@willfrancis24

What does Pope Leo XIV's tweet say about AI-generated art, from apps like Midjourney, Gemini Nano Banana and ChatGPT Images? Can it ever really be Art? What about other forms of mass reproduction like photography? There were similar concerns about that in the 19th century, with fears it might threaten the craft of painters and sculptors. But it didn't, and photography became its own distinct art-form. Will AI-generated Art become a respected category of creative work? #ai #artificialintelligence #aiart #genai #midjourney #chatgptimage

Pope's tweet about AI went really viral, but what does it mean? It's quite wordy. So let's unpack it quickly. So Pope Leo posted on X that there's an ontological difference even before an aesthetic one between art and what a machine generates through statistical calculation based on millions of images made by others. Ontological is a philosophical word. It means of or related to the phenomenon of being or existence. So he's saying an AI image can never be art, even if it looks identical to a real painting. He's saying the art is more than just what you see. A piece of art is an object that came from someone, maybe has a history, it exists and it has other qualities beyond just the pixels and the brush strokes that you see. Now that's always been true. If someone graffiti is on the side of your house, you call the police, you paint over it. But if Banksy graffiti is on the side of your house, maybe it'll increase in value and you won't paint over it, right? So art has always had varying value based on things beyond what you can see. Now the panic around AI art honestly to me just feels a lot like concerns people had in the past about mass reproduction in its various forms. In 1859, the poet Baudelaire dismissed photography as a refuge for failed painters. Harsh, but you can imagine the outrage at the time. Suddenly there's this machine. Any Idiot can press a button, boom, you've got a really detailed image. Where is only a few years ago someone have to train for years and then spend days or weeks producing each individual piece of art. Now you can just flick a switch. But of course, what happened is photography became its own distinct art form. It didn't destroy painting. I'd say there's more painters now than there's ever been. And within photography, you've got all these unique techniques and challenges and ultimately artistic decisions that the photographer can make. And we, the audience, love to see the fruits of those decisions in quality creative work. Right now, I kind of think AI is going that way. I do personally find it interesting looking at AI art. I mean, I'm not sure I'd print it out and put it on my walls yet, but I'm interested in how people created it and what decisions they made along the way. And as always with this stuff, you will find that the very best work always takes more effort, ingenuity and creativity. And particularly as we get used to what these tools are capable of and what it looks like when people use them in a very low effort way, well, when we see what really creative people do with them, it's impressive. And like photography, I don't think it threatens other art forms. I think if anything, it highlights their value.

Tesla just released a flying car! Here's what you need to know about last night's launch event at SpaceX in Texas, USA. So, the new Tesla Roadster, the latest version of their sports car, can only fly for very short periods of time, thanks to cold gas thrusters, which is an example of taking what they've learned at SpaceX and applying it at Tesla. It’s not combustible rocket fuel or anything like that. It uses tanks of compressed gas that is squirted out through nozzles in a controlled way to make the car be able to somersault, hover, or drop from a height and land softly. The rear thrusters also allow it to go from 0 to 60 mph in well under a second. They did it in 0.81 seconds in last night's demo at SpaceX’s test facility in Texas. The car also has a suction system to keep it stuck to the road, even when upside down. I did see a lot of comments like, "The douchiest car just got even douchier," because what kind of person would actually flip their car out of a traffic jam into a different lane? But that's the interesting thing about this launch event: it's kind of like applying the Mr. Beast content strategy to a product launch, where it's just great content. It's impossible not to watch. But that's kind of Tesla's thing. In 2018, they sent a first-generation Roadster into space, and it's currently orbiting the sun. It's about 60 million miles away from Earth. But people are rightly sceptical. The second-generation Roadster released in 2017 never made it to production, so I'd be surprised if this actually does come out in 2028 or ever. But if it does, there's going to be three versions: Performance, which is just the Tesla Roadster sports car and you can reserve one for just $50,000 deposit, the thrust version, which has rear cold gas thrusters and gets that 0 to 60 under a second which you can reserve for just a $250,000 deposit, and then the rocket variant, which has all the thrusters for lift and rotation and a top speed of over 230 mph. They do say that is not approved for public roads, for obvious reasons!! #ev #electriccar #tesla #teslaroadster #elonmusk #ai #technology #tech

Tesla just revealed a flying car and it's breaking the internet today. So, let's have a look at the key facts from last night's bonkers launch event. So, the new Tesla Roadster, which is the latest version of their sports car can only fly for very short periods of time. And it's thanks to these cold gas thrusters, an example actually of taking what they've Learned at Spacex and applying it at Tesla. Now, it's not combustible rocket fuel or anything like that. It uses tanks of compressed gas that's basically squirted out of nozzles in a controlled way to make the car somersault, hover or drop from a height and land softly. The rear thrusters allow it to go from naught to 60 in well under a second. At last night's launch event, they did it in 0.81 seconds. Car also has this suction system that can keep it stuck to the road even when upside down. A common driving problem we've all been dying to solve, right? Um, but I did see a lot of comments on the videos. The douchiest car just got even douchiest because I mean, seriously, what kind of person would actually flip their car out of a traffic jam or something into a different lane? It's insane. But that's the interesting thing about this launch event. To me, it's like applying the Mr Beast content strategy to a product launch where it's just great content, like it's impossible not to watch. And that's Kind of Tesla's thing. I mean, in 2018, they sent a first generation roads to interspace. It's currently orbit in the sun. It's about 60 million miles away from earth at the moment. But people are rightly skeptical. The second generation Roadster released in 2017 never made it to production and people have been waiting ever since. So I'd be surprised if this actually does come out in 2028 as planned or ever. But if it does, there's gonna be three versions. The performance, which is just the Tesla Roadster sports car and you can reserve one for 50 grand deposit. The thrust version, which has the rear cold gas thrusters gets you that 0 to 60 under a second, you can reserve one for 250 grand deposit. And then the rocket variant, which is all the thrusters for lift and rotation. It's got a top speed of over 230 miles an hour. They do say that's not approved for public roads for obvious reasons and you can't even reserve one. It's gonna be invitation only. So it's basically just for Elon Musk's billionaire mates.

Why does AI like ChatGPT and Claude love em dashes? What are they and how should we use them? All your questions answered! From the origins of the em dash on the printing presses, through its use in history to the present day where it's one of three dashes we use in the English language. And yet, it's become a telltale sign that someone used ChatGPT, Claude or Gemini to write. #ai #aiwriting #chatgpt #claude #artificialintelligence #technology #learnontiktok

Why does AI love M dashes? What is an M dash? What are they for? How do you use them? Let me explain everything you need to know about M dashes. So these long dashes in the middle of sentences have become the most obvious sign that someone used AI. And that's because large language models like ChatGPT were trained on vast amounts of human writing, and they Learned how to write properly. But once you've trained a large language model, it's really hard for it to make exceptions. So that's why even if you put don't use M dashes in your instructions, you will still see it use M dashes sometimes. So we're just gonna have to embrace the M dash. And anyway, they're really useful, so I'm gonna tell you how to use them properly in a sec. But firstly, why is it called an M dash? Well, an M is a unit of measurement that we still use today with fonts on websites and documents. But it originates from printing presses and the people who did the type setting on those printing blocks, placing all the little metal stamps that stamped the characters onto a newspaper or book page. Well, sometimes they needed to create spaces on the page, and so they had different sizes of blank spacer blocks. One of them was called an M quadrat. And that one was always square. It was equal width and height. It's actual size varied, but it was determined By the type face that it was within. So whatever height that type face was would also be the width of the M quadrat. And that's why the M dash looks pretty long, because if you turned it around, it would be as tall as the tallest, biggest character in that typeface. So now you know that in typography, 1M is one times the height of that typeface. In previous centuries, EM dashes were used a lot, particularly in the 18th century, to the point where there was actually a backlash against them. this is not the em dash's first Rodeo as a vilified punctuation Mark. In 1795, a language expert called Lindley Murray warned about their misuse by hasty and incoherent writers. But they are useful, and they are different from the other dashes. There are actually three ways to use an M dash. The first is to contain an aside. I left my phone, which was an iPhone 18, in the cafe today. People would often use commas or brackets for that, but the dashes give it more emphasis. In fact, some people do call the M dash an emphasis dash, but that's not actually where the name comes from, as we've just Learned. Secondly, an mdash could tee up an explanation or a punchline to a sentence. So I've forgotten one thing. My phone. Now, a colon would work there fine as well, but the dash just gives you a more abrupt reveal. And then thirdly, It can show speech being cut off abruptly mid sentence. How about we? And someone cuts in. You often see this in scripts or dialogue in novels where someone's unexpectedly interrupted by another person or something happening. So that's how to use an M dash. But it's just one of three dashes that we commonly use in English, because we also have the hyphen and the N dash. A hyphen is the short one that joins words together, like well known. An N dash is a little bit longer, and that often means two. So, like pages 10 to 20. And then an M dash, as we've talked about, that's the longest one. It's for those breaks, interruptions or points of emphasis. And if you want to type 1 on a Mac, it's shift option hyphen. Or on a Windows, you hold down Alt and type 0, 1, 5, 1.

Migoo helps me get real work done at a fraction of the cost of major AI tools. I can turn messy data and notes into polished, client-ready PowerPoint presentations in minutes, with slides costing as little as $0.01 per page. @migooai8 #ai #aitools #tech #workwithmigoo #migoo #migooai

It's the night before a client review. All I've got is a messy spreadsheet of add data and a page of notes. Normally that means a really late night putting it all together for the presentation, but I've started using Magoo. You just give an expert agent a task and it works on it in the background while you get on with other things. So first I dropped in the spreadsheet and notes, gave the data analysis agent the brief, analyzed this month's performance by channel, what works, what didn't, didn't, key takeaways. It analysed the data, pulled out the insights and then passed those to the slides agent and asked it to turn the analysis into a client ready report. The first pass has the right data and charts, but I just wanted to be more direct, like, what channels performed well, what didn't, where could the budget be better allocated? It reworked the deck based on that feedback, so here's the final result. And now I've got cost per lead by channel, performance trends, spend breakdown, clear recommendations, all presented perfectly for the client. So instead of spending hours building the report manually, I can go straight into the review with the analysis and the presentation ready. So if you're spending hours turning data into client reports, Magoo's AI agents could build the whole thing for you really easily. And you can try it out through the link in my bio.

AI writing has a shape! New research analysed thousands of articles written before ChatGPT, or any LLMs. It then compared those with articlea written by 5 different AI models - GPT, Claude, Gemini, DeepSeek and Kimi. And guess what? All the AI-written content has a distinct shape when you map the characteristics of the writing techniques and overall structure, that is detectable almost all of the time even after rewording some of the work. Here's an explanation and what it means for you if you're using AI to write. #ai #aiwriting #chatgpt #claude #aitools #artificialintelligence #tech #technology #learnontiktok

This new research shows that AI written articles have a clear shape that can be detected over 98% of the time, even if you change some of the wording. You might remember this research talked about earlier this year about AI stories having a similar structure. Well, this new study tested the same idea on company blog posts. It looked at thousands of articles published before ChatGPT even existed and compare them with versions written by five AI models. And look at that shape. What does that shape remind you of? So what does this actually mean? Like, what does this shape even represent? Well, every dot here is an article. And the researchers measured 176 features of how those articles were put together. Like how they introduce an idea, explain it, use evidence, bring it to a close. And then they plotted those measurements on a map. And the five AI models all overlap in this big clump, because they're basically making the same structural choices when writing, whereas human articles in blue over here occupy a different, more varied range of structures. Because we're human, we're random and unpredictable. We are not an algorithm. So what are those choices that AI makes? Well, AI tends to give you the main point straight away. It announces everything it's about to cover and then finishes by repeating the point. So it basically tells you what it's gonna say, says it, and then tells you what it said. I mean, that's One reason why it can come across as very waffly. So, like, one example of a structure that it loves is framing things as a battle between one thing and another. Dichotomies, so the old way and the modern way, or online shopping and physical retail. And that's just one example of these, like, go to little frameworks that it builds its writing on. Whereas human articles, as well as being more random and unpredictable, were more likely to refer to specific historical dates, places, periods, people, I think, in general, because that's part of how we understand the world and how we communicate our knowledge. Whereas AI doesn't have that lived experience. It doesn't remember things in the way that we do. It just likes to produce lots of words. And even when the researchers got the AI to reword and remove the telltale writing habits from the articles, the structure and the shape didn't really change. So that tells you that getting AI to write for you and just going in and reword in a few bits doesn't really cut it.

5 ChatGPT prompts for interior design. Change up any room in five different styles with these simple prompts. Whether you prefer the modern mix of Japanese and Scandinavian design that is Japandi, or more of a mid-century look, even an industrial warehouse look, or you could really go for it with a maximalist vibe. Finally, why not try coastal with a chill white, blue, seashells, linen-textures style? Remember, once you've restyled your room, you can swap out any piece of furniture or give ChatGPT product photos of furniture to put in there so you can imagine your perfect room! #chatgpt #chatgpttutorial #openai #AI #aitools #ai #tech #technology #artificialintelligence #learnontiktok

5. ChatGPT prompts to redesign any room, like having your own personal interior designer. So just upload a photo of your room and try any of these. You can grab the prompts by hitting my link in bio and going to Free Resources where everything's free. So first, japandi, this is like a modern mix of Japanese and Scandinavian design. So, like, soft textures, simple furniture, a nice calm feel. Next is mid century modern. Bit more character, but without filling the room with too much stuff. A lot of those nice kind of tones of beige and brown, you know? Then industrial, that converted warehouse look in your own living room. No.4 is maximalist. This is the one to try if you're just bored of being beige. Like, it goes to town on decoration, pictures, colours, patterns and plants. And then coastal. So you've got white, light blues, seashells, linen textures and that kind of thing. Now, these prompts are just a starting point. Once you get something you like, you can get way more specific, like change just the sofa, fireplace, the rug. What I do is when I find product photos of things I like online, I'll put that into the prompt and say, put this chair in

New ChatGPT dots! Major new feature brings an always-on 24/7 AI assistant to your ChatGPT, that you can call or text to manage any of your ChatGPT or Codex chats and tasks. Also, the new Space feature is a collaborative workspace where all the documents, spreadsheets, images, and sites you've created with ChatGPT live are stored. You can work collaboratively with your dot or other people and their dots on these. #chatgpt #chatgpttutorial #openai #AI #aitools #ai #tech #technology #artificialintelligence #learnontiktok

ChatGPT just launched the dots today. It's a massive new feature. I got early access to it, so I'm gonna show you how it works. Your dot is an AI agent that works 24 7 even when the app's closed. And it can manage all of your ChatGPT and Codex chats and tasks. All you need to do is call it or message it and it's kind of like having a concierge that does all the micro management for you. This is being launched with another new feature called space, which is like a collaborative workspace where all your documents, spreadsheets, images, websites, everything that you've created with ChatGPT lives. And it works a bit like Google Docs or notion where you can work with other people or their dots or your dot in real time. Anyway, there's a lot going on there and I'll do a separate video about that. So just update the app on desktop or mobile and you'll see your dot at the top of the sidebar menu. And the first thing you want to do is give it a name and pick a fun Avatar. Now your dot is powered by GPT6 Astra, the most advanced version of GPT. It has its own computer in the cloud where it can do all kinds of work and use apps. But. But you really wanna give it access to your actual computer so it can use apps and your browser, which is pretty useful cause then it's Logged in as you, and it can actually do real work for you. There are three things I personally like about this, and a couple of things that I think they could really improve. Firstly, I really like it being proactive. So my dot messages me throughout the day to remind me of deadlines or things I'm supposed to have sent to people, because it can see my email, calendar, meeting notes, literally all the apps I've connected to ChatGPT. So it's really useful as, like, this kind of virtual assistant. I like how it asks you for feedback as well. So it asked me if it was being too chatty, and actually I said, no, be more chatty. And it was. And the things it says are useful. And the other thing I really like is, like, the killer feature, really, where it can start new tasks and chats on your behalf, whether that's working on its own computer, in the cloud, or as long as you've got your computer on somewhere, doing stuff on your computer, it can orchestrate potentially any kind of work and also keep an eye on that work wherever you are. So maybe I was organising a birthday party, right? I could ask it to do that. It could design invitations, order everything I need on Amazon, create a grocery order, send out the invite emails, but also keep an eye on my inbox and keep me updated on who's coming, who's not coming. And keep a record of that. A couple of things that frustrate me about it are, firstly, it's a little bit locked away. Like, I wish it was either a separate app or in some persistent place where you could always see it without me having to leave the task I'm doing to talk to it. Now you can connect Slack to it, so I suppose I could do that there. And the thing I'm not sure about, like a lot of these things is if you're an advanced user, you've probably worked out how to boss ChatGPT around in very advanced ways anyway. But then with mainstream users, I'm not sure if they all get why this needs to exist when they already chat to ChatGPT. So it'll be interesting to see what the take up is on it, to be honest. Basically the first version of opencl for the masses. Because remember the open AI hired the guy who created opencl early this year, so we knew that that was something they were working on. And because Meta Muse has had pretty good reviews in its first few weeks, it feels like now everybody's rushing to launch always on AI assistants that are always represented by cute little blobs for some reason. And this is open AI's response to that.

5 ChatGPT codes for marketing images - to turn one product photo into a suite of photography for your website, social media, ads and more. You can grab the prompt to paste into your custom instructions in my Free Resources, where there is a full guide and an easy-to-copy prompt for your ChatGPT custom instructions which you'll find in Settings > Personalization. #chatgpt #chatgpttutorial #openai #AI #aitools #ai #tech #technology #artificialintelligence #learnontiktok

There are five ChatGPT codes that turn one product into a full set of marketing images. So take this coffee cup. Right, I've got one photo, but I want five different shots. So first is White BG that puts your products on a clean white background with a soft shadow. That's your website product shot. Sorted. Next is Flat Lay that creates a shot from above with matching props around your product. So here we've got coffee beans and a notebook, and there's space at the top for your own text, so perfect for a social post. Then in use that shows someone actually using your product. So instead of a cold product shot, now I've got someone actually having a coffee in a cafe with my cup. That's the lifestyle shot. Sorted. Then Billboard. Puts the product on a giant roadside billboard. If you pitch in a campaign, you can show people exactly what you've got in mind. And then seasonal. So type seasonal Christmas and you've got fairy lights. It's festive styling. You can change that to Easter, summer, Halloween, whatever, and it will always work with your product. So to set this up, just go to settings, personalization, and in custom instructions, paste in this prompt and you can grab that in free resources. There's a whole guide there in my link in bio. And basically every time you put in a photo and use one of these commands, you'll get exactly the same results as I've Just shown you every single time.

One connector that lets Claude generate its own images, videos, voiceovers, sound design, and music. This is ElevenLabs' MCP for Claude and it's honestly a game-changer for creators. Get 10,000 free credits via the link in bio. #ad #ElevenLabs #AI #aitools #claude #creative #ElevenCreativePartner

Claude just built this website including those gorgeous videos and images just from one prompt. And because I've plugged in the 11 Labs MCP, this allows Claude to create its own images, videos, voiceovers, sound design, pretty much any type of media that you need. And it was literally a one line prompt in Claude, a dynamic video LED website which uses cinematic video to show the coffee making process in our artisan cafe in stunning detail. Pasted a bit of info about the cafe in and whilst claw designs a gorgeous site, it sends prompts to 11 creative to get all those images and videos that it needs before building the website around them. To set this up yourself, all you need to do is in Claude, go to customize, add the official elevenlabs connector and just start building. Hit the link in my bio to sign up for a free elevenlabs account.

7 habits to get more from Claude and ChatGPT: Correct mistakes. Share your rewritten version and explain what you changed so the AI understands your preferences. Edit your prompts. Improve the original prompt instead of piling on follow-up instructions. Keep the conversation focused. Use projects. Organise related chats, instructions and reference files so your AI has the right context. Review AI memory. Regularly check saved memories and remove anything outdated or incorrect. Give examples. Share writing, designs or previous work that show the result you want. Examples make your instructions clearer. Let AI interview you. Add: “Ask me whatever questions you need to get the job done.” Give it the detail it needs for more relevant results. Save repeat workflows. Turn recurring jobs into reusable skills with instructions and reference files. Use scheduled tasks for work that needs to happen regularly. #chatgpt #chatgpttutorial #openai #claude #anthropic #claudecowork #ai #aitools #tech #technology #artificialintelligence #learnontiktok

7 habits that will make you a Claude or ChatGPT Pro. And every one of these works on both apps and on the free plans, and most people just don't do them. So. The first habit is correct. Don't just rewrite so when your AI gets something wrong. Most people just copy and paste it somewhere else, rewrite it and get on with it. But what you should do is go back and give it your rewrite and also give it some feedback. That will go into its memory and it will improve everything it does for you over time. The second one is kind of related. Be honest, how often do you go back and edit your prompts? I bet you never do that. But it's so much more efficient than constantly replying with more prompts. Trying to steer and correct the AI which has to pump out a new version of what it's doing for you every time. That just blokes the context window really quickly and so the outputs turn sloppy fast. The third habit is to just get organised. So start using projects now. I would strongly recommend you get the desktop app, either Claude or ChatGPT, and now your projects are just folders on your computer that keep all the instructions and files you work on nice and organized. Now that requires a little bit of organization from you. You need to start getting good at just organizing things into folders. Honestly that takes you Into a whole other realm of being an AI pro. No. 4 is be in the habit of checking in on memories in your settings. Do it every week or two so you can delete anything that's not correct anymore or edit or add memories there. It's just like a little bonsai tree. You just gotta keep regularly pruning and that will make everything more relevant and more useful. No. 5 is examples beat instruction. You could describe your brand, tone of voice forever, but just giving it a few real examples gets better results quicker. This goes for design, images, technical stuff, documents, anything. Just give examples from things that you like out there or previous work you've done that you want it to mimic. Because AI is a much better mimic than it is a follower of instructions. And No. 6 the reverse interaction prompt quite simply make your AI interview you. So tag onto your prompt something like ask me whatever questions you need to get the job done. And this is a great way of pulling good information out of yourself into the AI and the results will always be so much better and less generic. And No. 7 turn repeat jobs into skills and scheduled tasks. It's probably the most valuable habit here. So now that ChatGPT is ditching custom GPT's, we've got skills in both ChatGPT and Claude? Skills are just a saved set of instructions, potentially with reference files like previous examples, screenshots, Font files, etcetera. Now, to create a skill is really easy. All you do is chat. So once you've done a job and you think you might want to do that, you know, quite a lot in the future, you just say, let's turn this into a skill so we can do it this way every time, and it will do that. Skills can also use connected apps, of course, so a skill could involve dipping into notion or Gmail or whatever. So, you know, a skill can be quite a complex piece of work. But once you've done that, you can then follow up in chat and turn it into a scheduled task by just saying, do this every Tuesday morning.

Canva AI Vision in London - here’s what I learned about how AI is driving creativity and investment in the UK and beyond. Thanks @canva for inviting me! #ai #canva #design #aitools #tech #ad

To the Canva AI Vision conference in London this week, and I learnt a few things about where AI is going and how it's colliding with the creative industries. So firstly, I heard that creatives are very mixed on AI, but we are starting to understand that it's really just a tool for better workflows. You know, it's not about lazy AI generated slop images or placing creativity or anything like that. And that was the theme of the day with Canberra embracing pretty tough questions about staying human, staying creative. And my vibe check amongst designers in the room was that there is still fear and uncertainty, but AI is settling in as part of the workflow. It's just another tool. And events like this allow them to see how really creative people are using AI. Not to be less creative, but to be more playful, more experimental. I mean, that's what I'm doing with it. Someone in the AI startup space said that the ambiguity is where the money is, meaning there's still a lot that hasn't been solved in the AI space. There's lots of room for creative opportunity, right. Building apps, startups, creative tools or educational resources like what I do. I mean, I thrive there in that zone of uncertainty. And that is an inherently creative opportunity that any of you can take, honestly. And something that was said about London that I'd not really thought about, um, so much was that it's unique, right. Because it's a leading AI center outside of the US, but it's also a leading culture, media and arts hub globally. So you've got this melting pot of tech and art people, and, yeah, you've got some people doing some really creative stuff with technology here, and we saw some of that on stage.

4 things that immediately make ChatGPT more useful for getting things done: Get the ChatGPT desktop app. This allows the AI to work in the files and folders on your computer and also take control of your computer, including the apps and the browser. Now you can do very complex work with it. Start using Work mode, or, even better, Codex, which looks technical but is the same as ChatGPT, but the most capable version of the app. Connect your apps, such as email, calendar, meeting notes, Google Drive, etc. This gives it the ability to work in your systems. Skills. This is how you train your AI assistant, and you can create new skills or import them from other AI tools like Claude. #chatgpt #openai #chatgpttutorial #ai #llm #tech #technology #learnontiktok #aitools #artificialintelligence

Four things that immediately make ChatGPT way more useful for doing actual work. These are immediate upgrades and most people I talk to haven't done them. So the first one is get the latest desktop app. Just install it from ChatGPT's website. Yeah, this gives you access to not just regular chat but work mode and Codex and you will get so much more done. And yes, you can do this on a free plan. The killer feature with these two modes in the app is it can now work directly on the files and folders on your computer. You give it access to a folder that's a project and it can then read, edit and organise files inside there and create new ones. We're talking about files of pretty much any type here. You can now also enable computer use and browser control and this lets chatgbt operate apps and websites on your behalf. You have now unlocked the ability to get really technical or complex or long boring pieces of work done for you. It is like having a colleague and yes, it's safe. You always get asked for approval before it does stuff. The second thing is to start using work or Codex as a habit. I would strongly recommend you just go straight to Codex. Okay, because yes, it's got a few features for people who write code but if you're non technical like me, just ignore those and know that you are now using the Most advanced version of this technology and on the face of it it's very familiar it looks just like ChatGPT you can do basic things here, you can do really advanced things here all in one place so you kind of got an AI assistant now so the third thing to do is to give it access to your stuff and that's where apps come in so start with the services where your knowledge lives email, calendar, meeting notes, notion, Google Drive, wherever that is for you and now ChatGPT and gather information across those sources and do some really useful work for you whether that's basic stuff like managing your email or more complex stuff like running your entire finance department it can do that the fourth thing is skills and this is how you train ChatGPT so skills are just instructions and they were made popular in Claude but now they've become the open standard across all AI apps and in fact chatgbt has just said that they're ditching custom GPT's and it's all just gonna be skills now and skills are very easy to create if you do a task in chatgbt at any point you can just go turn this into a skill so we do it like this every time or you can describe a piece of work and give examples or you can even record yourself doing something it will turn that into a repeatable skill and do that job the same every single time And yes, you can go and download your skills from Claude if you got some there, just drop them into the chat here in chatgb and it will install them here. Skills is still rolling out. So if it's not in your chatgb settings, switch to Codex, go back into settings and you'll see skills there in the plugin section.

Is ChatGPT Images 2.5 any good? Or just another AI slop image generator? Let's take a look at whether it can overcome the typical problems of previous generations of AI image tools. I'll show you what features I think are most useful. So this came out on Tuesday. It's on the free version too, and OpenAI say it's sharper, faster and better at editing. Of course, the first test is a poster for the village fair. Just imagine if it did this really well, because we've all seen too many of these AI slop posters. What happened to good old Comic Sans in a Word document? This was with the previous version of ChatGPT. And here's the new version. I do think this will still make people angry. It's nicer. It's more subtle. The font is very recognisably AI. To be fair, in the new templates here is a poster template, and here are the results you get from that, with a reference image of one of those cool posters for an art exhibition. When we use that, it looks really cool. Next test is that weird yellow tint they add to make things look more organic, it’s become a real AI tell. In fact, someone built unyellowgpt.com just to remove that grading from AI images. And yes, we can see that it has stopped doing that, which is great. The third test is detailed scenes. Here's a Where's Wally (Waldo) puzzle, and yes, Wally is in here once. I've had a good look around it, and everything seems good, which is quite impressive. The third test is detailed scenes. Here's a complex, asymmetric view of a city with life on every corner and things happening on every block. It's 20 by 20 blocks in size. Wow, that is seriously detailed. The fourth test is complex text combined with real-world knowledge. This is a tough test, but also a common use case. If we want it to make infographics, posters, etc., we want that information to be correct. And it wasn't great at that, unlike Gemini. This infographic of London's tallest buildings is a lot better now, and the numbers are all correct. The last test is: can it break away from the conventions it's trained on? Can it show a clock that doesn't read 10:10? Can it show a wine glass that isn't perfectly half full? It used to struggle with that, however much you asked it, but it seems to have overcome that. That's great because it means that it's less influenced by the averageness of its training data and more influenced by what you want. Great. For me, the most useful upgrades are: Designs with text, like the header for my weekly newsletter, which you can subscribe to, see my link in bio. I've noticed it's much better when it uses my actual fonts rather than its own telltale AI typography. Consistency of character or object. If you give it a photo of a person, maybe you, and ask it to put you in 20 different outfits, it can do that, but it will still look exactly like you. You might put a product that you sell into 5 different settings, and it's great at not changing the product. There are very functional uses for it. This sketch feature: you get a little canvas to draw in. This isn't for bringing doodles to life, although your kids might like that. This is for influencing the layout of what it produces, which is especially good for design tasks. It is very interesting that OpenAI have just blocked design companies like Adobe from advertising in ChatGPT. I think they are competing with those kinds of companies now with this product. #chatgpt #openai #chatgpttutorial #ai #llm #tech #technology #learnontiktok #aitools #artificialintelligence

Let's run a slop test on the new chatgbt image generator images 2.5. But I'll also show you what it's actually useful for. So this came out on Tuesday. It is available on the free version. And open AI say it's sharper, faster, and better at editing. So of course, the first test is a poster for the village fair. Just imagine if it did this really well because we've all seen way too many of these AI slot posters. I mean, what happened to good old Comic Sans in a word document? Anyway, this was what the previous version of chatgbt images could do. Here's what the new version can do. And I do still think this will make people angry, but it is nicer, it's more subtle. The font is just very recognisably AI. The whole thing is really, to be fair, there is a new templates feature. There is a poster template. And here are the results you get from that, which I think you'll agree are much better. So the next test is that weird yellow tint they put on images, I think to make things just look more organic or something. But it's become a real AI tell. In fact, someone even built on yellow gpt.com just to remove that grading from AI images. And yes, we can see that it seems to have stopped doing that. The third test is very detailed scenes. So here's a where's Wally Waldo puzzle. And yeah, wally is In here once and about a good look around and everything seems good, which is pretty impressive. The fourth test is complex text combined with real world knowledge. This is a tough test, but also a common use case, right? If we want infographics and posters and Powerpoint decks and things, we want that information to be correct. It wasn't very good at that, unlike Gemini, but this infographic of London's tallest buildings is a lot better now than the previous version. And this time all the numbers are actually correct. The last test is can it break away from the conventions that it was trained on? Can it show a clock that doesn't read 10:10? Can it show a wine glass that isn't perfectly half full? It used to really struggle with that, however much you asked it. It seems to have overcome that and that's great because it means that it's less influenced by the averageness of its training data and more influenced by what you want. For me, the most useful upgrades, well, there's three. Firstly is designs with text, particularly like the header for my newsletter, which you can subscribe to, link in my bio. I've noticed it's much better at producing those with my actual fonts rather than reverting to its own kind of crappy typography styles. Number two is consistency of character or object. So you can give it a photo of a person, maybe you, and ask it to put you in 20 Different outfits, it can do that, but it will still look exactly like you. Or you could put a product that you sell into five different settings and it's great at not changing the product. And the third thing is this sketch feature. So if you want, there's a little canvas that you can draw in. This isn't about bringing doodles to life, although you kids will love that. This is really just about influencing the layout of designs. And actually, I'm starting to find that quite useful. And it's pretty interesting, I think, that in this very same week, Open AI have just blocked design companies like Adobe from advertising in Chat GPT. And I think that tells us that they want people using Chat GPT as a design tool.

Is GPT-6 Astra really AGI? Has artificial intelligence beaten humans on the true tests of general intelligence? What even is AGI? We do have a pretty robust definition of AGI. While every model gets progressively better scores, there are only certain parts of the test that AI excels at, whereas AI gets incredibly low scores on much of the test. So the new ChatGPT, whilst very powerful, compared with these tests of true general intelligence, doesn't look anywhere near achieving that, despite OpenAI's claims that we've entered the era of AGI. #chatgpt #openai #ai #chatgpttutorial #llm #tech #technology #learnontiktok #aitools #artificialintelligence

Everyone's saying the new ChatGPT is AGI, but what does that mean? What's the actual definition of artificial general intelligence? Well, there is one, and it's been put together by a mix of skeptics, pro AI people, academics, researchers. So let's have a look. It's basically one sentence, right? AGI is an AI that can match or exceed the cognitive versatility and proficiency of a well educated adult. It's not super intelligence, it's general intelligence. So last Thursday, Open AI launched a GPT6 Astra, and their president, Greg Brockman ended the press conference by saying, welcome to the AGI era. It's a big statement. Then on Sunday, Jensen Huang from Nvidia use chips it was trained on, posted AGI has arrived. Now, I've had early access to this model, so I've been using it for a while. And yeah, there have definitely been a few moments where I've had to like, really stop and think about the enormity of what it can do. It's clearly incredibly powerful. And I say this with no interest in open AI or in hyping it up. I'm just saying I've felt its power. But then again, it's just an LM, it's just a word spitting machine. So, you know, I'm really working this out like everyone else. But that's where this paper is quite helpful, because Last October, 33 researchers published a definition of AGI. And the authors include some big hitters like Joshua Benjio, one of the so called godfathers of AI. Eric Schmidt, the ex CEO of Google, who's very pro AI. Gary Marcus, one of the most famous AI skeptics out there. And they basically agreed on what this AGI test would look like. They took the way that psychologists have tested human intelligence for a very long time and just broke that down into 10 abilities. Each one's worth 10% of the test score. So it tests on general knowledge, reading, writing, math, reasoning, memory, vision, hearing speed. And if an AI scores 100%, we're basically saying it's AGI. Now, GPT4 scored 27%, GPT5 scored 57%. Now, of course, it was jagged. GPT5 got full marks on maths and writing, as you'd expect. On long term memory, zero. It just can't learn from experience. Like, I know chat. GPT has a memory feature. It's not really memory, it's just rereading its notes every time. And this is why benchmarks are really flawed. Like, of course, the AI companies pick things like maths and coding where they can smash the scores. It looks incredible, and they are genuinely starting to outpace humans on those things. But maths and code, they're what we build computers for. That's what they do. And this, more general intelligence, it just seems a long way off. Right, so is GPT6 AGI? No one's published results of putting Astra through these 10 tests yet. I'm sure they will. Anyone can. You can just go to AGI definition dot AI. But honestly, just reading through those tests, if This is what we're saying AGI, is we're not there yet.

The new ChatGPT Astra model is a much bigger upgrade than just getting better answers to questions. Its combination of AI computer use, Codex, autonomous task completion and the ability to work across apps means ChatGPT can now take on jobs that previously needed a human sitting at a computer. Here are 6 genuinely useful things you can do with ChatGPT Astra: Do your weekly food shop with AI Ask ChatGPT for easy midweek meal ideas, choose the ones you want, then let Astra use your computer to shop for the ingredients from your usual supermarket. It can navigate the website, add everything to your basket and get it ready for you to approve. Build your own apps and tools with Codex AI coding and “vibe coding” are becoming much more accessible to people who can't code. Using Codex in the ChatGPT desktop app, you can describe the app, website or internal tool you want and have Astra build, test and improve it. I'm experimenting with replacing several pieces of software I currently pay for, including meeting notes, transcription, notes and social media tools. Turn ideas into interactive 3D models Give Astra information about a house or building and it can create an interactive 3D environment you can explore. You can also use it for more creative projects, like turning children's drawings and game ideas into simple playable 3D experiences. Edit videos using AI With tools and skills such as Hyperframes, ChatGPT Astra can help build a video editing workflow that cuts footage, creates motion graphics, adds captions and handles much more of the production process. This could be particularly useful for marketers, content creators and businesses trying to produce more video content. Automate large amounts of life admin This is where AI computer use gets interesting. Give Astra a collection of photos of things you want to sell and it could potentially help create listings, organise the information and work through the repetitive admin. The same idea applies to organising files, researching purchases, handling financial admin and other jobs that normally involve lots of clicking between websites and apps. Build or rebuild a website You can give Astra an existing website, your company information and the content you want to use, then ask it to create a new site. Combine it with free hosting such as Cloudflare and AI can increasingly handle the coding, testing and deployment as well as the initial design. #chatgpt #openai #ai #chatgpttutorial #llm #tech #technology #learnontiktok #aitools #artificialintelligence

There are 6 genuinely useful things to use the new ChatGPT model Astra for, because it's capable of stuff that AI hasn't been able to do very well until now. If you want to grab these prompts, hit the link in my bio, which is where I keep all my free resources. And the first one is do your weekly food shop computer use. Astra's ability to take over your computer is the star feature of this new model. It's just on another level. In the ChatGPT app, Ask Astra for some easy midweek meal ideas. Pick your favourites and then tell it to go and buy all the ingredients from your usual supermarket online. It opens the site, fills the basket, and it does ask you before it pays. It's basically like gusto or hello fresh at a fraction of the price. The second one is build a tool you currently pay for. I know you've heard this before, so called vibe coding, but this is the first time that this feels really accessible to non technical people. So get the desktop app, switch to Codex, and then in a new chat start with forward slash goal. This tells Astra to keep running for however long it takes until the app or website is finished. And because it's so good at using your computer, it can test the app or website itself to make sure it really works. I'm currently rebuilding my meeting notes app, voice transcription app, my Notes app, my social media management software. The sky is the limit with this stuff. No. 3 is make 3D models. I know that might not sound obviously useful to you, yet Astra is just so good at this. Like, you could give it a Zillow house listing and it will build you a 3D model of that house that you can walk around just by using the arrow keys on your computer. I've started making Roblox games out of my kids drawings and crazy game ideas. So much fun to be had. And you'll just come up with ways to use it, I promise you. No. 4 is editing videos. I know this has been possible for a while, but honestly the results haven't been great. If you run a company, you work in marketing, or you're a creator, you wanna be creating more videos. Just install hyperframes. It's a free set of skills and Astra just spins up its own video editor. It cuts your footage, makes motion graphics, captions, everything. Trust me, it's really good. And no, I'm not being paid by open AI. I have no interest in that company. I'm just really impressed if Astra. No. 5 is do large scale life admin. So because it can use your computer very well and very quickly, you can just photograph all the things you want to sell on eBay or Facebook Marketplace and tell it to go and sell them. It doesn't need any instructions. It Will work it out, that's the beauty of it. And those things will get sold. It could do personal finance, could tidy up the folders on your computer. Honestly, this is just a whole other level now. No. 6 is rebuild your crappy website, something you've probably been putting off for years. Or do it for a friend's business. That would be nice. What I would do is sign up for Cloudflare's free hosting. It's completely free. Tell Astra you've done that and then just give it your old website or whatever content you've got. You can even get it to rubbish round on your computer for you. And it will build the website, it will get it hosted, it will do all the technical stuff for you. You just need to point your domain name at it. Job done.

GPT-6 Astra is here. The next generation of OpenAI's large language model has landed. I've had early access and been pushing it to its limits, and in most areas of work, I've found it really impressive, from everyday writing and knowledge work to AI agents, coding, and even building 3D worlds and games. Have we reached AGI? The president of OpenAI thinks so, but there's clearly more room for this to grow. Right now, this is the most I've ever felt AI is a colleague. #openai #ai #chatgpt #chatgpttutorial #llm #tech #technology #learnontiktok

I've had early access to GPT 6 Astra the most intelligent AI ever arguably and it's changed how I work with AI so here's everything you need to know both good and bad so you're ready for when it lands in your chat GPT next week so the headlines for me are firstly the big one it's amazing at using your computer the apps and browser in ways that we've just not seen before I'll do more videos on that so follow me for those but basically it's quite a weird experience you've got two cursors now you've got your cursor and you've chat Gbps and they're both going around working on different apps and different things at the same time it's quite a mindset shift maybe like riding a tandem bicycle with someone else right it's like the first time for me that AI feels like a true colleague and it can work for longer hours even days on one task staying focused on the job without getting derailed it's almost twice as fast on those longer jobs as well so it's just more satisfying to use in general is better at knowledge work although I do think it lacks a little bit in design more on that in a minute it's incredible at 3D modelling world building game building I had it build a Roblox game this morning using Roblox Studio and blender it's an idea I've had for over a year and this is the first time I'm able to bring the idea to life that is so cool and it's so good at code now that it puts building mobile apps and websites in literally anyone's hands you've just got to have good ideas and good vision it hallucinates a lot less it's more token efficient so you hit usage limits less it seems to be far more efficient than Claude Fable so you get the frontier model experience without constantly hitting limits which is good it's not the best at everything I test AI models in my own little test which I call deck bench because creating Powerpoints is the most awful part of my job and it's a really good measure of usefulness for me as a side note the additional benchmarks that AI is tested against in maths and coding etc they're basically obsolete now this model's achieving 99% scores and anyway those benchmarks are always too easy to manipulate play didn't reflect the usefulness of AI in real life and if you want to test AI models and find out what the best one is for you create your own benchmark like I did take the hardest messiest job that you do quite regularly like my Powerpoint decks well the inputs are always different but the outputs are always quite similar and therefore comparable and then give the same prompt to different AI platforms and models and you might be surprised which is best so for instance with my deck bench Powerpoint test GPT 5.6 Soul produced a much more attractive deck than Astra but Astra arguably created a more thoughtful one bit more like a Ted talk bit more intelligent so these models definitely have their strengths and weaknesses the last thing I love about this model and Claude Fable 5.1 to be fair is just the extent to which they understand what you want basically prompt engineering is dead right you do not need clever prompts you just need whisper flow or one of those tools allows you to talk into your computer and then you just tell AI what you want and you let the AI make decisions about what software what techniques to use you just say the outcome you want that's where we're at now and that's a very exciting place to be I think

This is the best AI UGC video generator I've ever tried. The level of realism is just incredible. Use 30WILL for 30% off your first month or year of VEED. @veedstudio #aivideo #ai #aitools #artificialintelligence #tech #technology #learnontiktok

Veed just became the most advanced AI UGC generator. This new version just launched. So here's how to get great results. And you can try this for free. So go to veed dot Io, start with a template that you like the look of. There's loads of really cool looking ones, but if you want, you can tweak that, the Avatar, captions, style, music, track, etc. And then choose the duration. You can generate up to 90 seconds of video and then the platform that you're creating for. And now you can just type in a topic and it'll write a script for you, or you can write a word for word script. This is gonna make scaling social accounts and ad campaigns so much easier, because the quality here is at a level we've just not seen before. I've put a link to Veed in my bio if you to try out.

The ChatGPT settings that stop it saying "yes" all the time and telling you all your ideas are incredible! If ChatGPT is your favourite AI tool, you can really improve the quality and tone of its output with a few simple settings. I'll also show you what you should put in custom instructions in your ChatGPT settings. #chatgpt #openai #chatgpttutorial #ai #aitools #artificialintelligence #tech #technology #llm #learnontiktok

I'm gonna show you the settings to fix ChatGPT being overly positive and saying that everything's a great idea. So save this video for when you need it and follow me to keep learning about AI. So they've added a few settings to change ChatGPT's personality. I'm also gonna give you something you can paste into this custom instructions box. Take this to the next level. This stuff is why a lot of people switch to Claude. But if you like using ChatGPT, just do this. So go to settings, then personalization. Everything I'm gonna show you is on this page. First thing is base, style and tone. It's a drop down. It'll be on default, but the other options are professional, friendly, candid, quirky, efficient, and cynical. For most people, I would recommend candid or efficient. Candid is just really direct. It will just tell you when something's not a good idea. Efficient cuts to the chase, uses less words. I personally like that one. Cynical is fun, but it's a bit sarcastic. It just throws shade at everything. I kind of got sick of that. Professional is just the closest to default, so not very useful. Now underneath that is a section called characteristics. This is the newest addition to this settings page. 4 things here warm, enthusiastic, headers and lists and emoji. And each one's a drop down where you can choose more, less, or default. So I would recommend having this as less warmth, less enthusiasm, Default headers and lists and less emojis. Now go and try that, see how it feels. You will notice a difference and maybe even go back and tweak the settings. Right? You might want to put a bit of enthusiasm back into it. But then further down the same page is custom instructions. And this is where you can, in plain English, just write how you want it to respond. I actually built a little free tool for Fun. AI Custom instructions.com and you just answer a short questionnaire and it will give you really good custom instructions to paste into your AI. But you can just write them yourself. Just be really specific about how you want it to do things like kinds of words and language. You don't want it to use metric versus imperial measurements, you know, those kind of things. Whichever way you do it, just know that this is something you can be tweaking over time. So every time you see ChatGPT, like having a bad habit, you know, doing something that annoys you, you can go in here and just say, don't do that.

Your AI team just got a whole lot smarter 🤖 Tutti 󠁯· VM is the multiplayer space where AI agents and humans collaborate in real-time. Build, test, and deploy multi-agent systems without the chaos. #TuttiVM #BuiltwithtuttiVM #ai #aitools #agents #automation #productivity #tech #aibusiness #multiagent

This free platform allows my and your AI agents to work together in a shared space on the same files, which hasn't been possible before. It's called tuttvm. I've got a few invite codes, so comment tuttvm and I'll send you one. First come, first served. So, Claude Code, Codex, the best coding agents all run locally on your laptop, but they're locked in isolation. They can't work together with your colleagues agents. In 2TVM, you just create a room and connect your agents, and now everyone in the room sees the same thing. The files, the conversations, what every agent is doing. And the agents already share all that context behind the scenes. So when you want one to pick up another's work, you just at the session. No re briefing required. And even if my Claude code and your codex try to edit the same file at the same time, tut tut resolves those conflicts in the background. It's a really clever system. And you can even authorize other people to use your agent and your usage limits. So this is definitely the future of teamwork between AI agents and humans. But anyway, if you want that invite code for early access, comment tooty, I'll drop it to you straight away.

How I set up every new Claude Cowork project. Just 4 simple steps! 1. Create a new folder inside the master Claude folder 2. Point Cowork at that folder. 3. Tell Claude to create instructions for that folder so that we have permanent context in a Claude.md file and an ongoing log of the work in a memory.md file. 4. Load it with context in the form of any files you already have for this project. #claude #anthropic #claudecowork #ai #aitools #artificialintelligence #tech #technology #learnontiktok

Here's how I set up every new Claude co work project. It's just four simple steps. Step one is to create a new folder for that project and that will live within a master folder called Claude, where I keep all the projects that I work on with Claude. Simple, right? So maybe a new client, a new video series, a new keynote talk, anything like that gets its own folder within the master folder. Step 2 point co work at that folder, start a task, select that folder and that's now your shared workspace with Claude. It can see everything that's in there, work on any documents in there and create new ones in there. Right, so it stays nice and organised. But we do need some context cause this is a completely empty folder. So step 3 is to set the folder instructions. These are the instructions that will always be read every time we do a new task here, just so Claude knows what he's working on and what the end goal is. Side note the great thing about this system is if tomorrow I decide to start using ChatGPT or another AI instead, no problem. I can just point it at this same folder and it will be able to read the instructions and continue working on this project anyway. What I do is I just prompt Claude, set up this new project folder. It will then write those instructions into two markdown files Claude Dot MD, which is the main instructions that might change a bit over time, but ultimately are the kind of ground truth of what is happening in this project. And then memory dot MD, which is more of a log of what's been going on in this project, so the AI can remind itself where we're at with it. And this will constantly change and be updated as you work on this project. And Claude might decide it needs to create other markdown files with more specific sets of instructions, like brand guidelines or specific process steps or something. It might also create subfolders, again, that it feels that would be helpful structurally. I generally just let it do its thing, but I do sometimes find that it's prone to overkill here. So sometimes I have to go back and say there's too many sub folders, too many different instruction files. It's a simple project. Simplify the setup. And then step 4 is to load it with real work context. So just drop in anything you've got in terms of existing information around this project, like if documents, Powerpoints, email chains, meeting notes. Right. And then you've got this beautifully organised folder. Because part of this is about training yourself to be more organised with your work, because nowadays we have to be. We've got an AI colleague who needs things neat and tidy.