The Exchanges

Q that. And so the registry is getting paid to create the methodology, which Which creates a misalignment of incentives where they're incentivized to create the methodology on my behalf, and probably because I am the customer, somewhat incentivized to make that methodology kind of what I want it to be. What are the other ways in which you think there's a sort of broken incentive structure in the registry world?

A Yes, so sometimes you'll create, you'll have them create a methodology, sometimes you'll get them to approve yours, sometimes they'll use an existing one. But yes, they, they want Basically, the registries historically have the incentive to issue as many credits as possible, because they get paid per project and per issuance, and even sometimes a fee when those credits are traded. So like, at the end of the day, they want to make every credit look as identical as possible to every other credit, and issue as many credits as possible. And because if they can't issue enough credits, then their whole reason for existing in the minds of the voluntary buyers who think they can buy a million credits at 10 dollars a ton, and in the minds of suppliers who are the people who actually pay them to issue the credits, it all falls apart. Right? And once you've issued some credits against a methodology that aren't very good, Then you still have the incentive to not change it, because then it makes you look bad for issuing those credits historically. And you're more incented to dig in on, no, our methodology is fine, actually, look at all these good credits we've been issuing against it.

AI assessment note: “registries historically have the incentive to issue as many credits as possible, because they get paid”

Q the first thing that would be useful is, I'm curious how you think about, there's this like multitude of different approaches that seems to be ever expanding, um, For how you can remove carbon from the atmosphere. Uh, but they presumably can be categorized into a few different buckets. I'm, I'm curious how you think about it at the highest level. What are the big categories of CDR that matter?

A Yeah, there's a couple ways to look at it. I think historically academics and others have tried to categorize like natural versus engineered solutions, nature-based versus non-nature-based, which is like directionally right, but not actually that helpful because like most of the interesting things are somewhere in between. Um, One thing that I found a little more useful is to kind of just think about like, what's the actual trapping mechanism for the CO two? Like where does, where does the carbon actually end up? Um, and there you can sort of think like, okay, well, it's leaving the air and another goes into biomass, whether it's algae or plants or trees or whatever, or like a cow or a person or whatever, uh, or it goes into some kind of mineral carbonate kind of thing, right? Like a limestone or bicarbonate in the ocean or whatever. Uh, or it, Is just directly extracted as a CO two gas and it's kind of not trapped, but it's a pure CO two gas and you got to put it underground or store it or mineralize it somehow. So within that, I would kind of think of it as like organic and inorganic, um, either like it's stored in biology or it's stored in mineral essentially. And everything kind of falls from that.

AI assessment note: “within that, I would kind of think of it as like organic and inorganic”

Q old market as we're calling it right now, and, and a little bit more exactly what's broken in the mechanics there and the value chain, and then we can talk about what the new market needs to look like. So first of all, define the old market as we've been talking about it briefly, and then give me your brief synthesis of like why Why all the incentives are misaligned?

A Yeah, so, for these purposes, the old market, we can sort of think of every Thing after sort of Kyoto clean development mechanism on to recently when carbon removal has started, started to grow, right? So that's primarily by volume avoidance credits. Uh, I mean, I think by volume, purely the most wrecks, but then what people discuss more is the avoided deforestation, red floss, other sorts of, uh, avoidance credits. Uh, there's some avoidance credits Credits that are probably reasonably good, um, but the vast majority of the volume is particularly low quality avoidance. There's also some reforestation, uh, some of that reforestation is likely good, but again, you have this weird pricing pressure where good reforestation costs north of 50 dollars a ton, and if your system is set up to tell people that 12 dollars a ton is good, then no one's gonna buy that if you make the 12 dollars a ton thing look the same.

AI assessment note: “the old market, we can sort of think of every Thing after sort of Kyoto”

Q up with, there's like a bunch of avoided deforestation projects, largely in Africa, that have been, you know, had credits issued from some of the largest project developers in this space. And certified through Vera on those registries, which have turned out to largely be bunk. So like, what actually, what's your sense of what actually happened there? Like, what was the chain of events that led to this situation?

A I would split it into sort of two categories. There is the projects that Generally did comply with the rules of the methodology, and the rules of the methodology just don't represent what's happened in the carbon cycle very well, and people are now figuring that out. And then there's the projects that, like, just did outright fraud, like, literally wiring money to not the people on the land, and instead to their executives, or any other sort of, like, just kind of general fraud that happens to be happening in this place, which also wasn't policed very well. But between those two things, like, there are well-meaning project developers that got caught up in this, right? Because, like, this was the only set of rules they had to issue credits. They don't know. They're told the rules are good. In many cases, they may have been doing a good thing, actually. But because the registries and the feedback loop with the registries is to just keep issuing credits and make each credit look the same as every other credit, everyone's incentivized the other way because the buyers don't want to go back and say, Wow, all this money we've been spending over the last 10 years was a waste. So the registries want to save face and be like, no, no, no, we're improving our methodologies, we're updating it, like, that thing is still real. And it just becomes this, like, house of cards where, like, every …

AI assessment note: “I would split it into sort of two categories. There is the projects that Generally did comply”

Q that because there's no buyer yet. And so there's no buyer to pay at that point. So that has to change, I guess, in the universe that you're describing. So like, what, what is the buyer? Is the buyer saying, Hey registry, I want to buy 10,000 tons. You know, I'll pay you a fee to make sure that they are high quality or like, how does it actually work?

A So there's two things that change. One is that the buyer pays the registry to evaluate deliveries and And two, the credits are issued after they've been delivered, not preemptively before. Like, at the end of the day, like, the buyer should send money to the supplier after they have done the thing, not preemptively based on a guess in the methodology about how much of the thing they might do, which then undermines any feedback looper incentive. So, and then in that world, when the buyer is sending the money to the supplier after they have performed the thing, Like is the case in all of these new carbon removal offtakes, because of course it is. You're buying something that doesn't exist yet. You obviously would only pay them after they've demonstrated they can do the thing. When you have a sort of ex-post paid on delivery credit, then you have a registry who the buyer pays to make sure that delivery is legit.

AI assessment note: “One is that the buyer pays the registry to evaluate deliveries”

Q a third party on top of that who then, and this exists too, who like goes through all the projects and says, okay, these are higher quality and these are lower quality, but as like a first person, Pass. You know, the registry, I would think, should be the one that you think says this is at least real or not real, if not, um, high quality, low quality, right?

A They absolutely should, and this is the role that buyers believed they were actually playing, because it's the obvious role you need them to play, and buyers were misled that they could effectively play that role in this old system, because the, their incentives We're optimized for the supplier's financial interests and the supplier's ability to produce predictable volume, not to actually understand the actual carbon cycle, which is nominally the whole premise of paying anyone to do an offset or removal, is you're paying them to do something to the carbon cycle. Like, that is the thing you are transacting that money for. And the traditional registry system abstracted so far from this to become mostly a paperwork shop with, like, occasional ground contractor visits. And it's just not the case. And then there were a lot of, like, very well-meaning and genuine, like, academic and policy efforts to try to fix the methodologies, right? Be like, okay, so like, the amount of, uh, baseline deforestation that would have happened in this region was actually much higher, uh, so you can't credit the stuff that you didn't have any effect over, or like, you drew your project boundary, uh, Right on the end of someone's land, and instead of deforesting in that area, they just deforested the area next to the box that you're calling your project, and like, sure, no trees were deleted in that squ…

AI assessment note: “They absolutely should, and this is the role that buyers believed they were actually playing”

Q Right, so that's a fundamental distinction, right? It's the opposite, right? Old model, yeah.

A Fundamental distinction. Because of course they work for the buyers. The buyers are the ones who want to make sure that the tons they're buying are legit, and good, and trustworthy, and usable, right? And even in the old world, it's weird, because like, They worked for the suppliers, the suppliers paid them, but, like, paid them with money from the buyers, and it was this weird path-through thing that screwed up the incentives, right? Like, it's not like suppliers just have magic money separate from people buying from them. It's always been the buyers paying, conceptually, but we did this weird path-through to break the value of buyers paying. So we have a registry with a direct relationship with a buyer, whose job is to evaluate deliveries of tons.

AI assessment note: “Fundamental distinction. Because of course they work for the buyers.”

Q in the soil, there's, like, runoff of the soil into the water. So, like, how does that, you're basically stacking these modules on top of each other, and the concept is that by stacking them together, if you account for all of them, then you have a complete and holistic picture of the carbon flux of the thing that you are doing. Like, am I sort of understanding that right?

A Yep, pretty much. And you have various requirements for levels of measurement versus modeling along the way. You have uncertainty and error propagation up through the stack. You have ways to categorize reversals, because it's not going to be perfect to start, but it's what you have to do if you're going to be serious about it. Like, and it's also the only way to track the earth science. Right, because the earth science on some of those components is going to change, and then you can change the module, and then you'll have a feedback loop with the supplier. Because here's the thing, the suppliers don't always know how much or what to measure. That's not their fault. Reasonable people can disagree about this, right? Like, how much more should you spend taking deeper soil samples, or more soil samples, or what shape should you put your control plot in the field, or like, these are not like, One, one answer is obviously doing fraud, and the other answer is obviously correct, right? Like, these are, like, actual, genuine design questions for the system, and what you need is a protocol that puts guardrails around it so that the supplier can make the right decisions, and then gives a feedback loop with every delivery. Because again, like, the question the registry has to answer is, Supplier thinks these tons were delivered. Were they actually?

AI assessment note: “Yep, pretty much. And you have various requirements for levels of measurement”

Q the market has not been able to solve yet, is like, should we be thinking of this as, ah, sort of low durability solutions are just lower value, and thus should command a lower price, but have a teeming market? Or, Low durability solutions don't make sense. They don't actually solve the problem. And we should just be focused entirely on high durability, which is what Stripe is doing, right?

A I think we need an actual way to quantify and at least try to think about the actual, like durability is not the only thing that matters about a solution, because many of the, what we could refer to as low durability, low end credits today, whether it's something like improved forest management or otherwise, it's not that they're low durability, they actually just don't exist. Right? So, like, you have this weird thing where the lower the durability is, in practice, it ends up meaning it's, like, less and less likely to be additional. Right? And that's not a rigorous rule, but I think the intuition is just that, like, the more aggressively you perturb the natural cycle, the more you're sure you did something. So, like, maybe one way to think about it, and that's kind of why we prioritize high durability stuff so much, is because we know at least, in some cases, it's additional, but that's not enough on its own. Right? So, like, one way to think about this, maybe, is any given credit, And this leads into, like, monitoring and verification and all this kind of stuff. And to be clear, there is a lot of value in, in forestry and in soils and in rewilding and all of these approaches. Some of it's just not carbon value. And I think if we could break that out, that would be really, really, really helpful for everyone. Um, but, you know, one way to think about it maybe, just like frame…

AI assessment note: “that's kind of why we prioritize high durability stuff so much, is because we know”

Q I'm curious how you think about, we've, we've talked about durability, we've talked about additionality, uh, some of these other things that are like, you know, determinative of the scalability and the attractiveness of a particular CDR approach, be it impacts on the, you know, all these other externalities. How do you, how are we supposed to weigh those in as we start to think about which approaches make sense?

A Yeah, I think, like, there's this weird doublespeak going on of, like, it's a commodity market, all credits are the same, that's why they make sense to use proportional to my emissions, but also, like, co-benefits are a thing, and, like, sure, maybe we kind of, on background, know that the additionality isn't really here with this thing, but it, like, supports this community that we want to support, or it has this biodiversity benefit, or whatever, and, like, both of those things are valuable, but you can't simultaneously have both. Right? Like, either it's a commodity, which means that, like, every unit of the thing is identical to all other units of the thing, or you want the, whatever you're now referring to as co-benefits, which is totally fine. It's just a different thing and should be evaluated differently.

AI assessment note: “either it's a commodity... or you want... co-benefits... should be evaluated differently.”

Q put that somewhere. Uh, you know, you need to be certain about the permanence of the sequestration, whatever you're going to do with it, but that seems fairly straight, fairly straightforward. And then depending on the other approach, like it's much, can be much more difficult to, to measure and verify. How do you think about where we are today in terms of M and V and where what's missing?

A I think today the situation is that You know, buyers are showing real leadership in giving companies that we don't know if they're going to work, but seem potentially plausible, the benefit of the doubt, and extending early purchasers and giving them an opportunity to deliver, and an opportunity to prove that they've delivered. And that is the correct first step, almost no matter what. So that's incredible to see. It is the case, also, that, like, a project themselves fully developing the mechanism by which they claim they deliver, the incentives are a little messy. It's also the case that, you know, an external party or exchange that sells that project's credits for a fee developing the mechanism is also weird, and clearly not how this should work. Um, but the who can the project work with to try to get it right, and who can the buyer work with to try to get it right, is still a little fuzzy. Um, and I think that, like, I believe that the majority, the significant majority of carbon removal companies and founders are, like, genuinely trying to do this right, and are not trying to scam people, and are not trying to do bullshit stuff, and they genuinely think they're working on a method that's going to work, and they want to do it correctly, and I think what we need to provide is infrastructure and a framework for them to prove that they actually are. Right. And what that could …

AI assessment note: “what we need to provide is infrastructure and a framework for them to prove”

Q we don't get there with mitigation in the meantime, every .1 degrees would cost another 11 trillion dollars to remove from the atmosphere. So like, It just drives home the point that, um, CDR clearly has to have a role. I think that has become, that has become obvious. But how do you think about where to place it? In a, in the broader objective, which is mitigating climate change.

A I think that tweet and that context is just, like, drives home the point that, like, it's really frustrating that we've been, to some degree, quite slow in reducing emissions, right? Like, it is, carbon removal is never intended to be a substitute for emissions reduction, and, like, to that point is going to be dramatically more expensive than the vast majority of reductions, right? Like, that's just how this works. Like, not burning something is going to be cheaper than Like, collecting the dilute remnants of the burned thing and doing something with it. Like, it should be pretty straightforward. I think that it's a different system, right? Like, decarbonizing all industries in the entire economy has a tremendous amount of inertia. And progress, like, massive progress has been made in, you know, cost curves of renewable energy, and you, you know this much better than I do. But I think we, at some point, have to be honest with ourselves about, you know, Needs to be accelerated as much as possible, but there doesn't seem to be a plausible way to reduce quickly enough to avoid, you know, significant overshoot and significant damages, and therefore we need as much CDR as we can, uh, to try to balance that a little bit. And in the long term, like, If we can't, you know, zero out emissions by, you know, mid-century or a little after, there's no way any of this is going to work, righ…

AI assessment note: “carbon removal is never intended to be a substitute for emissions reduction”

Q um, It should bifurcate the market in by pricing, and so you, you should have a market for higher quality stuff and a market for lower quality stuff, and you let the buyers decide, and, and, you know, that sort of answers your question. To what extent do you think that is a solution, or do you think of it more as just like a band-aid on a gaping wound?

A I'm not sure exactly what role ratings are going to play long term, um, but, like, your articulation that that will cause a bifurcation, yeah, that's, like, literally what happened this year, right? Like, this is the first time we've seen the market for the lower quality, cheaper credits contract at the same time as the market for, for newer, uh, more quality, more durable removals has grown. And, you know, the exact attribution of that from, you know, Investigative journalism piece to market liquidity issues to whatever it may be. It's like hard to say for sure, but I think the fact that the scrutiny has, has arrived matters, right? And it means that people are looking not just to, okay, let's fix the methodology slightly, but like, huh, maybe the system was designed wrong. How would you want to design it correctly?

AI assessment note: “I'm not sure exactly what role ratings are going to play long term”