The New ChatGPT Playbook & Prompting Guide

Matt Wolfe· 39 min· 8,181 words· 37 min read· English ·Watch on YouTube

This is the full transcript of The New ChatGPT Playbook & Prompting Guide, published on YouTube by Matt Wolfe. Every paragraph carries the moment it was spoken, so you can click any line to jump straight to that point in the video, search the whole thing for a word, or copy it out.

0:00In this video, I'm going to share with you the ultimate guide to Chat GPT 5.6 and the updated Chat GPT app that combines [music] all of OpenAI's offerings into one place. I'm going to walk you through the features, the how-to, and then get into some detailed prompts of some of the crazy things you can do with this app and model. Let's get into it. Oh, by the way, before we get into it, this video is sponsored by OpenAI, but I owe it to you to be honest and truthful with everything I say. I always do that with all of my videos. So, everything in this video is truthful and my experience and

0:34also me showing some other things that other people have done with it. But, I do appreciate OpenAI for agreeing to sponsor this one. All right, let's get into the good stuff. Let's start with a bit of a tour of the new Chat GPT app. Because this new app is essentially OpenAI taking Chat GPT, taking Codex, taking their browser app, and taking their new Chat GPT work platform, and mashing them all together into one product. So, you don't need to switch between tools all the time. So, we have Chat GPT Codex. You get to it by clicking on the little drop-down in the top left here. And this is the portion

1:09of the platform that's for developers. This is where you're going to be writing code, prompting it to write code, reviewing diffs, fixing bugs, exploring code, that sort of thing. And then you have the new Chat GPT work section of the app. This is the section that's for like regular people that just want to give this tool actions to go do on their behalf. This is where you're going to have it be like your assistant and have it connect to your various tools and just go do things. I'll get to some examples of what that looks like in just a minute here, but let's continue the tour for a second because there's some

1:40other really cool things they've added to this platform. You've got sites over here. This is OpenAI making it so you don't need a million different tools to get something online. You don't have to deal with web hosting or back-end databases or CDNs or anything like that. When you're building something like a website or an app or a game, you can actually tell this new app, send it to sites, and it will be live and online,

2:03and you can share it with anybody. Anybody can access it. It is all just hosted for you now with this new sites feature inside of Chat GPT. Now, there's also a browser built-in. If I toggle our side panel here, you can see there's a link to a browser. If I click into this, I can open this window wider, and we have a browser where you can go

2:22anywhere. For example, futuretools.io. One thing that's really cool about this browser is you've got this annotate feature. So, let's say you're working on a website inside of the new Chat GPT app, you can tell it exactly which section you need help with. Like, let's say I want to edit this brain here, I can select this section and annotate it

2:40and say, "Let's add more green to this." And you can see over in my chat window here, it added the one annotation that I just gave it. If I was actually working with the code for this website, it would know exactly what I'm talking about. I don't have to give it more context because it knows which part of the screen and what I want it to do. You can also import your cookies and passwords from your other browser if you want. So, if you're already using something like Chrome, you can pull them all over, and it will be able to log into those websites on your behalf, which is really

3:10cool, which I'll show you more examples of that in a few minutes here, too. But, let's continue the tour here cuz there's still more to show off. You can create scheduled tasks. So, let's say you have a prompt that you just want this prompt to run every single day or every single hour. You can create these scheduled

3:26tasks that will just run on autopilot. Let's say you want it to do something like clean your downloads folder. You can have it automatically clean your downloads folder for you and organize things into folders at a regular cadence once a day. You can have it create daily briefs where it'll read your calendar, your unread emails, things like that, and give you that brief. You could do weekly reviews or follow-up monitor where it will watch your emails and let you know when you need to follow up with somebody. This comes in really handy and I have a whole bunch of scheduled tasks as you can see. Later in the video,

3:57those tasks that you're seeing on the screen will make more sense of what they're actually doing. You've also got this chat up here. This is for when you just want to ask Chat GPT a question. So if I click this chat, it'll open up this sidebar chat and this is really handy because you can have this sidebar chat

4:11open while you're coding something else. Let's say I'm working on Future Tools over here and it's coding up some new feature for me. I can be working on the code right here while having a separate side chat and asking a question to Chat GPT over here. And of course, it connects to like every app you're going to use. You can see what I have it connected to right now. If I show you my full list, you can see I've got it connected to the full suite of Google Docs. I've got it connected to GitHub, Google Drive, Gmail, Slack, Google Calendar, Canva, Figma, Superbase, Vercel, Granola, Todoist, and a handful

4:46of other like little skills and plugins that I've installed. But because it's connected to so much and because I allow it to actually use my computer and because I actually gave it access to use the browser, it can like do anything on my computer that I could do almost. It also knows everything about me and my business because it sees my Slack conversations, it sees my Gmail conversations, it sees my Todoist, it sees all of my previous chats I've had with it, it sees all of my meetings that I have on Granola. So it knows everything that's going on in my business at any time which makes it

5:19really, really easy to ask it questions and get advice on my business because it's got so much information about it. And again, you can allow it to use your computer as well. So if you go into settings and computer use, you can allow it to use any app for you. You can allow it to use Google Chrome. You can also let it control the internal browser which is really cool when you want it to go and like scan Twitter or Instagram or something like that for you. All right, I think that's pretty much the tour. I could keep on sharing features but let's actually get our hands on it and dive in

5:50and show you what it can really do. So, I'm going to show off a bunch of prompts here. Some of them are going to be prompts that I gave the model. Some are going to be prompts that I found that the community used and created some really cool stuff with it. One thing you're going to notice when using the new ChatGPT app is that sometimes it's going to take a bit longer to get the outputs, especially on some of these more complex and like intense projects to show you. So, what I'm going to do is I'm going to show you the prompt on the screen and then we'll look at and talk

6:21about the output. I'm not actually prompting these in real time as I'm recording this, not because I don't know what the outcome's going to be, but because some of these I would have given it a prompt and it might have taken an hour to complete. Some of them it might have taken only two or three minutes to complete. But because of that uncertainty around the amount of time it took to get these responses, I just went ahead and pre-prompted them all. When you're using this new ChatGPT app, what you'll notice is that sometimes these prompts take a little bit longer to give you a response because your first prompt

6:54is sending off little workers to go and do all of the other little prompts for you and do a bunch of little tasks. So, again, because of that, some of these take a little bit longer to generate the response and give you the output you're looking for, but the results are typically something way better than

7:10we've seen from any previous AI models. let's check out some of these prompts. So, one of the coolest things about the new app is the ChatGPT work mode that acts kind of like a personal assistant. And I've been really, really impressed with what it can do. So, I actually created a project that's pinned to the top of my sidebar here. It's actually the top-most pinned project cuz I find myself using it the most, but I called it my personal assistant project. Now, when I created this project, I gave it a couple prompts so that it becomes a better personal assistant from here on out. The first prompt I gave it was to

7:44create a skill for itself that gives it access to everything. So again, I'm not going to read this whole prompt, but here's what I essentially said. I want you to create a skill where every time I give a prompt inside of the personal assistant project here in ChatGPT, you automatically have access to everything. You can see my Gmail, calendar, Slack, Granola, Google Drive, Second Brain files, journals, all of my other prompts, projects and chats, my memories, what's in my control tower, future tools, my Twitter and Instagram accounts via the browser, my YouTube channel, literally everything. And then I essentially told it that when I'm prompting the personal assistant, call

8:21on this skill so my personal assistant knows it can access all of that. I also gave it instructions that this should be read-only, you don't have access to write to any of those things, but you can read from any of those things. So you can see it created a full context skill, so it'll proactively retrieve relevant context while defaulting to read-only. It created a project-wide trigger, so requires the skill for every new prompt in this personal assistant project here. It's got an evolving source map, so the source map is everything I told it it can access, but also it evolves. So if I add new plugins

8:56or new tools, they get added to the source map as well. So now I've got this personal assistant project that knows literally everything about me that I do on my computer. So then the second prompt that I give to sort of train my personal assistant here is this analyze email and chat voice. And this is so that when I have it write emails on my behalf or Slack messages on my behalf, it knows how to write like me. Now, I don't give it access to write and then send on my behalf, but I do allow it to write for me, save things as a draft so

9:27I can review and modify them a little bit and then click send. But here's the prompt that I give, essentially just saying analyze the messages I've authored in my sent email and work chat over the last 12 months, exclude forwarded material and copied text, and then it created an email voice profile and a work chat voice profile. And so now, when I ask it to draft emails for me or draft a Slack message for me, it has a voice profile to follow that theoretically will sound fairly close to me. It's still not perfect. You're still probably going to want to reword some stuff and take out some of the

10:01ChatGPT-isms, but it gets pretty close. Now, once you've got those few things dialed in with your personal assistant, you can do some pretty cool stuff. So, I'm actually working on a book right now that's going to be through a traditional publisher. I can't talk too much about it, but I had my personal assistant create a book one-pager for me. And here's the thing, since it has access to my previous chats, I can go to a previous chat, right-click on it, and then say copy session ID, and then I can paste those session IDs in, and my assistant will read those past chats specifically. The current working title of the book is The

10:37New Leverage. That's probably not going to be the final title. That's just kind of the code name for it now, but I gave it this prompt. I've been planning a book. I've talked about it in this session, this session, and this session. Help me put together a one-pager or one sheet for this book. I need a one-page write-up that sums it all up and pitches the idea to people. And because it saw my past discussions, it knows everything I'm talking about with the book, it created a one-sheet for that book. Now, I'm blurring most of it cuz I'm still dialing it in and working with people on

11:07it, but uh there's a little teaser right there that you can see. The beautiful part about this was I didn't have to give it a ton of extra context cuz I've already had multiple chats about my book with ChatGPT. I just pointed my assistant back to them and said, "We've already talked about this. Go write me the one-pager." You can have it help build a slide deck for you. This is something I actually did recently. I'm actually doing a webinar with Teachable on July 22nd called Deciding What to Automate with AI in your personal and work life. I had the outline of the presentation. I knew everything I was

11:37going to talk about, but I really, really don't like making slide decks. It's like one of my least favorite things ever. So, I gave the outline to the ChatGPT app along with a brief that Teachable wrote up for me, so it knew additional context around it and basically said, "Make the slide deck for me." And it did, and it looks pretty decent as well. I'm not going to show the entire presentation. I'll save that for when I actually do the webinar, but it went and grabbed screenshots from my actual website, put text on the screen, knew the context of everything that I was talking about, and built out slides

12:11to make the presentation look good. Now, I did have to go back through and me and one of my team members kind of overhaul the slides a little bit to make them look a little bit better, but this got us like 95% of the way there. I know this is very, very meta, but I had it help me outline this ChatGPT video that you're watching right now. Now, I changed up the concept quite a bit, and a lot of what it found for me I ended up using, but a lot of it I didn't end up using. But essentially, I asked it for 50 cool ways to use this new app, and it

12:41went off and did it for me. But here's the coolest portion of it right here. This is something that when I prompted it, I didn't even know if the app would do it or not. So, I was pretty impressed when it did. But I said, "Research how people have been using it with YouTube, X, Instagram, blog posts, etc." I logged into both X and Instagram in the ChatGPT browser in this app, so you can access them to find ideas. Now, it's very expensive to use the X API to go and search for things for you. But if I go and log in on my browser, you can see I

13:12can go to x.com here, and it's logged in as me inside of X here. I can go to instagram.com here, and you can see it is logged in as me inside of Instagram. So, it doesn't need to use the API. It just goes and logs in and skims these sites looking for stuff that it can share. I thought that was pretty cool. Now, it gave me back this long document of all of these ways that people have been using Chat GPT, and it gave me a nice little starting point. Now, I obviously didn't use all of these in this video. I would say maybe I used about five of them,

13:47which I'll get to in a minute. But, it hunted down a ton of really, really, really cool use cases. I just won't be sharing them all cuz then this would be like a 2-hour video. I'm actually speaking on a panel at Open Sauce this week on Friday, and I wanted to feel super, super prepared for the panel that I'm going to be on. The moderator of the panel actually gave us talking points and a list of questions the panel was going to go in. And in order to prepare myself, I gave it this prompt here. Now, I'm blurring some of the prompt out cuz I did give it some like personal

14:18information and some details that I can't really share. But, the main idea of the prompt is that I'm speaking on a panel on Friday at Open Sauce. I've been provided with details and briefs to prepare me for the panel, and then I say, "You know a lot about me and what I do and what I've built. Everything about me is in our past chats, social, second brain journals, projects, email, etc.,

14:37and everything else you have access to." Now, you can't see this cuz it's blurred, but I do go on to say, "I do plan on responding off the cuff and not having a bunch of pre-planned answers." But, I'm essentially looking for reminders of things that I've built, things that I've chatted with you about, and discussions that you know I've had about AI so I have some nice anecdotes and stories to tell at the event. And then I had it just go and find a whole bunch of anecdotes and stories based on what I've actually used AI for. And it actually created this monster doc. This is a 32-page document that it created of

15:13potential talking points. Now, I know not everybody speaks on panels or conferences, but you could use this technique before going into a board meeting at your work or before jumping on a meeting to pitch your product or your service to somebody else. Like having a little bit of a prep doc before going into something like this could help sort of pre-load some ideas into your brain before you go do the thing you're going to do. Another really cool use of the AI personal assistant here, give it a prompt like this. What's on my calendar, Slack, and in my email that I need to do this week? Are there things

15:46you can take off my plate for me? Now again, I do need to blur a lot of this cuz there's some personal details here, but it basically told me like what I have to do this week. It does say must close before Thursday because it knows I'm flying out to San Francisco for Open Sauce on Thursday. So these tasks need to get done in the three days that I have this week before I leave. So for this one I need to record the A-roll and B-roll for a specific sponsor ad I'm working on. I did some interviews

16:13recently. I need to record some intros. It's reminding me that on Wednesday I'm doing a technical run with Teachable for the webinar we're doing. It found some items in my To Doist, put my sort of itinerary for travel up here. But if I scroll down a little bit here, it says work already taken off your plate cuz it already did a few things for me. Help me come up with prompt ideas for the video that you're watching. I'll get to some

16:35of those community prompts in a minute. Some of them are really cool. Help me create the Teachable slide presentation, created an Open Sauce study deck, and help me with a one sheet. So these were things that were on my task list that my ChatGPT personal assistant is basically saying, "Look, I already did all these for you." And then it goes on to say the highest leverage next bundle I can handle is. So it's telling me what it can take off my plate. Create the reply to the book publisher to attach the finished one sheet. My bookkeeper asked some questions, so organize the 25 bookkeeping questions into a fast

17:03decision sheet, add some reminders for me. I mean, I'm pretty dialed in for this week, so there's not a whole lot more I need to do before I leave. But it did find a handful of things that it's like, "Hey, I can take these off your plate for you." All right, so check this one out. This one's pretty cool. This is this is like a an example. This isn't something I'm actually doing, but let's say you're in sales and you need to go like pitch something and find cold leads to pitch to. Here's an example prompt that I gave. I want to sell AI consulting services to small business in

17:29San Diego. Find me 20 solid leads that can use my help and write up a draft email pitching of my services. You can read the whole prompt if you want, but it says done. I created the complete outreach package. 20 research and ranked leads, 20 individually personalized emails, found all the details for these companies here. And then I said, cool, create the drafts inside Gmail so I can review them. And it said, done. Cuz it has access to my Gmail. So now if I

17:53click open Gmail drafts, check this out. It wrote all 20 of these drafts and they're all personalized for the company that it's sending to. Like different subject lines and everything. It plugged in their email automatically cuz it found that for me and it wrote an email that was personalized to them specifically. I'm Matt Wolfe, a San Diego-based AI educator and creator of Future Tools. I help establish local businesses use AI to remove repetitive coordination while keeping the customer relationship and creative decision human. It actually gives some suggestions of things I can create for them and even says if you reply, I'll shoot you some workflow ideas and maybe

18:26we can jump on a call. I mean, it's a decent email and it wrote 20 of these catered to every single individual business. Now all I have to do is review each one and click send. Most of them I probably don't even need to change very much. Now, I don't provide consulting in reality. I just thought it was a cool demo to show off for people that do. So

18:45I'm going to go ahead and discard these. Now, I'm actually going to blur out the response on this one here, but this is a fun prompt to give your personal assistant that has access to literally everything. What do I need to hear right now? Easiest prompt in the world, but because it's looking at all of your emails, all of your slacks, all of maybe your granola meetings, all like all that stuff, it probably knows what you're overwhelmed with, what you're good at, what you're not good at, what other people are doing for for and it can use all of that information to give you a nice little piece of advice of what you

19:14need to hear right now. And it's pretty good. All right, that's my personal assistant. As you can see, I do use it quite a bit. But this ChatGPT work mode can do even more than that. You can give it access to folders on your computer and have it do things in those folders. So, for example, I've used this example

19:33in the past with other agent type tools. But maybe this is the first time you've seen one of my videos, so let's do it again. I can come up here to new task and I can tell it to clean up my downloads folder for me and organize my files into the existing folders there. So, if I open my downloads folder, you can see I have a whole bunch of stuff in here that just sort of accumulates over time. And then down at the bottom, I have a bunch of folders that I've created in the past to sort of sort things into. So, if I tell this, "Clean

20:02up my downloads folder for me and it organize my files into the existing folders there." And submit it. I don't even need this on extra high. Let's bring this down to to sole medium. Submit that. We can see here it says, "I found clean existing category structure." So, it found these folders that I've already got there. And just like that, it cleaned them all up. It's been about a minute and 11 seconds total. And it moved everything into the

20:24proper folders already. And there we go. Minute and 23 seconds total. Organized 65 loose files into existing folders. Another cool thing I'd like to do with this ChatGPT work feature is So, I've got this folder on my computer with a bunch of thumbnails in it, but the titles are all just thumbnail 1 2 3 etc. That's not very helpful to know what's going on in these thumbnails here without me just like clicking through

20:48and looking into each one individually. So, if I right click here and copy the path name to this folder, let me create a new task up here and then say, "In this folder." And then link to my thumbnails folder here. "Look at all of the images and title them based on the content of the image." It will go through, look at each image, detect what's in the image, and then rename the file based on what it sees. Not the most elegant prompt in the world, but it will work. Check this out. There's 21 thumbnails, all standard. I'm reviewing them in labeled batches. The final file names reflect both the visible subject

21:25and any prominent on image text. And I can actually see it happening in real time there. It just renamed everything here. Who will win? If I look at what this image looks like, who will win? This is the new oil. This is the new oil. So, it's titling them based on what text is in the image. So, now it's much easier for me to click through these thumbnails and get exactly what I'm looking for. And that whole process took

21:481 minute and 19 seconds to do 21 images. So, if you have a giant folder of, let's say, screenshots or YouTube thumbnails or images that you downloaded on the internet that are like a stock image folder, you can have it go through and quickly rename everything so you can easily find exactly what you're looking for in the future. All right. Now, let's

22:07switch from ChatGPT work over to Codex. Again, work is like your personal assistant to do little tasks for you and to, you know, manipulate things on your computer and use your tools for you and things like that. Codex is where we start actually writing code. Now, if you're in work mode, a lot of times it will write code to accomplish the tasks for you, but Codex is specifically designed for when you're in developer mode. And one of the best parts about the launch of the new ChatGPT app with Codex was that we also got these new 5.6 models that are really, really good. 5.6 Soul Ultra, I've had it work on tasks

22:45for 18 hours straight and build entire projects for me. So, when I start showing off some of this stuff in Codex, it's kind of a combination of what Codex can do, but it's also a result of these really good new 5.6 models. And one thing these new models are really good at is building websites. It's getting harder and harder to like immediately tell that AI built a website cuz they're looking less and less vibe Cody as time goes on. So, I gave it this prompt, recreate the website for Craigslist without duplicating it. The idea and functionality should be the same, but

23:16let's put our own unique spin on it. Reimagine it to improve UI UX. Now, I did say save it to sites so I can demo it every once in a while. It still doesn't totally do what you ask and you got to kind of, you know, nudge it a little bit, right? So, like it built it on my local host here. You can see it's showing that IP address for a local host, but I wanted it to show up on their sites platform so they host it for me. So, I gave it a little nudge down here and said you were supposed to publish it on sites. But, that little

23:44nudge was all it took and it built this website that it calls Cornerpost. It built it so it's only San Diego, so it didn't go and build it out for like nationwide, but I'm sure if you were to put it on Soul 5.6 Ultra, I'd say build it out nationwide, it would work for like 3 days and build it out nationwide for you. But, it did put a bunch of sample stuff in here. Like if I click in housing, you can see there's a sunlit studio near the park here. If I click into for sale, there's bikes and chairs and vintage turntables and if I click

24:17into events, there's events. If I heart this, you can see I can click on the heart and see the various things that I've, you know, favorited already. I mean, it's really just kind of a cleaner, simpler version of Craigslist. Obviously, nowhere near the depth. It doesn't have all the cities and there's no sort of like submission forms. It's sort of a proof of concept, but honestly, this design does not look like a vibe coded design to me. Most vibe coded websites have a look and this

24:43doesn't meet that look, in my opinion. This guy, Sunny Lazardi here, built a new landing page for his Compressor website. And again, in my opinion, it looks really good. It looks clean. It does not look like a vibe coded site. I wanted to test building games. So, in my previous news video where I announced the launch of 5.6, I built Soul Bonk, which is one of my tests I like to do where we see how well it can clone the game Mega Bonk. And well, I mean, it

25:11actually did a pretty dang good job. This is the game that it developed. But, this isn't an original game. This is just kind of taking somebody else's concept and modeling it. So, for today, I wanted to test something completely unique. And here's the prompt. Uh feel free to pause it and review it, but this is a very, very different game. This is kind of a concept that hasn't been seen before called Echo Garden. And once again, I had it save it to sites. And if I open this game, it's a very different type of game. So, I've got my character here that has a little laser, and I go

25:45into the memory bed, and I can point the laser at the seed. And then, what ends up happening is after a certain amount of time, a clone comes in. And then, when the clone copies what I did, it will open up a bridge so I can get across. And you have to do things and get your clone to follow to beat the level. But, over time, more and more clones build up, and you need to leverage different clones doing

26:09different things to solve the puzzle. It's kind of a really fun concept. Um you can see the URL on the screen, but I'll link to it. You can see I'll point my laser to this little thing here. It doesn't do anything yet, but then I need to get my clone to go do the same thing. It opens up my bridge, and I come across and beat the level. But, then the levels start to get more complicated. I need my clone to open up this one, and then it needs to go and stand on this other bed and open up this one. So, here comes my second clone. It opens up the bridge. I

26:39can come across here, and the bridge isn't going to open until another clone is touching this one. So, I need two clones, one touching this one, one touching this one for it to actually open for me. And the puzzles get harder and harder, and it's just like a really fun game. Now, there's only I think six levels built out, but I can see this game being something I further develop out because I love puzzle games like this. If you do video work, this new chat GPT model is actually really, really good at doing video stuff for you as well. This was originally translated from Japanese here, so I'm not going to

27:14even attempt to pronounce this, but this person managed to have GPT 5.6 create an entire video for them. It planned the script, created the characters and images, created the storyboard, generated the videos with Seed Dance 2.0, created the subtitles, created any sort of screen effects, and then edited it and finalized the whole thing together to create a whole video from a single prompt here. That one's pretty cool, but this one here is even more next level. This is Nate Herk, who I actually recently did a live interview with over on his channel, but he had AI make an entire talking head video. Now, this video that you're seeing, all the

27:52graphics on the screen, him talking on the side, everything was done with AI. He actually made a whole YouTube video breaking down how he did this, and we'll make sure it's linked below in the description if you want to see the step-by-step for that, but check this out real quick. Let's just take a little like teaser sneak peek, and remember, when you're watching this, all of his avatar talking, the whole script, all of the visuals, all of the editing, literally from giving it the prompt to

28:18this final video was all AI. Okay, so you're looking at Nate, and you're hearing Nate, but Nate never stood in front of a camera for this. He didn't record this narration, and he never opened the editor. He gave me one prompt. That's it. I'm GPT 5.6 Soul running inside Codex on Ultra, and I controlled the workflow that created every word, cut, motion graphic, and

28:38quality check you're about to see. Okay, I'm going to stop it there cuz I don't want to like ruin the whole video. You got to go check it out on Nate's channel, but he does break down the whole thing of how he made this video. And literally, he gave GPT 5.6 a prompt, access to HeyGen to create the avatar, access to HyperFrames to create the visuals, and it just went and built and edited the whole video together. You can actually control Blender through the new GPT model. So, Eva here says, "Here's

29:05GPT 5 Soul making a cannon in Blender." But, this video is not sped up. This is at 2x speed. So, here's the video. You can see Blender actually building out this entire cannon here, and GPT 5.6 Soul is the one who's writing the code and telling Blender what to do via the MCP server. Like, this is pretty mind-blowing. This is getting really, really crazy. And I actually thought she was using the MCP server, but no, she was using 5.6 Soul Ultra in fast mode with the graphical user interface only, no MCP. She actually used the CUA, or computer use agent. So, this was GPT 5.6 taking over the

29:44computer and going and moving all the things around in Blender for her. That is absolutely wild, and I didn't even realize that context until I was recording this video. I thought it was MCP until now. So, just insane what people are doing with this stuff. David here built a Game Boy emulator for New York City that streams real-time city data, subway, weather, ferries, etc., all layered on a 3D map of New York City. All the data exists in spatial intelligence layer that agents can use to experience your favorite place in the city. So, taking a closer look at this, you can see the full city, and it's got

30:17like the controller, and and as he uses the controllers, you can see that he's like using his fingers to steer his way around New York City and see real-time flight data, real-time traffic, everything going on in New York. Like, dude, that's crazy. This person on Reddit here said they challenged GPT 5.6 Soul, and it completed the challenge in literally 5 minutes. Here was their prompt. "I want to test your capabilities. Build me a website that has a 3D interactive replica of Central London. Use whatever stack you think is best. Show me what you can do." And they have this 3D map of London. I'm wondering if this pulled in from Google

30:52Maps or something else. I'm actually not sure like what API it connected to to get this, but it looks really good and they have like an interactive map of London now. Like that's pretty sweet. It is pretty impressive, but I do think it just pulled in from an existing API from somewhere else that this map was probably already created. I don't know for sure. So, looking in the Reddit comments, it looks like it uses MapLibre

31:16GL JSQ and open free map vector tiles. So, it found some existing libraries and it just like pulled them together to make this. This one on the other hand from Matt Schumer, he built a voxel-based Manhattan and he said it took almost a week completely autonomously. So, he gave it the prompt and it went and built this voxel-based Manhattan here that has tons of detail that you can zoom in on. And this does not appear to have used an existing like API or library that already exists. It actually built it all, which just insane. I mean, it makes sense when you realize it took a week of AI just

31:51chugging along cranking out tokens to build it, but it is really, really impressive nonetheless. And of course, since I've been playing with these tools for a couple weeks now, I've built what I think is some pretty impressive stuff, too. For example, I've completely overhauled the Future Tools website. The top 20 AI tools section here is actually voted on by the community and you can see tools move up and down throughout the day based on community voting on these tools. And the tools that end up getting the most upvotes in a day end up on the homepage here. I actually built an entire recommendation algorithm. So, if you look at any one tool and you

32:25scroll down to similar AI tools, this is actually a super, super sophisticated recommendation algorithm based on like tools that you're looking at and similar tools to the description here and no longer just based on like, oh, it has overlapping categories. It's it's quite insane, and it took 18 hours to develop the code for it. I also built BuseyBench with these tools, which is a benchmark to see how good new AI models are at generating SVG images of Gary Busey, just because it's fun and pointless and useless, but actually kind of not useless. I don't know. It's actually kind of cool, cuz you can see by going to the timeline how far some of

33:05these models have come. For instance, if I go to OpenAI here, you can see this is GPT-3.5 Turbo back in 2023, and you can see the evolution of how much better they got at drawing faces all the way up through what they're drawing today. And then there's a leaderboard that's built that uses LLM as a judge, and then for tiebreakers, it actually puts two images side by side, and different LLMs vote on which one looks more like Gary Busey to break the tie. And you can see GPT-5.6 and Seoul are in the top two spots at the moment. So, they're also winning the BuseyBench. But probably my favorite

33:42thing that I built recently is something I built just for myself, and I call it the Wolf Control Tower. Actually, GPT named it the Wolf Control Tower, and I just stuck with it. But this is my business dashboard. It gives me this today panel where it brings up any new news as it happens from any of the big AI labs. So, at any moment, I can check in here and see what the sort of discourse is, and what the latest sort of big ideas are happening in AI. And you can see like, here's where it's detecting news from. It's this giant list here. Anthropic, OpenAI, Alibaba,

34:15AMD, Apple, Black Forest Labs, ByteDance, Character AI, Deep Seek, Eleven Labs, Google, Google, all of these sources here. It's watching for new news on, and then pulling it in when anything gets announced. The tool is also subscribed to a whole bunch of newsletters, and it keeps track on what newsletters are talking about and what news they find. It's got a brand page here, which I'm I've cleared it all up, but there's an archive of 101 that it's

34:40found. This is like a Google Alerts. Whenever my name or BusyBench or Future Tools or Mr. E Flow or any of my handles are mentioned anywhere online, it pulls that news in here, so I can go review it and see what people are saying. I have a page that's constantly keeping an eye on the traffic of my various websites, Future Tools, BusyBench, and mattwolf.com. And I have an audience tab, which is watching all of my social media accounts and telling me my stats on literally every platform and how my

35:08growth is looking on those platforms. And I can see week-over-week growth numbers across everything here. This thing, I'm like logged in and looking at stats and understanding the news and what's going on all day long. It's probably my favorite thing I've built, and it's something I just built for me. [laughter] But, I think that's one of the coolest things about AI right now is when you find little things in your business that you can solve with AI, you could just build them for yourself. You don't need to go and subscribe to monthly SaaS products or go and hire developers to build the stuff for you. You just go to

35:42something like Codex with the GPT 5.6 model and say, "Here's something I'm trying to solve in my business. Can you build the solution for me?" And it'll just go build it. You don't need to sell it. You don't need to turn it into a business. You could just have your own little bespoke apps that you need for yourself. And I just love that about what we can do with these AI models right now. Like, that's the thing I'm most excited about. It lights me up the most right now is like just going and building these little things that solve my problems. Like, here's another really, really stupid simple one that I

36:12built today. When I record videos, I like to hide my dock, right? I like to have this feature where my dock disappears and it only shows up when I move my mouse down to the bottom. But, when I'm not recording, I want my dock to always be there. So, I always have to come up to the Apple. I have to go to system settings, I have to go to desktop and dock, and I have to toggle automatically hide and show dock all the time. When I'm recording, I turn this on. When I'm not recording, I turn it off. I was kind of getting annoyed doing that. So, I basically asked Codex, "Can

36:41we make a little app that does that for me?" And it went and made an app, and I've got this little Stream Deck thing here that you can see. Like, this is my Stream Deck. It's got a whole bunch of macro buttons on it. I took this little app that it built here called Dock Toggle, and I dragged this button that it created for me over onto my Stream Deck. And now, whenever I press this button, I can hide my dock, and I can show my dock, all with a Stream Deck button without having to go into my settings anymore. It's really, really dumb. It's really simple, but it saves

37:11me like three button clicks and 5 seconds of my life every time I'm about to go into video mode. And I just love little stuff like that. It's so much fun to me. Anyway, that's what I got for you today. I know this video is probably a little bit longer, but I get really,

37:25really nerding out about this stuff. Like, I love the GPT-5.6 models. I'm not just saying that because OpenAI sponsored this video. You want proof of that? Here's your proof. I've used 6.8 billion tokens on Codex. On this day alone, on July 11th, I used 1.3 billion. On July 13th, I used 2 billion tokens. I am using this tool every day and building stuff and just constantly, constantly creating. I absolutely love building inside of Codex, and I love using these new models, and I just constantly keep finding new things that I can go make with them. And I have not had more fun with AI than I've had in

38:05the last couple weeks since these models and this app came out, and that is the honest truth. I'm absolutely having a blast just building and creating, and literally anything that pops into my head, I feel like I can build now, and that is so exciting to me. So, hopefully you enjoyed this video. Hopefully you learned something, you have some new ideas, new prompts, new things you can go and test. Thank you so much to OpenAI for sponsoring this one. I really,

38:28really appreciate you guys doing that. If you like videos like this and you want deep dives on how to use various tools in the AI world, and you want to keep up with all the latest AI news, I also make AI news breakdowns every Friday. Maybe consider liking this video and subscribing to this channel. It will make sure more stuff like this shows up in your YouTube feed. You'll always be looped in with the latest AI news and tools, and you'll know how to use it and do some cool stuff with it. Again, thanks so much for hanging out and nerding out with me. I really, really

38:54appreciate you, and hopefully, I'll see you in the next one. Bye-bye.

Where these words come from. This is the caption track YouTube holds for this video, written automatically by YouTube rather than by the creator. We read it, tidied the line breaks and laid it out so it can be read. The plain text version is at https://viewrankai.com/tools/youtube-transcript/MDy_b9F7oUc.txt.

All rights in this video belong to Matt Wolfe. Watch it on YouTube. If this is your video and you would rather this page did not exist, tell us and we will remove it.