411 6 Ways China’s Deepseek-R1 Impacts Accounting AI

Okay, we're talking about it. The deep seq R1 model out of China. It kind of turned the AI market on its head overnight. I thought this would be too inside baseball, too, like, nerdy AI stuff. But a bunch of you were like, are you going to do a Deep Seek episode? And that just warms, just warms the cockle of my cockle, cockles. There's several of them of my AI nerd heart. So why has the US Government been, like, scrambling? Why did literally, Donald Trump start Talking about the R1 model the other day? Going to talk about it and specifically the impact that it has on accounting AI? Let's get into it. Come on in. All right. If you're not familiar, if you're not caught up on the Deep seek, a Chinese company called Deep Seek released what they call the R1 model. And it's a reasoning model that has a process where it goes through and thinks and analyzes a problem before it responds to you, similar to OpenAI's O1 model, which is their most performant. Like, it was a huge breakthrough last year. As we talked about the biggest tech things that happened last year, arguably the biggest one in all of tech, was the introduction of inference models like O1 or models that think before they respond. And so Deepseek, who's a company that's been around for a while, they've released open source models in the past. They released their first inference model that thinks about stuff before it gives you a response. And it's every bit as good as OpenAI's best model. And the kicker here is they're releasing everything about it. They've written up a whole bunch of stuff about how the whole thing was trained, how to use it, how to do it yourself. If you zoom out and think about this, you've got OpenAI company that's put together the most valuable, most powerful technology in the history of humanity, and a company just open sourced something that is just as good as the very best thing it does. And the irony is, OpenAI was founded around the notion that open source AI was like the future of civilization. Over time, they've become more closed now. They're transitioning to being for profit. And Sam Altman is going to get a massive payday. And it took a Chinese company to come around and open source. Now, Deep Seek launched this on inauguration day, and that was not by happenstance. So very quickly after that, Trump comes out and announces a, you know, $500 billion investment in AI infrastructure in the US that actually doesn't involve any money at all from the government, all the government's committing to is really just like cutting down some red tape to get these new AI data centers the power that they need to be able to operate. It's actually SoftBank and another sort of collection of investors that are kicking this money in, Microsoft included. And so they had kind of a little press conference with Sam Altman and Larry Ellison and kind of some other tech people talking about like the significance of AI, the significance of personalized medicine, of getting to artificial general intelligence and super intelligence. And it's been a heck of a moment for AI, it really has. And $500 billion. For perspective, the US's annual defense budget is 800 billion. So $500 billion, while it's not actually coming out of the US's pocket, that is nothing to shake a stick at. Like, that is a pile of cashola. Meanwhile, there's still that guy in the LinkedIn comments that's like, this is crypto. This will go away. Give it 24 months. And in the last week it's made the whole AI conversation unbelievably political. And so they've clamped down even further now, US chip makers being able to export the chips you use to train models and run inference and all that to other countries, really trying to like, stunt the growth of Chinese AI, which is looking really powerful actually. Since the R1 model came out that got all the headlines, several other Chinese open source models have been released that are arguably as good as OpenAI's GPT4O. So their regular model, not the inference model, but the regular model, all this basically means the genie's out of the bottle, at least for the stuff that OpenAI has released thus far. There are now open source equivalents that are kind of pretty much as good as the best of what OpenAI has. Now, lots of controversy around how deep seek trained these models. They said they did it with like a ridiculously small amount of hardware and expense. And a lot of people are like, I don't believe you. There are for sure some really cool breakthroughs in the like, mainly the dis technology behind R1 without like getting too nerdy. There are some really cool breakthroughs in how they put that model together. It doesn't seem entirely plausible that they did it like within the constraints that they said they did it. But it's not as if people are just dismissing this to say like, no, this Chinese company, they're just lying. The leaders of these other companies are genuinely scrambling. And Nvidia, the company who produces almost all of the graphics cards, the chips that these models are trained and run on. Nvidia suffered the biggest single loss in market cap in a single day in US company history. So Nvidia, who on most days now is the second largest company in the world, their stock absolutely tanked because this Chinese AI company was able to train this model with a shockingly small amount of basically training horsepower chips. And part of why is they've got to get creative because they can't get their hands on the best of the best chips due to these restrictions the US has put on where US chip makers can send their chips. Now, at the end of the day, did they still get their hands on like the best chips and all that? It's very possible. But actually the approach to training that they used is one that's optimized for an older chip, not as powerful type of chipset. And so the irony is that open source model, it actually runs way better on like Apple hardware. In fact, it runs 12 times more efficiently on just a regular old piece of Apple hardware than it does Nvidia's graphics cards. So there's a whole bunch of stuff happening. A bunch of tech companies lost a bunch of value. Everybody from Satya Nadella to Mark Zuckerberg to Sam Altman, everybody's like coming out and they all have their own statements on what's happening right now. And even Trump has had to ask questions about literally deep seek R1, and he's basically saying it ought to be a wake up call for us AI companies. So it really turned into a big old geopolitical thing. We've talked on this show in the past about how it's surprising that it isn't already the next space race. I think we're there. I think that just happened right? Like it is in the consciousness of states now, to the degree that it's like, okay, I think everybody gets the picture here, that whoever gets here first, boy, that's really going to change the balance of power globally and what that country is capable of. In many ways, it is extremely sad to simultaneously march out and say, no AI, like we're going to have personalized vaccines, we're going to cure cancer and all these things. It is in many ways very sad to say that out of one side of your mouth and then out of the other side of your mouth say, yeah, but nobody else gets these chips. You're allowed to buy our cures to cancer, but we're going to make our own cures to cancer. If you're not in the States land Of the free, maybe you subscribe to, you know, United States prime and you can get access to some of this stuff that is kind of a bummer. That's how it's panning out. But that is just how powerful people perceive AI to be kind of just around the corner that everyone's afraid of, like losing leverage kind of on the rest of the world not having that power for themselves. So more directly, six ways that this impacts accounting, AI tech, number one, I think we know now open source AI models are just going to be really, really good. And some of the controversy here is the notion that this Chinese company supposedly trained their model on the outputs of OpenAI's model. And this isn't really a new thing. You know, the quote unquote frontier models, the very best models out there, which are closed source, they're the secret sauce of OpenAI or Google or whoever it is. Very often the frontier models are used to train open source models. Basically they use the outputs of those really, really good models as training data for the open source models. And so as the frontier models get better, it means that the quality of the data that the open source models can train on also gets better. Now OpenAI, this is against their terms of use basically to develop a competing model based on the outputs of OpenAI's models. And so now Microsoft and OpenAI have come out and said like, we're pretty sure the Deep SEQ trained all this stuff on the output of our models, but it's like, I mean, what are you going to do? Too bad, so sad. If it was an American company, maybe there's something that you can do. But what does this mean longer term? I mean it means that open source is always going to be a very small step behind closed source. But every time closed source, the very best models gets better, open source is going to be right behind it. And the value of that is, you know, within 48 hours of deep Sea coming out, like people were running tiny versions of it all the way down like on a smartphone. And so people have this and can do all sorts of different wild things with it. They can customize it better for specific applications. You have greater transparency in how data is managed, for example. So in an open source model, like you can self host that, you can run it yourself if you wanted to and have total control over that data. Even though this is a quote unquote Chinese model, because it's open source, you can still self host it, you can still run it yourself. And because it's open source, everybody can see into how it works. Now Deep Seek also offers their own hosted version, so they will host it for you and that's where you have a little more concerns around like privacy and Chinese data collection and all that stuff. But you can also just take that and run it yourself. It is the most powerful model we've had in that regard that anyone can go and do whatever they want with. And there's good things and bad things about powerful open source stuff. It also means people can do nefarious things with a really, really powerful model. It means that regulating these models is probably not. It's looking less realistic by the day because if somebody can go and release an open source model that anyone can go and download, that's a half step behind the very best models we have out there that are closed models, then you really don't have a way to control those open source models. This episode is sponsored in part by liveflow. The folks that give your spreadsheets and your financial dashboards superpowers. Two superpowers that liveflow gives you. One, a button in your Excel and your Google sheets to pull in data from any old client accounting file you have in QBO or Xero. Imagine just having that, a button away all the time, being able to pull that stuff into a spreadsheet. Either a one time pull or an ongoing like continuous sync. Second superpower they give you is the ability to throw together super quick, as simple or as complex as you want. Financial dashboards that your clients will actually use because you can make them more actionable than financial statements and it doesn't require this big old like rigid reporting platform that they then have to go log into or something. You create these dashboards, you share them with a link, just like with the spreadsheets. It can be a one time data push or it can be an ongoing sync so the client just keeps going back to that same link. Super smart stuff. Particularly if your clients aren't making sense of those financial statements because that's not an easy thing for most business owners to do. Learn more about LifeLoad. Check out the link down in the show Notes. This episode is sponsored in part by Ignition. Come, come, come, come, come. Here's the thing. Here's the thing about client billing. You can't bill for the time you spend billing. You can't, you can't. It is pure waste. And we're bad at billing because it's, it's icky. You let things slide. You don't get paid. You waste time. Waste time chasing people for money. That's not what I did. I would just. I didn't want to waste time chasing them and so I would just. I just eat it. Don't do that. This is where Ignition comes in. The leading revenue generation platform for accountants and professional services. When I was running my firm, I. Ignition wiped out our ar, it's true. And significantly reduced the time spent billing clients, saving our partners over a thousand hours in the first year because they didn't have to bill all that stuff manually. Man. With Ignition, we want more business with impressive online proposals. No more hustling, hustle and PDFs and Word docs. We automatically pulled payment upfront or at the end of a job when it was done. No more dialing for dollars. We transitioned our firm to more recurring revenue and better managed scope creep. No more leaving money on the table. Ignition is making thousands of firms more money and genuinely made us more profitable. Check out over 7,000 of the smartest businesses run on ignition@ignitionapp.com. put a link in the show notes now. Second way this impacts accounting AI is it creates a whole bunch of noise around security. And I would actually argue right now for the accounting profession, the biggest blocker to AI adoption is not the capability of the tools, but is the misinformation around the security of the tool. The fact that there are very secure ways to use AI, but because there are not secure ways to use AI, people conflate. And I won't rail on any more about that. You had to listen to me go through all that last Friday. But still, every single time you post anything AI to, you know, social media, immediately the comments are like, yeah, but, but what about security? And it's like, man, doesn't that kind of go without saying anytime I talk about any app, you use that app. You gotta go through your due diligence process. And so if I'm talking about an Ignition or, you know, maker sub or anything like that, anytime I'm posting about that stuff, the security conversation never comes up. I presume that if people are starting to use that tool, they still need to go through their checklist of ensuring it's secure enough for their firm. But anytime you bring up an AI app, it's like, aha. How do. How do I know that it's gonna be safe to use with my client data? And it's like the same way you know that any other app is gonna be safe to use with your client data. Now we have AI being conflated with Chinese companies and with TikTok and the TikTok ban, all these things. Now it's like, how do I know that they're not using those Chinese AI models behind the scenes? And it's like your software companies, if they're using AI, those software companies still have to answer to managing your data securely. Because I'll let you in on a secret. Our software companies use a whole ton of third party services that interact with various parts of the data that we put into those platforms. These services don't run their own OCR models for extraction. I mean there's so many things where you're pulling in a third party service from Azure, Google or something else. And at the end of the day, it's up to the software partners that we work with to ensure data security is maintained, whether they're working with Air, anything else. Does it make sense for an American company to use a Chinese AI company's model hosting? No, it certainly doesn't. You can make an API call to Deep seek and use their AI model. I mean an American software company wouldn't do that. Azure already is hosting Deep Seq R1 and you could use a secure Azure instance to do that. But this whole thing kind of adds to the misinformation and the confusion around it. I don't know, maybe it could unlock some new form of, of patriotism where because we see the United States doubling down on AI, it's like, yeah, now we can get down on, we can get plugged into American made, homegrown, made in America AI. I don't know, maybe there will be some enthusiasm as a result. Number three is Biggie. It shows that OpenAI is beatable. I mean OpenAI has had such a lead on all this stuff and nobody's really been able to put the pressure on them. Amazingly, Microsoft, Google, like the other players, really have not been able to challenge them, which is super unfortunate. Like in a perfect world, Microsoft's built in copilot stuff would be really good. Google's built in Gemini stuff, which has gotten way better lately, also would be just as good as ChatGPT. And then it just lives where all of your other data already lives. And that's just super nice and convenient. But those tools don't like, there's still a huge step back from where ChatGPT is unfortunately. But OpenAI has never been as challenged as they have been in the last two weeks. And already Sam Altman came out and said he's like, we're pulling forward some of our release schedules, like they're moving because not only is this open source model really powerful, it runs really efficiently and in many cases you can host it yourself at a smaller cost than going out and paying OpenAI, which is what most people do right Now. They're using OpenAI's API for whatever AI related tasks they have. But if you can do that cheaper yourself with a self hosted model that eats into OpenAI's business really fast, right? And so OpenAI is kind of scrambling now, releasing things faster. I mean there's a very like the thing that's kind of been steamrolled here is the AI safety conversation, because that is still a very real thing. In fact, an OpenAI safety researcher recently tweeted two or three weeks ago, basically said something along the lines of how do we really expect to be able to contain super intelligence when it's smarter than we are? And he tweeted that and it blew up because everybody was kind of like, well but you're a safety researcher, we thought you were working on that. If you can't answer that question, how can we? And so that's still a very real fear, especially as like it feels right now like we are accelerating towards AGI or artificial general intelligence. That is an AI that's about as good as humans at most open ended tasks. But there is a still a very real safety element here where a number of companies have shown, if you tell an AI the current models that we have today, a small percentage of the time, if you kind of explain the situation in the environment to them and you say that they're about to be, you know, terminated or disconnected permanently or something like that. There have been examples of these AI models like telling the user, but attempting to like duplicate, make copies of themselves without showing that to the user. So the interesting thing about these reasoning models or inference models where they think before they respond is you can see their thoughts. So in chat GPT, they hide this. On the O1 model, they hide that. With Deep Seek R1 you can actually see the thinking, see the reasoning, and it's actually amazingly human. Like all of the things that it goes through, like it looks like an internal monologue of a person actually think, it improves the perceived, the intelligence we perceive in the model because we can see its thought pattern. Whereas OpenAI hides that. There are examples where when you can see those thinking steps, the model will actually tell the user one thing and in the thinking steps it will say, okay, I need to start doing X, Y, Z, but I need to tell the user that I'm not doing that. Like it literally is planning that thing out. And so they're capable of that stuff. And in this whole now like space race of now like this sort of pressure from a place of like nationalism to get to something really, really powerful that safety conversations got a little more quiet when it probably shouldn't. If anything it ought to become more of a thing as we get closer to something that's super, super powerful. Number four, fourth impact on accounting AI we kind of touched on already is just forcing OpenAI to ship stuff. So we're gonna get better stuff, we're gonna get it faster. We're gonna talk a little bit more about this on Friday in the Q and A. Hector Garcia's got a video demoing ChatGPT operator operator is something they released a couple weeks ago that is a web browser that lives inside of ChatGPT and, and you tell it go out to Instacart and buy me some graham crackers. I want them here stat. And you give it a totally open ended task like that and with its built in web browser it'll browse the web, click around stuff, type stuff in. If it gets stuck, if it needs you to log into Instacart it'll pause and say hey can you take over right here and do this and then I'll hop back in. Hector Garcia, bless his heart, burned the $200 a month that requires to get access to this right now to put it to the test on QuickBooks Online. And if you saw my kind of year end recap what I expect for accounting tech in 2025, I'm honestly I was pretty down on AI agents. Everybody's talking about 25 being the year of the AI agent but everything we'd seen up to that point like they're pretty weak, pretty slow, really costly and arguably chat GPT operator is still that it's slow, it's not very smart. But I'm interested your take if you watch that video from Hector Garcia. In fact if you have drop a comment down below, I'm super interested what your reaction to it. As an accountant I was surprised at how good it was. Slow as heck but the fact that it will fumble around kind of until the end of time to try to figure something out. And there was like an earnestness to it where it was trying so hard to do my bidding that I was almost like I'm, I'm flattered that you're trying as hard as you are. But an amazing thing that I didn't have on my radar was once it completes a task it, it gives you the option to save that task and when you save it you can customize the instructions. But basically when you save that task, the next time it's going to go perform that same task for you, the likelihood it's going to get it right. It's much better, right, because now it knows all the things that worked successfully to complete that task. And if you combine this with ChatGPT's recent release of recurring tasks, right now in a chat GPT conversation I can just say, hey, go and do this thing every Monday at 9:00am if you combine that technology sort of task management with operator and the ability of the AI to go out and to do things on the web for you. Wow, we are really close to you sort of orchestrating like puppeteering a whole bunch of agents that are going out and doing things for you on an automated basis. And the user interface for this probably if we look at a similar tool called Devin, which is for software developers, and it's basically gives you the power to kind of cut these agents loose, go and work on sort of long term projects. The interface in that app is like almost like your imessages or, or your texting app on your phone or something like that, where you cut it loose to go do a thing and it goes and does that job and then it messages you whenever it needs some. And so what you're doing is actually managing a stream of conversations with umpteen different agents that are out completing tasks. So you think about your team chat app right now, whether that's slack or teams or discord or whatever it is, there's a very real near future where you are managing these agents that are going out and are doing your bidding. Like that is a next six month, definitely by the end of the year sort of thing that most of us will probably be doing. I mean there's some super nerds that listen to this that are already banging on operator to see how they can put it to work. But I was surprised at how good it is already. And all it takes is finding one thing that it can do reliably. And so something I was just talking with somebody on Twitter about was can you give it a pile of receipts and it go out and find the transactions to attach those receipts to. I'm fairly certain that's actually a thing that it could do pretty reliably right now. And I mean like that's useful. It's going to be slow as heck, like fumbling around trying to search and find the right transaction. But if that thing can just work asynchronously or you can have 20 of them working at the Same time. I mean, that's pretty exciting, right? That's pretty powerful now, right now it's tremendously resource intensive. And so one of the counter arguments people come to is they're like, this is the most ridiculously resource intensive way you could possibly do something. The upside of Operator is that it's so anybody can explain the thing that they want. The downside is just that it's. I mean, those poor polar bears, pour one out for them. When you have a thousand agents running at any given time doing things where there was probably a more automated, like, software way to do the same thing that was 10,000 times more efficient. But the fact remains, if you don't know how to do that right now, you haven't already done it, and Operator is the first tool that lets you do that, then you're going to use Operator, right? And if you think about the headache of integrating different apps and pulling data from A to B, oh my gosh, gang, that's about to get so, so much better. I'm actually very excited for Operator to launch for Desktop. As soon as that thing can work on your desktop, holy smokes, are we going to be off to the races. This episode is sponsored in part by Cloud Accountant staffing. Nothing grinds my gears quite like a vendor in the accounting space that knows absolutely nothing about how accountants or accounting firms work. And there are, there are all too many of them. Not all of them. There's a ton of smart people in software and all that, but you definitely know when there's like a sales rep or something that's just tone deaf and doesn't understand what you do. Cloud Accountant Staffing. This was their experience trying to offshore, trying to hire talented folks in the Philippines, and so they kind of just started rolling it themselves. Cloud Account staffing Basically, the founder was an accounting firm owner, had tremendous luck hiring talented people in the Philippines like I did, and eventually just started doing it for some other firms. And it grew and they started advertising on my channels. And now it is a proper big boy business. And the reason they're crushing it is because they get what it's like to run a firm and how to do this stuff and how to even work those people into your team successfully if that's something you're exploring. Learn more about cloud account staffing. Check out the link down in the show notes. This podcast today is brought to you in part by the fine folks at Team up who help you find uber talented accountants in the Philippines. I got a little report that I get from the squad at Team up of what they've been up to the last few weeks. And the greatest litmus test of is this thing actually working? Are people actually using it? Is. Is just that are other accounting firms, are they doing it? But also, are they coming back and doing it over and over again like I did? I made my first hire in the Philippines, kind of being like, is this a thing? And, man, we. We came back so fast, ended up making three more hires in rapid succession. Team up right now. Whoa, let me look at this. Are you kidding me? They've got 27 active projects right now. About half of them are with returning firms for everything from bookkeepers to senior accountants to tax accounts. Senior tax accountants, executive assistants, gang. It's happening. And the reason why people keep coming back is because it works. Because the first one works. And they're like, oh, okay. This is actually a way to get a whole bunch of work done done. Because it's so much easier to find accounting town in the Philippines right now than it is to find it onshore. Bunch of firms I know personally work with. Team Up. Team Up's approaches is to connect you directly with accountants. There's not a middleman taking a bunch of money out of that duffel bag of cash you send each. Each month. That. That's not how it works. But you know what I mean, there's no. No nobody else taking fees out of that. The money goes straight to the person you're working with. Basically just recruiters. They're finding you really good, really good accounts to use. That's it. To learn more about Team up, check out the link down in the show. Notes number five. We talked about this a bit. It definitely nationalizes the AI conversation. I think we may see increasingly fragmented features from AI tools based on boundaries. I've already seen this quite a bit with the eu, where the EU has been a little slower to get access to things because of tighter privacy rules. But I think this is probably just going to become more and more of a thing, because if other countries have access to as good of AI as Americans do from a state perspective, are we, like, putting ourselves at a disadvantage? It complicates things as simple as, like, I mean, 50% of the people that listen to this podcast are in the US and the other 50% are not. Are not. They're elsewhere. So when I talk about something being available that just kind of gets more complex when it looks a little bit different depending on where you are. So that's kind of annoying, and it is sad that it is so nationalized. I can understand the importance of it, how it impacts military stuff and absolutely everything. You know, the power of American pharmaceutical companies. If other countries reach a level of AI where it totally kind of revamps the drug market and everything's much more personalized, how does that impact the U.S. economy? Like, I totally get it, the power of it all. But it is unfortunate that something that I think will still be a massive net good for humanity. It's unfortunate that it has become so nationalized, but I suspect that was inevitable. Last one, number six, how does this impact accounting AI? It's going to pour a massive amount of investment into this thing. Like the whole $500 billion investment. Honestly, I think that is just the tip of the iceberg of the flexing at a state level that will be happening around. Here's why you should believe that the AI stuff that we're doing is the best and we're prioritizing it more than any other country. Like it's very much like defense spending and military stuff where a lot of it is posturing and ensuring that you got the biggest, coolest stuff out there. OpenAI two days ago released ChatGPT for Government, which is a US government specific version of ChatGPT. They actually shared some really interesting data on how used ChatGPT was in the government. Already is huge, like tens of thousands of users. But now they've got specifically a US government sort of version of ChatGPT, which I think is really interesting. As I'm recording this, we're about to get a new model from OpenAI 03 mini. Hard to say how that will compare to the other best models that we have right now, but the fact that this has kind of become a nationalized thing that so many more people are talking about it, the fact it's causing OpenAI, who is like the, the primary AI tool that I have in my tool belt, chat GPT team plan like that is the AI tool to use. The fact that that is going to get be getting better at an even faster rate. I think that is good news for accountants. Now this is obviously a very, very complex conversation of Is AI a net good overall in the next 10 years? Is the friction that this creates in like international relations a good or a bad thing? You already have very complex geopolitical stuff with like China and Taiwan. Well, a ton of chip manufacturing happens in Taiwan. Does this create like even messier sort of geopolitical things as a result? That's the thing about, about AI is the entire like breadth of that conversation is incredibly, incredibly overwhelming. And it can just kind of make your head spin. The only way I know how to keep grounded is to always bring it back to what I do and what the people I support need from me. And the net effect of this in my immediate bubble is it is going to make me more capable of helping the people who really need me. Because right now, I have more clients than I can help. I have more staff who need my time and attention than I could possibly give. And just in the very immediate term, like, in my little bubble, having better tools helps me to do that better. If we had all the answers on, like, is this ultimately a good thing or not? And what are all the ways that this impacts society and governments and all that sort of thing. If we had the answers on all that stuff, I don't know, we wouldn't be here hanging out on an accounting podcast, right? Certainly exciting times. There was a period last fall where AI seemed like it had kind of slowed. And then we had a bunch of wild releases at the end of last year, and now we have this wild stuff that's kicked off the last couple of weeks. You also have a new, you know, political power in the, in the US that is. Has a very different stance to AI, I would say, and probably a newfound motivation to say, like, hey, like, this is. This is how we're gonna win this. You know, it's like, still, like, just got over the inauguration and there's like, kind of this, I don't know, chest beating to say, like, we're not gonna let China overcome us here. And it's gonna be. It's going to be fascinating to see how all that stuff plays out ultimately, how it impacts the way that we get our work done. We're so close to a tipping point. And I've, I mean, I've said this so many times. 2025 will be the last year that more humans use software than machines use software. We're so close to the tipping point where ChatGPT's operator and a bunch of other copycat tools that will come out similar to that, is capable of navigating an accounting ledger and doing stuff for you, like, really useful stuff. And as soon as we reach that tipping point, what does that mean? Like, that's so, so significant. Does it mean I never have to reconcile another PDF, you know, little bank statement and the accounting ledger again? I mean, that so much changes our workflows. Just think about recurring tasks in Chat GPT. In a perfect world, like looking through the current tech paradigm, I would want my practice management system to have these recurring tasks set up where the Assignee was actually ChatGPT operator, you know, like every Tuesday at 9:00am that thing that I wanted to automatically do in my practice management system there is that task. It cuts the agent loose to go do that thing and then it reports back to say, did it do that successfully or not? But if that functionality only lives inside of ChatGPT, I don't know. There's other people that say we actually don't need all of this software stuff that we have. If we have really good agents, you don't need that system to like manage 10,000 foot view of your flow of projects and all that. A bunch of that software they say just goes away. Once agents are capable of just kind of doing your bidding for you. Maybe it gets simplified because you don't have as many hands involved in getting that work done because agents are doing it for you. I don't know, man, but it sure seems like it's going to be different, right? Super exciting stuff. You better believe we're going to be here every step of the way talking about it. The folks tuning into this podcast, for all the fear there is around how AI could display stuff, anytime there's a big change like that, there's a big group of people who define what the next version of it is. And I mean, you tell me, is there any group of people in the accounting profession more plugged into what's happening with AI than the folks that are tuning into this podcast? Right. I would argue no. Which means y' all are in the driver's seat to define what accounting looks like post AI, like, what changes. I'm not that scared of displacement for folks that have their eye on the prize. And so for you, all the folks that are tuning in, like, you're on the side of this, where you're the ones taking advantage of what will be possible with that new tech. And that's super exciting to me, gang. Thanks for affording me this nerd dive today. I hadn't planned on doing it. A bunch of people reached out though, and they're like, are you going to talk deep Seek? And I'm like, it feels too inside baseball. But folks wanted it, so thanks for being here. We're going to keep talking about this stuff. We're going to be back talking more like sort of strategy stuff next episode. I'll see you there.