Introduction

I'm going to give a bit of talk on, essentially, it's partially using language models and AI. It's partially using other stuff.

And in terms of backgrounds, so I am not an engineer. I didn't come from a technical background. My background's mostly been in capital markets.

But I'm really energized by using this stuff. And so I've kind of gone a long way with using no-code coding tools. And it's made me realize that actually, In order to build quite useful things in your life, you really don't need that much technical skills. And there's a lot that you can do without needing to be an engineer.

Personal Journey

So I guess this system that I'm going to demo and show kind of has some background story to it, which ultimately I think the beginning started something like five years ago when, let's just say it was a really tough time in life. I was just not happy with my life in a lot of metrics, like was essentially going to bed at 2.30 in the evening, was like 30 pounds heavier than I am, financially not in the sort of same place that I'm at now.

The Power of Goals and Systems

And so I think I read this book from James Clear and he talked about, well, why do some people succeed and other people don't succeed? And we put all this energy into setting goals. And so as part of my annual goal process, I would go back and review, okay, how did I do versus last year's New Year's resolutions? And inevitably, you always, you know, look at like, okay, I didn't match up.

And I think he came at this, like, part of the book, he talks about that, look, everybody has the same goals. We're all trying to get to the same outcomes. But what differentiates the people that win and the people that don't is actually the systems that they have underneath that. And so what are the processes that they do every day?

And how do you win in the small moments, whether that be in this half hour, this day, this decision that I'm making right now? And those compound over time if you can make good decisions and kind of add them up. And so he gives a couple rules talking about how to improve habits and what's helpful. And one of the things he talks about is basically making it as small as possible to not do the thing. So in terms of improving a habit, he might just say, okay, if you wanna lose weight, just simply log your weight in a spreadsheet.

Starting with Simple Metrics

And that's basically where I started. I was like, I'm gonna log a couple metrics in my life, whether it be time I wake up, did I do a workout today, this and that in a spreadsheet. And so started with that, put it in Excel every day.

And over time, basically, I went from zero to like a little bit, and then right back down to zero. But over time, what I realized is that actually by tracking it and paying attention to it, it became a more sort of compassionate way of dealing with improving habits. Because even if I had a downtime for a couple months, I could still look back and be like, well, at least it was better than last year. Or at least it was better than the year before.

And so, sorry, that was just like, how often am I doing workouts and exercising? So I started then expanding this to lots of other metrics in my life, whether that be anything that I want to improve in life, I would start to say, okay, what are metrics or KPIs around that that I could track, put into a sheet, and cover? And pretty much inevitably, everything that I paid attention to would gradually get better. It wouldn't happen in a straight line, and it would take time, but I think people really underestimate the power of compounding really small changes over time.

Expanding the System

So then what I realized, though, is a lot of people in my life were like, oh, this would be great. Can you build a spreadsheet for me? How could we do one? So I'd set up these templates for people.

And in every single case, people would go do it for like three weeks, and then they would fail. And so what I realized is that actually out of that, the key is you've got to make it easy.

Leveraging No-Code Tools

1which how you make it easy is you use systems like make.com. So if you're not familiar with make.com, it is a no-code coding platform where you can basically plug one app to another app to another app to another app and get it to do lots of different things. And then within that, you can also incorporate calls to language models to do different things as well. And so for someone that doesn't know how to code, it really gives you an ability to do stuff in the background.

And as soon as I started taking things like, OK, I want every time I do a workout this to go into my spreadsheet automatically so I don't have to think about it. I want every day the data from my sleep data to go into my spreadsheet so I don't have to think about it. The more I made it easier, the more it became easier to build and build and add to the system.

So then I basically started saying, okay, can I do this for lots of other metrics? And so that was across you know, things that I wanted to learn across different measures of finance and across different things in sort of fitness.

Demo: Airtable and Task Management

And then basically the other big realization, which I'm going to demo Airtable and show some stuff that I've built in Airtable, is that I kind of graduated from Google Sheets to Airtable. I'm just going to run real quick through sort of what I've built is basically a centralized place dashboard for lots of metrics in my life. And then I've built on top of that or connected to that is a task, sort of a task application that I incorporate some elements of AI.

Incorporating AI and ChatGPT

And the reason why I talk about this in an AI chat is mostly everything that I've built here was me kind of questioning ChatGPT and saying, how do I do this? OK, I got an error. How do I fix this? And this sort of iterative process.

But effectively, in terms of dashboard, so I've got a dashboard that covers lots of different elements. But I track basically time I woke up, and I'm putting all these things from different apps into one centralized place.

McAfee's interrupting. It looks a lot better on a big screen.

So from that, I then decided, okay, this is great, but how am I managing the things that I'm doing in terms of my to-do list? And we all have our own different systems in terms of your to-do list.

But for me, combining, okay, the things that I wanna do, my goals, the metrics that I'm getting out of my life, and then the tasks that I'm doing, what are the things that I'm doing to kind of move me towards those goals? Putting that all into one centralized place, which was what I was just going through, was really helpful.

And so I've kind of gone on this journey through it with task management systems where originally I just started with pen and paper.

Where was I? Yeah.

So originally, I just started with pen and paper. And over time, then I adopted a tool called Todoist, which was great. But I think I ran into issues with that in terms of it didn't structure everything the way I wanted it to structure.

So essentially, for every task, from reading some other books, it was like, OK, I need a tag. Does this task take more or less than an hour to do? And is this important or not important? And by basically building a matrix out of that, I was able to build a dashboard that was able to kind of focus me in on what are the most important things that I need to get done every day.

Innovating Task Management

And so I think now kind of getting into the elements where I've incorporated language models in different ways. So I guess the first one was just kind of using it as a coach along the way. But the second one really is coming back to this is taking unstructured data and then structuring it.

So essentially in my to-do list I've been testing different formats of like how do I write down everything at the end of the day. Because for me like when I'm on the bed like I don't like having screens around me so I like doing it physically. But the challenge for me was then I would have to-do lists in my app and then I would have other things on my paper which wouldn't sync up.

And so then I started basically, oh this is Toomey who usually sits next to me while I'm going through and doing my to-dos. But I started basically creating these sort of printouts. And so I basically built a sheet that I was like, okay, if I print this out for my tasks, And I didn't bring any with me, but I've got a stack of them lying next to my bed.

And what I do is I basically write down the task. And then I write down, OK, what category does it sit in? And is it important or not? And so I'm just kind of drawing on a page. But when I take a photo of this and then upload it to the system, that converts it to JSON. So essentially, you're taking unstructured data and moving it to structured data, which

With the launch of GPT-4.0, it's actually super fascinating how easy you can just take photos of things and then get the data out of those images and use that in a thoughtful way.

So then once I built out this dashboard and to-do list system, I've then been adding different elements to it. And so here I can kind of match, tag my tasks.

AI-Driven Daily Pep Talks

But I think sort of the interesting AI element that will get me into something further is this automation that I've got running in the background, which So essentially what this is going to do, it's gonna generate me based on my to-do list and based on a couple other variables, it's gonna generate a sort of a daily pep talk from the speaker of your choice.

And so if someone could just give me a task, I'll add it to the task so we can show that it's live and not like a recorded thing. So if anyone wants a shout out task, I'll add it here.

I'll just put London.

Yeah, there you go. And I'm going to say it's going to take longer than an hour.

So once I've created this in my Airtable, essentially what I could do is go to my tasks, my record review. It should be there. Let's just see.

Plans. There you go. And then I'm gonna mark this as, mark done. Oh, I lost my mark focus button, but I can do that here.

So. Oh, it isn't focused great.

OK, and then in terms of speakers, we have the option of I had many more options, but I've I've slimmed it down to just Joe Rogan or Samuel Jackson. Can I get hands for Joe Rogan? Hands for Samuel Jackson. OK. All right, great.

So from my dashboard here, I can then go pep doc generator, hit generate. This is going to tell me that it's run. And then what's happening here is you'll see there's a bunch of different modules that there's something happening here. And effectively, it's fetching information from various different ports. It's organizing that information. It's then putting that into a prompt that has a lot of structuring around the prompt that then produces the text. And then it's taking the produced text, going to 11 Labs, producing the audio recording. And then from there, it's saving that in Google Drive and giving me a link back in the dashboard.

So 7.22. So folders, pep doc uploads, and 7.22.

Rise and shine, Alex. Today is Tuesday, July 16th.

I don't need a weapon. I am one. Current conditions. It's 18.5 degrees Celsius with partly cloudy. Expect patchy rain later, high of 20.2 and low of 13.4.

Here's your schedule. One, MindStone meetup at 1.30 p.m. Two, Mariam Alex at 5.30 a.m. Three, Chinese class at 3 a.m.

Top three to-do items. One, cycle from Land's End to London. Two, find examples of interactive Airtable interfaces. Three, create demo for MindStone AI meetup.

Remember, Alex, you're the weapon. Own today like a boss. Thanks.

So the last time I did this demo is in New York, so I still had it at Eastern settings. I'm not doing Chinese class at 3 AM. Yeah.

Behind the Scenes

But so let's just get into the nuts and bolts of how we can put something like this together and what was actually happening behind the scenes.

So essentially, on this path, it looks complex from a distance. But actually, when you kind of get close to it, you realize it's not that complicated.

Effectively what it's doing, and I've broken it down into some steps, is first it's going and fetching information from a couple different places. So one, it's gathering the weather data. Two, it's gathering what's on my calendar. Three, it's gathering what's in my to-do system.

Right? And so essentially what we're doing here is basically just doing an API call, which you can do quite, I mean, if you think about these tools like Zapier and Make, they have something like 20,000 native apps that you can access on the system. But once you realize that you can make third party API calls, that basically opens up the world to millions of different endpoints that you can fetch data from.

API Calls and Prompt Structuring

And so then basically what we're doing is taking that API call to the weather data, extracting the JSON, and storing it in a Google Sheet. And what I've found in terms of prompt structuring and prompt coordination, I have yet to find a better system or a better tool than a well-organized Google Sheet. And to demonstrate, I'll talk about what makes a good prompt.

So essentially with a great prompt, I think there's kind of three elements, right? So there's the instructions of what you want it to do. There's the input variables of what's the information you're passing in. And then there's the output instructions or the output format that you want it to look like.

And so I have a sheet like this for every automation that I build. And so here I give it some background context on myself. I've written a few paragraphs that sit in the background.

I toggle the speaker here, which I said I had more, but I've limited it to two. And then I give it a tone. So one of the things I do here is I add instructions to the prompt for how I want it to sound.

Now, What you'll find if you don't do this is that ChatGPT sounds very mechanical. You can tell if you got a post or an email or something like that from ChatGPT. But I think there's a couple of hacks around this.

And my favorite one is basically giving it a bunch of things that you've written before and then asking it for a detailed set of guidelines for how a language model could replicate and sound like you. And so in all these things, I have different tones that are basically sitting in the background. And for this one, you can see if I expand it, So things like maintain an informal and friendly tone, et cetera, et cetera, et cetera. ChatGPT actually generated these, and then I'm just using these and sticking them back into the prompt.

And so with all this information, when that automation runs, it's going and fetching all those information, and each one it's printing to a different part of this Google Sheet. So it's basically printing them all here. And this is just the really raw data. Now what this just does is it just organizes all of the background information.

So essentially what I have here in Google Sheets is all the background information I'm putting in. Then the prompt that I'm passing to ChatGPT, which if you pass in a very long, so this prompt is, I can tell you, This prompt is 2,600 characters long, as opposed to the normal character, which might just be improve this email. You get way better results when you have a really detailed prompt that dictates what it is you're looking for.

And then here, natively within Google Sheets, you can then exactly just pull in. So if you use the ChatGPT plugin in Google Sheets, you can just get ChatGPT to run natively within your Google Sheets. And that's what's happening here. So this is the text that was just generated.

So coming through this quickly, essentially if you look at like, this was the first element of when I just said, hey, create a pep talk for me based on this information, versus once I gave it a really well-structured prompt, it came out with a substantially better outcome. This I've talked about, this I've talked about. I'm happy to answer questions on it or go into more detail afterwards.

Voice Generation with Eleven Labs

And then the other bit here is then Eleven Labs. So if you haven't used Eleven Labs, it's a really fun tool.

Essentially you can jump on and you can give it a recording of any amount of footage of a speaker. Of course, there are some things where you shouldn't be taking someone else's voice and you do have to then confirm in that voice that you would like it to essentially be able to create a voice. But if you give it a recording of a voice, you can essentially create a voice to mimic that.

There are ones from celebrities out there like the Samuel L. Jackson one and these ones already.

Summary and Key Takeaways

So yeah, in summary, so in summary, I guess a couple key points.

One, I think if you're not tracking your habits, I would really suggest doing it. It changed my life massively, and it's essentially if you want to do it well, I think building a couple systems that kind of do that in the background, super helpful.

I guess take two here is that you really don't need to be that technical to put this stuff together. It really just kind of takes you willing to kind of spend some time and fiddle with it. But I think as these models get better and better, then I think these sort of systems, you can just continue to add to that over time.

And I've seen some really impressive ones where people are using it to answer all their emails and generate all this content and do all these different things. But it really starts with kind of building a base and having a centralized database.

Connect and Learn More

And then, yeah, last thing, I guess, is if you're interested in this stuff and want to talk about it or have detailed questions, we are launching, well, one. I'll put my details on the screen. So if anyone wants to connect with me, this is my LinkedIn. Happy to chat about automations or whatever you want to chat about.

And then separately, I'm gonna share a separate link after this, which I'm working with MindZone to build a course on AI mastery, which is gonna launch in September. It's gonna be a three-week course. And effectively, we're just gonna cover how you can build AI tools from a no-code background.

So if you're a super senior AI developer, which there are some here, it might not be the most appropriate content for you. But if you're a business leader or if you're an entrepreneur and you want to incorporate this stuff in some way, we're going to cover sort of the groundwork of how you can put this together. Not this specifically, but there'll be other examples.

Yeah, and thanks for your time and attention.