a user might be coming from, you know, when using this? And in your example is maybe a little bit different where I'm going with this, but it's like, okay, if, if we have an optimal answer and we know that that user has a bias against that answer, do we also need to find something that is the optimal intersection of, of what they should do unbiased and what they're going to do because of their biases? That becomes a whole lot more complicated, I guess.

Well, that's also where just prompting comes in, where the user just needs to come back and say, hey, hang on, I don't want to do xyz. I have a strong preference to do ABC because of these reasons. Take this into consideration when making your recommendations and you'll come back with a different set of recommendations.

Right. So that's the magic of it being AI and not just a database searcher.

Right, Exactly. But there is another aspect to this, which is that these large language models and the technology as it presently exists, and again, it's just iterating extremely rapidly, but kind of inherent within A large language model, you ask these currently existing GPTs or these AIs as they're called, you ask them the exact same question six different times, you're going to get six slightly different answers. There'll be 80 to 90% overlap, but there will be slight variation. And over time we're already seeing an increasing convergence, even just in the last couple of years of development. But there is this kind of inherent design for answer variability. It's not a bug, it's a feature, it's the way that they're designed. Well, the challenge with that is when you are making agronomic recommendations, in some contexts, precision is not optional, variability becomes a problem. And so there is this other aspect of Field ARC that we're building, and we're using it internally, but it hasn't been released publicly yet. Where we're building this algorithmic engine that has learned from the database that we have put together. And this algorithmic engine, you can now put in any type of laboratory data of microbiome analysis or of nutritional profile analysis, and it will make a set of recommendations for you that are not based on the LLM. They are based on algorithms. They are precise, repeatable, consistent, and those two engines talk to each other. Or you now have the algorithmic engine that can communicate to the large language model and back and forth. So you can ask fieldarc to describe the rationale or the logic sequences for making a certain set of recommendations. And it can describe those accurately, but the recommendations will be precise and consistent.

Really interesting. That goes a little bit in line with what I was going to ask you about, which is the feedback loop. How do you complete the feedback loop, you know, for aea so that people who are using fieldlark, you actually find out kind of what's happening in the field and use that to kind of continue to train the model. How do you sort of complete that feedback loop?

There's two layers of that. The first is that each user, like these large language models, these engines now have these very long context windows where they remember all of your historical interactions. And so simply by the process of interacting with it, you can tell it, hey, this is my local context and my situation here are the dynamics of my fertilizer supplier relationships, whatever information you want to give them, and it will remember that you, as the user, train it on your specific context and it remembers that. So that's, that's one level, but the other level of feedback from the user to us as a company is how do we improve the technology is at this point I was about to call it manual. Manual is not quite the right term, but it is, it is very much person to person. Even though we have a fairly large user base, the best innovation always occurs as a result of conversations with the users saying, hey, I want to do this, I'm trying to do this, I need this functionality. And then incorporating that feedback directly, there's no replacement for a real conversation. I mean, there's, yeah, we get feedback on social media and by email and many different directions. But the most thorough and deep feedback comes from personal conversations also generally from people who are very heavy users. Right.

And you have a team of AEA agronomists out there using this app, you know, alongside their, their farmer customers as well.

Yes, we've been using it internally for quite some time. And the beauty. And I spoke a bit about the algorithmic engine that we have as well. So right now our work, particularly on the higher value specialty crops, we rely and use SAP analysis very intensely. And so the beauty of this system is that we can upload SAP analysis data and it makes a set of recommendations and provides a written explanation for those recommendations. And this is so valuable we have several team members who, they might get Lab reports from 60 or 80 different blocks in a single day and that might be a recurring stream five days a week. Well, how do you manage that volume of information data coming in? So it used to be this consumed tremendous amounts of time and energy and mental bandwidth to look at all this data and make a set of recommendations. And now we've shifted where the recommendations are no longer human generated, they are now human reviewed. I think that human component still makes sense because there's still context that people know, say, oh, actually these three blocks are in the same geographic location and I'm only going to load the sprayer one time. And so they all need to have the exact same recommendation. So that human review piece is still essential and I think will be for quite some time. But the technology is iterating very rapidly. So the way that I think about this technology, you know, the prevailing paradigm for a lot of tech development over the last couple of decades has been about how do we most effectively replace people. And I think that has the potential to be problematic in agriculture because there is no replacement for human observation and intuition. And there are skills, there are things that AI can do and that computational capacity can do much better than humans. And there are human skills that are much more effective than anything a computer can do. The most powerful, it's not AI versus human, it's AI and human combined. The way that I've been very deliberate and very intentional in how we develop field ARC and the various other technological tools that we're working on as not how do we replace people, but how do we make people better, how do we facilitate them being better observers and them asking better questions. So one of the ways this is manifested, if you type a request into fieldlark, if it's missing context, in order to make good recommendations, fieldarc will come back and will say, hey, in order to make the best recommendation, we really need to know this additional information as well. And then when it gives you a response, it'll give you a very long detailed response. If that's the way you've said it and what you've asked it for, at the end it will say, here are other possible questions for you to consider, to dig deeper. And it'll give you four or five follow up questions. And one of the most common pieces of feedback and response is people have been reporting the incredible learning journey and discovery journey that they have been going down following those follow up questions and just following the thread.

Hmm. Because they seem, they say, yeah, you know, that is a good question. I do need to know that. Yeah, that's really cool. Talk more about kind of the human side, like through this process, what has been clear to you about the things that humans are just going to be better at and why, you know, why that's such an important part of the equation.

A couple of decades ago, Gary Kasparov wrote the book Deep. Think about the competition that he had with IBM's Deep Blue, if I'm remembering my names correctly. So this was the first competition of human versus computer of the world. Grand chess master playing chess against what was essentially what we would today call machine learning or deep learning. And eventually, of course, the computer won. And Gary wrote this book about his experiences and his analysis. And it's a fascinating book. It's worth looking or asking ChatGPT or Gemini or someone to generate a rapid overview of the book and Cliff Notes version. But in essence, Gary said that there's different types of competition. You have human versus human competition. You have AI versus AI, you can have human versus AI. And he wrote about the strengths of each of them. So the obvious strengths of the computer are just simple raw compute power. But when it comes to strategy, our unconventional strategy, or the kind of these intuitive aspects, humans can beat the machines consistently. And so the most powerful combination is actually human plus AI working together. And I think this is particularly appropriate for agriculture because so many of the decisions, the agronomic decisions that a farmer makes and the decisions that agronomists get challenged on and get asked to work together on. So many of those decisions are decisions about edge cases. Do I make an application today or do I wait a week? And so it's all these edge cases where the combination of intuition and lived experience is extremely important and cannot be duplicated with technology in the near term, can it? In a decade or two, maybe to an approximation. But what technology can do and what it should be doing, in my opinion, is it should be guiding humans and amplifying their strengths. If their strengths are intuition and observation, I'm just picking on two. Then there should be scouting apps that are technology associated, that guide the farmer or the crop scout or the agronomist to ask better questions and to be more observant on specific characteristics, it should ask the question about specific characteristics of what people are observing in the field. And so it's that combination that I think will be the most powerful.

And along similar lines, I know you have experimented with this concept of a digital clone, you know, using your decades of experience to try to digitize that in some way. I'm really curious about this because I have had other people approach me and say, you know, I have a lot of expertise. I'm worried AI is coming for my job. Now. How do I get ahead of it and create my own? You know, they always called it an AI, but it sounds like maybe what they're referring to more is a digital clone. Can you. Can you talk more about that journey you've been on?

That's been an interesting one. Kind of came out of left field a little bit. It came out of first, I was constantly exploring the landscape and trying to understand how things are evolving, how things are going. And at the same time, I also. I get this barrage of emails and social media pings of people wanting to ask questions and asking for feedback. And I enjoy helping people, even people who are not AEA customers. It's so rewarding. I get a couple of these emails every week where people from around the world have implemented the things that they learn on the podcast, and they experience these remarkable successes and they write to tell me about them. And that's very rewarding and very fulfilling. But just from a very simple bandwidth perspective, it's just the opportunity. I don't have the bandwidth to respond to the hundreds of requests that I get every week. So I came across this platform. I don't even remember what took me there, but I came across this platform called Delphi. And Delphi, essentially, the foundational way it works is if you have an existing large content library, either privately or publicly, you can simply point it to all of that information and it will learn from all of it. And essentially you create your own private learning data pool that is distinct from that of the Internet. So in my case, I have this history of the hundreds of webinars that I've taught, the podcast interviews, the blog articles that I've written, and I was amazed at how rapidly I was able to build a database of 5 million words. They consider the kind of the minimum data set you need to be about 2 million words, preferably 4 million. I hit 5 million immediately as soon as I started. So now people are able to ask that digital clone. And if I have ever addressed that topic at any point in the past, it will answer that question as if though it were me. It's a bit freaky, and the space is obviously, it's constantly evolving, but I think that within. It's entirely conceivable that in a relatively short amount of time, many people could have these digital clones if they chose to.

Yeah, it is a little bit creepy when you first hear about it, but I could see the application for sure, because like you said, I'm sure you get a number of questions and you probably get the same question over and over again. And pointing someone to a Frequently Asked Questions page is not quite as personal as saying, like, hey, you know, here is a way you can get John's response at scale. That's pretty cool.

Well, you know, and I also come from this cultural background where personal relationships and community is so important. And I've tried to be very thoughtful and intentional about how we develop technology and how we use technology and how we avoid having technology replace relationships, as has been so common for the last 20 years since the introduction of social media. And so in the way that I've programmed, if you will, the directives that I've given this digital clone, it is very clear the way it introduces itself and the way it describes itself. I am not Jon Kempf. I am John's digital avatar. But I want to create this very clear distinction that you're not talking to John.

Yeah, I think that's really important. And in general, I appreciate your thoughtful approach to this, considering externalities, considering community and connection and honesty and transparency. And one of the questions I had for you on Field Lark is when these LLMs first came out, I was using them all the time. I just thought, this is really interesting just. Just seeing what its capabilities were. And then I started seeing all these reports that like, oh, every time you ask ChatGPT a question, you know, so much water gets used or so much resources, you know, do you ever get that question? Or how do you reconcile that with. With field lark as far as kind of like the environmental impact of tools like these?

Only about every week. I think the answer that we should take as a society, and the answer that we're taking as a company is to the degree that we can, is when we have all these significant energy costs, how much value are we creating that is offsetting that? So there's always this cost benefit analysis that needs to be done. If we're using AI just to simply create better social media posts, then maybe we need to question our priorities and the impact that we're having. But if we are using water for cooling AI data centers, and we're using that to facilitate ecosystem regeneration on scale, where we have much more uniform and consistent rainfall in a given region, for example, then there is a significant cost benefit ratio in favor of doing it. I would propose that we should think about it not in terms of using it or not using it, but in terms of what are we using it for. And for that matter, we would be wise, I think, to ask that question for all the technology that we're utilizing. Well put.

You made a comment early in our conversation about, you know, these are hard times in agriculture for a lot of folks, and sometimes that's, that's kind of where the opportunity comes. From your vantage point at aea, what are you hearing from farmers about? Is this sort of, you know, Overton window open of I just need to rethink how I do things and kind of open to new ideas on how I approach my operation.

That's a good question. And the answer is I have so much empathy for people and even, I mean, I was on this threshold back in the early 2000s. We were on the threshold of not farming anymore because of severe disease and insect pressure. And so. And I've been in that threshold even more recently on farms that I've helped manage and been closely involved with. I think two things can happen, and I'm speaking very broadly of people as a group here, and of course everyone is an individual, but what is happening as a group? I think many people are becoming much more open. They're much more open to doing things differently, but also they're much more paralyzed by fear and by perceived risk and by being afraid to do anything different. So there's kind of both of those Things happening at the same time. You know, I saw a couple of posts by Zach Smith and a couple of others on social media talking about, here we are, it's early November and everyone is talking about the gas tanks rolling across the fields of Illinois and Iowa and everywhere else where we're now injecting anhydrous ammonia. We have an abundance. We do not need any additional information. We already have enough knowledge. Enough information, enough data, enough science. We all know that that is a bad idea from an agronomic perspective. That is poor stewardship. We Talk about the four Rs, right product at the right place at the right time. Well, that fits none of those. That's the wrong product in the wrong place at the wrong time. And yet we still do it on scale. And it's because it's familiar and it's comfortable. And because it is familiar and comfortable, it is perceived as low risk or low risk reward ratio, even though it actually isn't. And when I interviewed Pam Marrone, formerly from Marone Biologicals, on the podcast five or six years ago, and she made a comment relative to their work on biocontrol products that I found very interesting, she said that if many of the historical control products she was speaking in terms of fungicides were held to the same performance standards as the bio controls, they would never qualify and get onto the market. And there is a question in my mind about to what degree that is true. Well, I'm persuaded that it's true to a large degree of commercial fertilizer applications versus biostimulants. I need to qualify that because I said that with a very broad brushstroke. There's no question that appropriate management of nitrogen and phosphorus and potassium produces significant yield responses when done appropriately. And there's also no question that many of the things that are traditional practice when analyzed through wise agronomic lens appeared to be quite idiotic.

That is a provocative but fascinating statement. I never thought about that way. But I'm also not one to question Dr. Marone either. This kind of brings me back to a rabbit hole. I almost went down, but I used self control and said, no, I'm not going to go there yet. But you had talked about a product of kind of winter nitrogen fixation product. Can you go back to that? Now that we're talking about nitrogen a little bit more, this is a bit of a non sequitur. But I did, I did want to get back there. What exactly is that?

We developed this approach. Let me give some context. First, I learned about 15 years ago, came to the realization, just through connecting a whole bunch of dots, that what happens with soil biology and biogeochemistry in soils in the winter is quite different from what happens in the summer. Soil temperatures are colder, and that means biophysical reactions are slowed down, microbial reactions are slowed down, chemistry interactions are slowed down. And if you look at this from a macro geological perspective, if you consider organic matter levels in cold soils versus warm soils, if you just look at the lignite ore, for example, Lignite ore deposits in Alberta are rich midnight black. If you go to, like black earth humates, if you go to the New Mexico deposits, they're this golden amber honey color. And language is constantly evolving. But historically they were referred to as humic substances or humic acids versus fulvic acid, where you have these shorter chain, smaller molecular compounds further south in warmer conditions, the further north you go in colder conditions, you have longer, more complex compounds. Well, we observed some interesting things happening with soil analysis results. And without going into all the detailed background, what I came to realize is that when you stimulate biology during the winter months, you get much better formation of stable organic matter, what is correctly referred to as mineral associated organic matter, during the winter months, than you do during the summer. Because during the summer it just cycles too fast. If you have crop residues, they get vaporized or metabolized and released as carbon dioxide. Nothing wrong with that. We need that to happen. It's not a question of good or bad, it's just different. And what happens during the winter months when you have this formation of mineral associated organic matter as a result of carbon cycling? There is also the opportunity to capture nitrogen and associate nitrogen with those various forms of carbon in a very efficient way that doesn't leach. It's not leachable when you have nitrogen in the form of bacterial proteins or that is associated with these various complex carbon forms. So I came to realize that many of the inoculants that we were applying and things that we're doing, it's very common to put applications on at spring, but you're losing your golden window of opportunity of the winter months when you have colder soil temperatures. So we developed this practice. I'm not even going to call it a product, but this practice of putting on what we called our soil primer applications. And the primer, the first iteration of it was a combination of a product called Rejuvenate, which is a very strong biostimulant, and a microbial inoculant. And the microbial inoculant contained A number of different species, such as Azodobacter and Azospirilum and others that have the capacity to sequester and fix nitrogen. Well, now, today we still have those products in that mix. We still have our soil primer, but we're also evolving the Rejuvenate. Rejuvenate was an incredible product. It's been one of our number one sales products for many years, just simply because of the effectiveness and the results that it has. Inflation has affected the cost of ingredients to the point where it's one of the more expensive products that we sell. And in the current macroeconomic environment, that just doesn't work very well. So we've been working for a while, and this fall we released, or this summer actually, we released an alternative to rejuvenate called Revenant Charge. But the bottom line simply is that when you have the combination of microbes which have the capacity for nitrogen fixation, combined with a strong biostimulant, combined with some degree of crop residue or available soil carbon, combined with cold temperatures, not warm temperatures, in most agricultural soils, you can quite readily fix 60 to 80 units of nitrogen from fall to spring. In some soils, you can do more. In some soils, you can't do that much because they don't have enough carbon. So there is a degree of context dependency here, as there is with all biological products. But I have a very high confidence in that response because it is very consistent and reliable. So I think that's. That's somewhere we're really missing the window of opportunity. In many conventional agronomic managed soils, instead of putting on biological inoculants in the fall, we're putting on anhydrous ammonia in the fall, which has the exact opposite effect.

Yeah. And because in this case it's the mineral associated organic matter, it's, I mean, more likely to not only be there, but be there when the crop needs it. Right throughout the growing season.

Exactly.

Okay.

In essence, with this form of nitrogen, you get time to release during the season. The reality this, it mineralizes slowly. There is nitrification that occurs, but it occurs in most soils at a measured pace that the crop can absorb it effectively. So you get very little to no leaching of this form of nitrogen.

John, I really appreciate getting to share some time with you. I mean, I really do learn a lot every time I've got a chance to talk to you. So thank you very much for doing this. Is there anything else, though, that obviously we could go on for hours on this stuff, and we're barely scratching the surface. But anything else that we want to make sure we properly prioritize today before I let you go.

I can't think of anything else. But Tim, I want to say thank you for the work that you're doing. Thank you for hosting me and having me on. I enjoy these conversations and space is iterating rapidly. I'm also in your position as a podcast host and I know the challenge that there's often, at least in my case there's more guests that I want to interview than I get to. But I enjoy these conversations. I'd enjoy being back on as agriculture evolves and as the situation evolves to report on what new stuff is going on. So thank you for having me and thank you for all the work that you do and and for our listeners who are listening, you can check out our work at Advancing EcoAg and check out Fieldlark.

All right, well that is going to be where we wrap up today's episode. As he just said. Make sure you check out their website, advancingecoag.com check out Fieldlark AI which is just Fieldlark AI. Also, if you're not listening to the Regenerative Agriculture podcast, super interesting content there on John's podcast. So thank you to John for being back on the show again. And this is where I tell you that that whole interview was done with John's digital clone. No, I'm just kidding. It really wasn't that. It was the real John kept, I promise. But interesting stuff about the digital clone as well. I think this was a thought provoking episode. I'm excited to hear your thoughts as always. You can email me tim@gred.com find me in the DMs on LinkedIn or X and look forward to hearing your feedback and thoughts on everything shared on the show, including today's episode. That's it for today. Thank you so much for your time and your attention. I really don't take it lightly. I'll be back next week with another story of ag innovation.