Erica: Welcome to the Executive Function Brain Trainer podcast. I'm, Dr. Erica Warren.
Darius: And I'm Darius Namdaran and we're your hosts.
Erica: Hey, Darius, it's great to see you today. I'm excited because you're going to take us back into AI world and tell us a little bit about what we're going to do.
Darius: Well, AI in executive functions has kind of gone to the next level again. it happened in January. There was a quiet shift that happened and it started with developers and is now sort of trickling out to, more switched-on techie people. And it will gradually trickle out to regular folk probably over the next year. Well, definitely over the next year they'll become more aware of this. So this is kind of breaking, you know, depends what type of person you are. Basically what's happened is we're used to chatting with AI, okay, in the chat interface. And so AI will happily advise you or create some piece of content or whatever, but it doesn't do stuff for you. Okay. And part of Executive function is getting stuff done. It's kind of like, are you going to stay focused on what are your priorities, or are you going to just get distracted and do other things like go on your phone and do what Facebook wants you to do, or Instagram wants you to do, or are you going to do what you want to do? And fundamentally, I think of executive functions as the ability to orchestrate your own life as an executive, to get what you want done or to be who you want. You know, so do be, feel. All this is directing your energies in the direction you choose as the executive of your life.
Erica: So interesting. You're almost saying if you can do what you want, even if you can't do it, if you want to do it, you can do it because you can offload it onto AI.
Darius: There's a lot more you can now offload onto AI. I'll give you an example. Claude came out with Cowork. Okay. So we know we've got Claude Chat. So if you go into Claude, it's like ChatGPT, it's their version, it's Anthropic's version called Claude. Okay. And you can go in there and chat normally, we all know about that. But they've introduced this thing called Cowork. Okay. And it starts to work with you as a coworker.
Erica: Okay, interesting.
Darius: And it's not copilot like Microsoft has been promising for a while, which is, I don't know it very well, but what happens is you download an app onto Your computer, the Claude app, Windows or Apple, they normally release the Apple one. First, you download it and it, then with your permission, can go and look at your files and organize them and do stuff with your files.
Erica: That sounds scary, that makes me frightened. But just, I'm just go, I understand, go forward.
Darius: But then that fear turns into some excitement and wonder the moment you take, let's say your downloads folder of images, okay. And you say, oh, gosh, I've downloaded all these images, and they've got all these random number names or whatever, and you just say, look, go through these photos, okay? And please suggest how you would organize them. And it would say, right, I can see that there are 300 photos in here. I can see there's a bunch of them that are the logos of your company. There's a bunch of them that seem to be, you know, images for a website. And you've got some manuals in there, images, screenshots of different things. And some of those screenshots are of this product, some of the screenshots are of Canva and another ones of Microsoft and other ones of that. So how about we organize them into five folders, you know, Microsoft, Canva, your logos and so on. Can I do that? And you go, yeah, sure, go. And it will create those file folders, it will organize those into them, put them in the right place, and it will say, can, can you see what I've done? Are you happy with what I've done? And you go, yeah, I like that. That's great. Approved, done. Okay. Right. So it doesn't do it.
Erica: Now, if I said I don't approve, then it wouldn't, it would put it back to what the wet it was.
Darius: Yeah. So what it does is it creates like a, a shadow of, it's called a sandbox. It creates a shadow activity on what you're doing. So it will change the folder structure the way it thinks, and you'll say, that's what I want on mine. And it'll go, okay, I'll do that on yours. Do I have
00:05:00
Darius: permission to move it? Yes, done on yours. So you can see it before it. It does it.
Erica: Okay, so I want to. I'm going to play devil's advocate with you a little bit. Is this safe?
Darius: Well, that's a really good question. With, With Claude Cowork. It is safe, yeah.
Erica: How do I know it's safe?
Darius: Well, because what they do is it creates this. It's called a sandbox. Okay. And this is a technical term where it Creates a mirror of what you've got, a very temporary mirror. And it operates in that mirror world, parallel world, where it cannot change anything on your hard drive unless you give it permission explicitly to do that. so you have to continually give it permissions.
Erica: I'm more concerned about safety, as in my privacy. How do I know that it's not looking at everything on my computer? I mean, bottom line, on my computer, it's got my life from the last 25 years. Whenever I get a new computer, I just copy everything onto the next one and onto the next,
Darius: Ah, yeah, great question, great question. I get where you're coming from. No, so what it does is it doesn't get access to the whole of your hard drive straight away. Okay, so what happens is, let you know, when you create a new chat in ChatGPT or Anthropic Claude, you can click the plus, and you upload a, file or folder or something like that. Yeah. When you, you hit the plus and then you give it a particular folder to pay attention to. So you might say that this is my downloads folder or this is my folder on this project. Let's organize it. Okay. And so it just gets to see that folder. So you could choose the whole of your hard drive if you want, or you could choose the that particular folder, and it will go through that particular folder. So you limit it to that area.
Erica: Tell me, so this is something you download onto your computer and you're saying that's safe to download this onto my computer?
Darius: I would say so. I'm not guaranteeing it, but Anthropic think it's safe and they are a big player. They don't want to do anything dangerous. There are other versions of this that are not safe. Much more advanced versions, like. Well, they're not safe. There's OpenClaw for example, which is. I've got my version of OpenClaw and I call it Roy. Okay. And Roy's got much more access. So much access that I actually bought a separate computer for Roy to run on. So Roy's got his own Mac mini, he's got his own computer, he's got his own monitor, he's got his own apps, he's got everything. And he's only got a few things that are of mine that, are synced to my file. So if he updates it on his, I say, yeah, I'd like that over on mine and we can transfer it over. So I use Roy to manage my notes. Okay, so what happens is that you know how you go and you create something on ChatGPT or on X or whatever, wherever you are, they're all over the place. Or in Gemini or NoteBookLM or whatever. All these notes are just everywhere again, you know, and it's valuable stuff. So I'm done with all of that. So I'm like, whenever I make a note, turn it into a markdown file and drop it into my digital brain in Obsidian, and I have. What I basically building is a digital memory of all of these interactions and assets into one place that is not dependent on OpenAI or anthropic. If I change a model tomorrow, I. I've still got all of my memories, all of my notes, all of my chats in my place, not their place.
Erica: And what's the purpose?
Darius: Well, the purpose is, I mean, it
Erica: makes sense because you're in the middle of a huge, really powerful project. But for like, just like a student, like, how would a student. Or just like a regular person that just has a regular job?
Darius: Okay, so regular.
Erica: How would they. How would, like, your average person benefit, from using something like this? How would you advise them?
Darius: Renaming files is a big deal. So I've just reorganized those files. I'll just take that. And I go, oh, yeah, that folder structure is great. That makes sense for me. And then you decide, oh, gosh, no, I don't like that folder structure. Actually, now that I see it, I would like it more like business and then work and then put those icons, logos and so on into. And my business section and then my work section. All right; I'll reorganize those. That would take time for me. All right, create a new folder, drag that folder over here, do this. That's all technical executive function tasks.
Erica: Yeah, yeah. And I mean, I also like the idea of renaming things because, again, I mean, when particularly. Well, like my Google Files. But I guess it can't help me with Google Files.
Darius: It can go into your Google hard drive with your permission and start reordering.
Erica: Interesting. That's right, because there is something where you can download your Google Files. Interesting.
Darius: And no, you don't download your Google Files. You give Claude Cowork permission
00:10:00
Darius: to go to your Google Drive and start organizing your Google Drive.
Erica: Wow.
Darius: And it will actually open up a Google browser and start clicking. It will say, right, I'm going to go and click on your recents. Okay, I see in recents you've got this. Shall we move this to there? And it'll click on that and I'll put it there and I'll do this. You Know the whole idea with ADHD of body doubling, right? Where sometimes, okay, so body doubling with ADHD in its simplest form is where when you've got a hard administrative or focused task that you've got to do that. You know, I don't do this very well. I kind of put this off or I get distracted or whatever. You know that feeling of when there's someone beside you doing it with you or just even in the room, you're just a bit more focused, you've made a commitment. Yeah, let's do this together. They do it together. That's body doubling. So there's different levels of body doubling. There's the body doubling of let's go to university as students, the library, and just sit together and work together. That's body doubling. You're not doing the work; you're not talking about it. And then you walk away and then you go and have a break, and you come back and you just feel more in the zone. that's kind of body doubling. There's another layer of body doubling where maybe this is more how to deal with processing. So I met a lady with dyslexia who was running a charity, and she had to do applications for the grant. And I said, how long does it normally take you? She said that three days to do, a grant application. I said that fair enough, they're horrible, aren't they? But, you know, worth 20 grand, 50 grand, whatever it is. And she said, yes. But when I sat with my coworker, we did it in two hours. And I said, why did it take two hours compared to three days? And she said, well, what I would do is I would talk and she would go, that's over here. And she would click on that, and she would click on this. And then I would say, oh, we need this over here, we need that or remember this and remember that. And she was orchestrating it verbally rather than clicking on this, clicking on that, and overwhelming or working memory so much that she would, oh, yeah, what was I doing? Right, I'm over here. and she loses her focus, her inhibitory control, she's lost her focus. She comes back to it and then that's an hour gone, another hour gone, another hour gone, Three, three days gone, and it's finished. So laborious. So basically that body doubling person was acting as her working memory to just click through things. Oh yeah, I know where that is. I know where that is. She could stay focused on the actual executive level of deciding what goes into this grant.
Erica: Right. It's almost like, she had a dozen workers and said, you do this, you do this, you do this, you do this, you do this, and let's come all back together. And it is.
Darius: But it's not even as sophisticated as that. Okay. It's nowhere near as sophisticated as that.
Erica: Okay.
Darius: Having developed software for people with dyslexia, I know how much working memory, clicking here, clicking there, and then clicking here takes up, because you've got to think, oh, where is that? Oh, it's over here. This is the dropdown. I'll read the dropdown. Oh, it's there. Now, what am I typing in? I'm typing that. And you've already used up three units of working memory with just one thought. And maybe other people have got five or seven units of working memory. They've got another couple of units to use up to keep on that process. But I've used it all up just clicking on things. I've completely lost track of what I was actually wanting to put into that software. And we just get used to this, thinking this is normal, but it's not normal. That's. Working memory with software is a nightmare.
Erica: Yes. Yeah. No, no. I mean, I have that problem myself. My working mem gets taxed when I'm on my computer because there's just all sorts of things coming at me, and I'm on my way to do something, and something distracts me. I mean, it's. It's. They're just. It's intense.
Darius: So, okay, imagine you had a coworker right beside you, and you said, look, fill in this form for me. And you're like, right, okay, what. Where's the website? Yeah, it's on this government. Whatever form it is, it opens up the webpage. There it is, this URL. okay, it's asking for your name, your first name. All right, put in Darius. Right, great. In Namdaran Put in this, put in that, and then so you can have a conversation. It'll click through for you and, do stuff for you on there that would use up your working memory.
Erica: Now, what I am noticing from what you're saying is that it enables people to use different processing areas that they typically haven't been able to use in the past. So, as you were saying, you know, we have to click through and we have to hold in our mind the map of where we are, and then all these overlapping programs that have different maps and, and. And that can be very taxing. Now, some people, they're not going to want the verbal piece because they're More tactile. But really, up until now, this interface that we have been using
00:15:00
Erica: is more tactile, and it doesn't really accommodate those people that are very verbal.
Darius: Yeah. And the interface. Yeah.
Erica: So interface tends more simultaneous, as you're always seeing the big picture, but the big picture gets a little bit messy. So it's helping you to organize things sequentially or whatever. When things get a little bit disorganized, it can organize it for you. So it's constantly keeping things organized. Organized.
Darius: Imagine you had someone that went into the messy room of your bedroom and sorted out your bed and sorted everything and put it in its place. That's what it's like for your messy computer. Okay, so let's go back to our analogy. Super simple analogy. Okay. For example. It's not just an adjacent real. You go, right, yeah. All those logo images. You know how you sometimes get these logo images and you've got a little one and a medium one and a big one, and you got a PNG and a jpeg, and you're like, oh, my God.
Erica: I don't.
Darius: I get so confused with all this. And you're like, look, I'm so confused with all of this. How do we rename this in a way that would work for me? And then the AI will say to you, well, there's a few ways that people typically do this. They might start with logo large and then put in the date. Or they might say logo large for website, logo large for printing. Or, you know, so I don't need to think about, PNG and things like that. It just says what it does on the tin. And I. Oh, I like that idea. Why don't we try that? And it will go through. Okay, I'll rename it, renames those 20 images like that. And you go, oh, I like that. Actually, could you do that again, but this time add the date it was done, because I don't like going into when it was created. And, yeah, sure. Da da, da, da da. That's cool. Oh, that's good. That's good. And by the way, now that we got the 20, could you just create some sub files, which is like one for website and one for printing? Oh, yeah, sure. There you go. Like this. That's great. Yeah, yeah, that's good. Oh, that's really good. Yes. And by the way, I've been meaning to email my business partner a proper email at that logo there, and it's been on my mind to do list for like a month. And you go, right, sure. Is it, this one here and it's like, yeah, I think it's that one. Could you just add that to an email and tell him I'm really sorry and it'll open up your email, it will then add in the jpeg. It'll. Then you'll talk to it and you'll say, say something like this. Say something like that and it writes and you go, that's, not like this. All right, yeah. Oh, that's good. I go in and edit. Right, send. I've sent it. Done. fine. Great. Now, where were we? And it's like, oh, we were going through your folder structure. Oh, gosh, yes. Isn't that incredible? Where were we?
Erica: It is incredible. So, okay, this is Claude Cowork. Claude Cowork. You can download it for free, and you have some. What do you have access to? Free?
Darius: I have no idea. I'm like on the 200 quid a month plan or whatever it is.
Erica: but how, much is 200 quid?
Darius: $200 a month? Wow.
Erica: Wow. Most people can't afford that.
Darius: I know, I know, but what does
Erica: it cost for just like the average
Darius: joe, 0 pounds, $0 free for, you know, pretty generous use. You know, like you're not a power user but use it periodically, etc. And what it'll do is it'll say, right, you've used up your allowance for the, this four hours. You'll have to come back in a couple time and use up your next allowance like it used to do for, AI tokens. And then there's like a $20 a month tier, then there's $100 a month tier, then there's a $200 a month tier. And at the hundred to $200 a month tier, it starts acting like a personal assistant that you would pay $1,000 a month for sort of thing, you know. Yeah. So it depends on how sophisticated you get with it. And I've just shown you the most obvious silly type use of it, you know. But for me that's significant because I've got dyslexia, I've got adhd. All of this procedural administrative stuff is a nightmare. Okay, I'll give you another example.
Erica: Okay, well, first, how does it feel now? You said it was a nightmare, all those procedures.
Darius: Feels so much better. I have to go to the next level. So I use another next level product called OpenClaw, okay. Which is an open-source version of this. And at the moment I'm using Hermes, which is sort of the next level
Erica: up from all these Programs. It just sounds overwhelming.
Darius: I know.
Erica: Like listening to you. My. I'm going like.
Darius: I know, but. But the point is, when you hear it from me and it's really hard work for me to get it all working. Yes. But in six months to 12 month’s time, when you're listening to this or you're thinking about this and you're looking out for it, and you start thinking agents. That's what Darius means when he's saying agents. Okay, so when people start saying an AI agent, that's what I'm talking about. Sponsored by Ivy. Imagine turning your meetings
00:20:00
Darius: audio into a live mind map instantly so you remember what matters. Well, try Ivy for free now at, Ivy App. that's IVVI App.
Erica: sponsored by the Executive Functioning Coaching Assessment. A, ah, quick online assessment that uncovers challenges and develops personalized strategies for success.
Darius: So you've got different levels of AI development. You've got the AI chat level and then you've got the AI agent level. And so the chat level is where you have. It's kind of like going into a doctor's waiting room and you have a consultation with a doctor or a lawyer, an expert, and then you go out the door, and you have to do everything yourself, right?
Erica: You're just having a discussion. You're just having a discussion, and they're just, they're brainstorming with you, they're coaching you.
Darius: And the next time you have a discussion, they don't remember their previous discussion. You know, when you go into the doctor or some professional, and they're like, they're, they're bluffing. They're kind of like, oh, yeah, yeah. Hi. Hi. Hi, Darius. Yeah, yeah. How are you now? How was it since we last met? And they're just picking up from you? Oh, gosh, yes, yes. and you're just reading the notes while you're chatting or whatever, that kind of professional. I'm too busy to actually have a relationship with you and remember what you're saying type thing. That's what AI is like right now. Every time AI wakes up, and you start a new chat, every single time, it's like, it's just being born is like suddenly being dropped into a new, world. Like, think of a movie. Suddenly somebody wakes up in this new setting and they're looking around going, oh, my goodness, I've just been teleported in here. What's going on? You know? And then someone starts talking about talking to you and say, Darius, you're a great doctor. You need to tell me a little bit about how you did that surgery. Oh, gosh, yes. I'm a great doctor. Surgery. Yeah. Yeah. What do I know about surgery? Yeah, surgery. Yeah, yeah, yeah, yeah. What kind of surgery are you interested in? Well, you know, you're an expert in, the gut microbiome and cancer. You know, gut surgery and so on. Oh, yeah, yeah, yeah. And in your brain, you're going gut surgery. What do I know about the microbiome? That's what the AI is doing all the time. And it does not remember the last chat you had with it. Okay. And what you'll see in the chat, if you have a long conversation, one chat, once it gets to the end of its context window, 2000-005000-00100,000 context window, it's gone back to that because it's over. It doesn't remember. It starts again. And you're saying, I thought we just talked about this. And it's like, oh, yeah, yeah, yeah, yeah, we did. Yeah, we definitely did talk about it.
Erica: So are you saying that, Claude. Coworker.
Darius: It remembers. It remembers.
Erica: So it really gets to know you?
Darius: Yeah. So what it does, okay, is it creates a system of notes. Okay? Think about,
Erica: Is that. Is that saved on your computer since you downloaded it?
Darius: Yes. You can see what it remembers. So you know the movie for 50 First Dates?
Erica: Haven't seen it.
Darius: Okay. Or it's like 50 First Dates or something. So is this. This woman is such a lovely movie. This woman loses her memory, okay, in a car crash. And this guy, they fall in love in this day, you know, this one day, head over heels. And then the next morning, she wakes up and she can't remember him. And he remembers her and the whole family that's, supporting her and so on says, look, mate, you just can't have a relationship with her. This is just wrong. She's lost her memory. She remembers all of us, but she. She doesn't remember new people or new things, okay? And that's what it's like with an AI. Once it's done its training, it's remembered everything up until that date. Everything after that, it does not remember. Okay? So what he does is every day, he goes back and shows her a video of the day before and says, this is before we meet. I, know. You don't know me. And he brings this VHF with a TV on top, hits play, and she watches it and she goes, my goodness, is that what we did? Is that who we are? Yeah, we're in love. This is you. This is you sending a message to yourself in the future, like for today. And she speaks to herself, you know, remember this, remember that. This is important. You've got a dad, you've got this, you've got that. It's like dementia sort of thing. And he does that for 50 days and then they start living together and they're in love and they continue. But every morning she has to wake up. Same with AI. So what happens is these models sort of create. This is who Erica is, this is what Erica likes, this is what we did before and leaves little notes to itself because every time you finish that chat, it's gone. And a new AI picks up from where you, left off with the new chat and it speaks to the next AI for the next chat saying, remember this, remember that, remember this. Yeah.
00:25:00
Darius: And so it creates a memory of who you are. It creates a memory of all of your notes and things that you've done. But it has to do it very, very carefully because AIs get, have a limited working memory too. And they get overwhelmed with a lot of reading and they don't know what's important and what's unimportant. And so it builds up this kind of tree of knowledge on what notes you've got and what information you're doing and so on. So it builds up a note taking base for you, which creates a memory for itself and a memory for you.
Erica: Wow. Wow, this is extraordinary. So I'm going to pull you back to the question, how can this help the average person? Let me step back. How can this help a student, A high school or college student? I guess it can help them keep their files organized so they wouldn't need to get a huge membership. They might be able to do it on the free membership or. Oh yeah, yeah, yeah, right.
Darius: Free membership's fine for students. Yeah, yeah.
Erica: And then, and, and, and then it's just a matter of downloading it. On how easy it is to learn to use. I mean are there, how did you learn? Did you watch videos or did you just YouTube explore YouTube.
Darius: Claude code. I would I would type in Claude co work for students and see what you students are doing with Cloud Cowork. Like you could type in Notebook LM for students or whatever and in 15-minute video, two 15-minute videos, you'll have enough to do way more than what I've just said or just open up Cloud Cowork, and it's got a whole bunch of suggestions. You click on one of them. So I tidy up some folders. Shall I make this for you? Shall I do that for you. Would you like a spreadsheet of all of your commitments over the next three months?
Erica: And where can you find Claude Cowork? Because I imagine that there have got to be people out there that say, oh, this is Claude Cowork, and it's not, you know.
Darius: Okay, well, we need to put a reliable link in the show notes. Swipe up. We'll put a reliable link in the show notes.
Erica: Great. So we will put that in the show notes so people can feel confident that, they are using the right link.
Darius: The right link is Claude AI C L A U D E dot AI and it's got orange sort of crisscross logo that looks like a tumbleweed higglety pigglety. And that's Claude Cowork. And they are ruthless about anybody competing against them or faking them and so on. So it'd be reasonably easy to, to find that reliably. And it'll be in the show notes. But the point is not necessarily to go and do this right now, okay? The point is for listeners to say, this is where things are going and when this becomes super simple and super easy to use, start using it. And I wouldn't say it's super easy to use at the moment. You know, I would say it's still very early days and you have to be quite sophisticated to make it work and to kind of trust it. And people are trying to crack the code of how to make this work for regular users who will just jump on and use it.
Erica: Right. For someone that just wants to organize their desktop or their, are their files,
Darius: you could do that right now. Right? Really easily. But the future is, what you can do right now is just way, way more even than that. And I'll give you an example of where it's going from a real executive function point of view. That's just first, that's the aperitif. That's the little taster, sweetie you get when you start at the restaurant. You know, the main meal, second or third course of year six would be something like this. Okay? I've got a version of this called Open Cloth, okay. Which is like a more advanced version, full access and on, more access that I can just let it do its own thing. I don't want it touching my own computer. I don't want these things touching my own computer. Just leave it alone, okay? But you can have your own computer, make your own mistakes, wreck your computer or whatever, that's fine. But you can't touch mine. That's where I'm at. At the moment. Okay. Which is why I bought its own MacBook Mini for $500. And we can talk to each other like I would buy an employee a computer. I bought it a computer, and I want you to do X, Y and Z. So what can I do with this one? Let's call it Roy, okay? Because I call it Roy. It's my nickname. So I'm the executor of my father's estate, okay? Which is a nightmare. I hate being the executor of my father's estate. Because no one thanks you. You don't get paid for it. You cover all the liability. Everything lands on you, and there's no Thanksgiving for it at all. Bottom line is, my lawyers submitted a tax return. And I looked at the tax return and I was like, hold on a minute. I think we're used. After we submitted it, I think we've not claimed for some exemptions. Spoke to a lawyer and said, well, you know, there's this peculiar little rule in
00:30:00
Darius: the UK where if you give a gift to someone out of your earned income rather than your capital, your investments, but you earn some income, and you use that extra income to give it a gift to someone, it's got no tax on it ever. At all. But the problem is you have to go through seven year’s worth of bank statements. You have to evidence all the money that they've made, all the money that they've spent, and you have to prove that they didn't reduce their standard of living in any way as a result of that gift. It wasn't sacrificial. And then it came out of their excess income. And you have to prove that it came out of their excess income. And it involves seven years of accounts. That's a lot of money. That's a lot of time. And the accountant's like; we don't really have time to do that. Just accept that it's okay. You might lose you £1,000 or one and a half, £2,000, but, you know, it might cost you that much just to find out if you're eligible. Okay, Roy. Okay. Two weeks later with Roy, I had scanned seven years of bank statements on, three different bank accounts. So bear this in mind. Month one, three pages long PDF. Month two, month three, month four. 72 different PDFs for one bank account and then another bank account or his credit card, another 72. Another bank account, another 72. None of these are digital. None of these are in a spreadsheet. All just hard copies, etcetera, Scan Them all. Roy took each one, scanned it, turned it into a spreadsheet, organized it, made sure everything was reconciled correctly, et cetera, did the same for all of them, integrated them all one year into one spreadsheet. And then it said, oh, the financial year starts April to April. These are all from January to January. I'm going to change them all from April to April and organize them by financial year. Then for my tax return, it says, right, we need to categorize everything according to these different categories. For this tax submission, which is not normal categories, categorize them all. You know, then it's like, right, we've categorized them all. We need to summarize them all together and work out what his allowance is. Right. There's this gift from you. This gift from there. Is there excess in it? Yes. By the end of those seven years of accounts, saved us £20,000 in tax.
Erica: Well, that paid for Claude for many years.
Darius: Yes. And what it meant was I could, hand on heart, say I did absolutely everything I could possibly do to protect the beneficiary's interests. Yeah.
Erica: Without it taking you. I don't, God knows how much time.
Darius: Yeah. Or I could have given it to an accountant. And the accountant would say, right, it would cost about 400 to 5, $600 per year to present. And then at the end of those, that's $3,000 to go through those seven. And then they might say to it works out that you'll only get $1,000 back. And you're like. And then the beneficiary will say, why did you waste that money? do you know what I mean? So here's the message in all of this. All those things that, you know, you should be doing, but you don't have the time or headspace or capacity to be doing. When you've got an AI agent, it becomes so much more doable. And it's not that you end up. Yeah. It just. I can't tell you, that was a good example how much freedom and peace of mind that gave me, you know, because if another accountant did it for me, I would still have to go through all of again to be sure, because I still carry the liability and they make a mistake.
Erica: The bottom line is AI is going to be able to do more our taxes pretty soon.
Darius: Oh, yeah.
Erica: And it's going to be a lot more reliable. I mean, I know that I'm always a little ambivalent. I'm like, I don't believe that my tax person didn't make some mistakes. I mean, he got my address wrong the last two years, and if he got my address wrong, what else did he get wrong? And how much money have I lost? And blah, blah, blah, blah, blah, blah, blah. But it's too complicated for me. But I'm thinking within the next few years, it's going to be easy to do your taxes well.
Darius: And probably now it's easy to check your taxes. So, for example.
Erica: Oh, I could just put my. I could just.
Darius: You could put it in my.
Erica: And say, are there any mistakes?
Darius: Yeah. Ooh.
Erica: So Claude, AI knows the tax system well enough to be able to say whether there are any mistakes.
Darius: Oh, yeah, probably better.
Erica: And is it safe to put your taxes into something like that? I guess you don't necessarily have your tax ID number on there. That's a cool idea. I'm going to.
Darius: Yeah. And the thing is, here's what is different than a chatbot. You can put that into a chatbot. Okay. You can
00:35:00
Darius: put the chatbot in. Okay, into chat, paste your taxes in, add it in, and it'll give you feedback. However, let's say you put in your accounts, okay. Into an agent who has a digital memory of all the things you did in your business for the last year or your personal life in the last year, it would go through the tax return, and it would cross reference it and it would go, I. You bought a computer, Erica, a new computer last this year. I don't see that on your tax return. Now. Your accountant doesn't know you bought a computer. You forgot you bought a computer in that tax year. I'm just using it as an example. But it remembers because it's looking out for these sorts of things. And you go, oh, damn. Yes. I use my personal credit card. I forgot because my business credit card wasn't with me in my wallet. But I just needed to get it done. And you get in touch with your accountant, you go, I just noticed that that wasn't in there. And you get that tax deductible. So it's all those silly little things where you forget certain things. And this is the value of notes, of, taking notes and having a record that is accurate of what you're doing. And ironically, note taking is becoming even more valuable when you've got agents because the agent can look at your notes and go, oh, I get where you're coming from.
Erica: Wow, very interesting, Darius. Very eye opening. I look forward to exploring this cloud Cowork. I think I'll download it.
Darius: Yeah, there's going to be others. Like, you could go to perplexity computer, they're doing one as well. OpenAI will bring one out if they've not done so already. By the time you're hearing this, it's developing every week. there, there's various versions of it. Personally, I think if you're more technically minded, you start thinking about, no, I shouldn't be recommending that, but you know, like Open Claw and Hermes and things like that. But the one thing I would say in all of this is please take notes. Please put your files, important things in one place. Like if you've got a chat in Claude or in ChatGPT and you think, gosh, this is really good, don't leave it in its notes. Say, turn this into a markdown document or a Word document and it will take it, turn it into a Word document and then save it to your favorite note taking app or Google Drive or whatever. But put it into one folder somewhere and you may never look at that again for the next six months or year. But I don't guarantee in the next year to two years you will say, gosh, I wish I had sort of those conversations because that captured my thinking process, you know. And that's one of the most important things that AIs need for you is not the finished document but the thinking process. Because if it understands your thinking process, it goes, oh, that's how Darren Science organized. Yeah.
Erica: And find the little loopholes or.
Darius: Yeah, well, not organized, but think your way.
Erica: Well, it might be organizing.
Darius: It might be, yes. And that's all super, super superficial. But at that point we'll be thinking much more complex things like Darius wants to do his tax return, his lawyer, his accountant wants all the information. And then it will proactively, it will say, ah, right. I remember last year the accountant asked for X, Y and Z. Last year we had to do this, that, and the next thing and so on. And last year we had this problem and last year and this and that and so on, and it'll pull together and it'll make a plan. And I know this is the way you like to do it, Darius, et cetera.
Erica: But, but can we trust it? I mean, I'm going to go back to that. I'm going to go back to the, you know, can we trust that it didn't make a mistake?
Darius: No, no, we can't. No, we can't. So that's where you need human beings to check. So I'll take the example of my accounts, okay? The, the, the.
Erica: Or how about the example of just double checking my taxes I put my taxes in there. I say, did my tax person make any mistakes? And they say, oh, yeah, he made a mistake here, here, and here. Then I go back to my tax person and say, is this true?
Darius: Yes. No, you don't. You go back to another instance of AI, and you say, you are a tax auditor, and you're the tax man. Review this, okay? And then it says, right. This is not quite in the right place. I would prefer this. I would prefer that, and so on. So it's not necessarily mistakes, but there's an inconsistency here. And so it starts moving from mistakes to inconsistencies to. This is. So
00:40:00
Darius: you ask it to mark the work as well. So you got to bear in mind how the AI is thinking, okay? Every time you ask it to do something, it's kind of like, I'm not sure. I'm in this room. Who am I? What am I meant to be doing? And it's saying, are there any mistakes here? So it'll go, all right. Am I meant to look at this as a forensic accountant or just point out some few things?
Erica: So you have to be very specific out who, what paradigm they're coming from, what perspective, who they are, because you're right. So a doctor may look at the same situation, very different than an artist or. Yeah, I get it. I get it.
Darius: And also, you know what it's like when you check your own work for typos and then you miss it.
Erica: Yeah.
Darius: You've been really thorough. The AI has been really thorough and diligent. And then you come back the next day, and you check it again. There's some obvious stuff there. Or you give it to an editor, and they pick up other things. You have to do exactly the same thing with AIs, where you give it a fresh start. And you go, now look at this as if you're an editor. Great. Now look at this as if you're a tax accountant. Great. Now look at this as if you're the taxman. Great. Now look at this as if you're going through overwhelming.
Erica: But I guess you're just saying if it's something really important, like your taxes.
Darius: Absolutely. If it's important, you do that. You don't just take the first run. And when I was doing this, as an example, those seven submissions, it would go through it, and it would say, these items haven't been categorized. Okay. They've just been left as question marks. Do you want to categorize them? And I go, no, I don't actually. And then other little things, and then that's not quite right. This is not quite right. It's not like it. So the other.
Erica: So you're still conducting.
Darius: You're still, you're still conducting.
Erica: You're still in charge.
Darius: You need to work out what the grade level is for a pass. Okay? So in the, in the real world, when you get work done by professional, they're not going for 100% accuracy. They're going for enough accuracy to not get into trouble. That's the bottom line.
Erica: Yeah.
Darius: And everyone in a legal profession or accounting profession or whatever, profession, where will I get into trouble? Where am I? Okay. And that line is different for any sort of situation, you know, and so we've got to kind of start realizing where that line is. Okay. So for my accounts, I was like, if we got audited by the HMRC, what would HMRC say? And hm HMRC would look at it, and they would say that this is pretty good. But we would want more evidence, right? I don't want that. I want all the evidence to be in here. All right, well, if that's the case, then we would need this sheet with this information. All right, well, go get that. Have you got that? Yeah, I've got that. I'll make it for you. Done. All right, fantastic. So you, what you do is you create a grading system, and you say, mark this out of 100% as if you were an accountant. Mark this as if you were the supervisor of that accountant. Mark this as if you were the taxman. And I would tell it to mark its work.
Erica: Interesting. Okay.
Darius: And it comes out with a mark.
Erica: The bottom line is it's going to get more accurate. But until it gets more accurate, just make sure that you're double triple checking and looking at it from different perspectives.
Darius: or in a way, another way of saying it is make sure that you are getting it to mark its own work. And then you check what it's marked.
Erica: Okay.
Darius: That's the key. Because if you then take on the responsibility of saying, God, I've got to read through all of this and check it and so on. No, nobody wants to do that. But the thing is, AI agents are really good at marking things because that's how they were trained. So when it starts marking something, you start noticing the way it grades it and go, well, if I'm, you know, the supervisor, I would say, let's mark it for accuracy, let's mark it for readability, let's mark it for auditing. Let's mark it for this let's mark it for that. It's got an inbuilt marking structure. And then it will test it against that. And by Jove, it will get and say out of 10 for each element. And so some will be a seven, some will be a six, some will be a nine, and some will be a three. And you go, right, let's go through all those threes. And you get everything above five. Great. Is everything above five? Does that mean that I'm safe? Yes, you're safe. The accountant safe. But five means there might be extra work for you in the future. Right, let's get it up to a seven. Up to a seven. And it's the same for, like, I work with some people who are like, counselors. And a counselor can sometimes be called by a court if they're counseling someone with anxiety or whatever, commit
00:45:00
Darius: suicide. The coroner says that we want to see your report of your meetings. Not to see whether or not you're guilty as a counselor or whatever, but to prove that they had suicidal thoughts and this wasn't a murder, but realistically it was a suicide, etc. And some counselors have adequate notes, and they get called in to the coroners for a day to give testimony. But if they have detailed enough notes, they never need to be called in, which is why their supervisor keeps saying to them, look, okay, these are adequate six out of ten, but you need them, at an eight, ideally, to just save yourself the hassle of when this happens, because it's not if, it's when. And that's what AI can do for you if you tell it to go and mark your own work.
Erica: Wow, this was very enlightening. I really appreciate this discussion. I think it's. It's very interesting to see where AI is going and how it can help us.
Darius: Pleasure. Until next time.
Erica: Until next time. Sponsored by learningspecialistcourses.com Courses and resources that support educators and coaches.
Darius: Sponsored by Ivy. Imagine turning your meeting's audio into a live mind map instantly so you remember what matters. Well, try ivvi for free now at ivvi.app. That's ivvi.app. Thank you for joining us at, the Executive Function Brain Trainer Podcast.
Erica: Check out our show notes for links and resources and follow us on social media.
00:46:38