STOP Using ChatGPT Like It's 2023 (Do This Instead)

0:00When ChatGPT first launched, it genuinely looked like magic. You typed a question and a machine wrote back. But, that was 2023, and if that's still your routine, you're not even using half of what it can do now. So, in this video, I'll go through the seven habits to drop and what to do instead. And by the end, the same subscription you've already got gives you back a lot more. 2023 was the year AI went mainstream. ChatGPT was brand new and it hit 100 million users in 2 months, faster than any app in history. And the thing that frustrated everyone back then was that it forgot

0:30you completely the moment a chat ended. So, every conversation started with who you are, what you're working on, and how you want it to answer. That's the first habit to drop, introducing yourself at the start of every chat. It doesn't work like that anymore, and the fix is in the settings. Open settings, then personalization, and there are two things in there worth setting up. Custom instructions is where you tell it once how to answer you, and every chat after that starts from there. Mine tells [music] it to keep answers short by default and ask me one question when my request is vague, plus use US context

0:58[music] and skip the compliments about my questions. By the way, every prompt and template I use today, including my exact block, is packed into one free doc. Get it from the link below and follow along. Memory is the other one, and it's the part that's changed the most since 2023. It picks things up as you chat, like your job or the project you keep coming back to, and it uses them in later answers without being asked. [music] And

1:21you can read everything it's remembered. There's a page in that same menu listing every single memory, and you can edit or delete any of them. That page is worth 2 minutes of your time because half the people I show it to have no idea it exists. It can also draw on your past chats, so the context you built up over months gets put to [music] use. And that's why the answers start fitting you. And if any of that makes you uneasy, the off switches are on the same page. Memory can go off entirely, or you use a temporary chat for anything you don't want remembered. On free, you get

1:49a limited version of memory. The paid plans remember more and for longer. Set those two things up and and never have to introduce yourself again. You're still typing everything in by hand though, and that habit's next. Habit two is copy-pasting. Retyping a document into the chat, then copying the answer out and fixing it somewhere else. That was 2023 ChatGPT, a text box that only understood text you typed into it. Now you just give it the thing itself. Drop in a PDF and ask questions against it, or photograph a page of handwritten notes and it reads them, or screenshot an error and ask, "What am I looking

2:20at?" So, I'll take a rental contract, drop it in and ask, "Which clauses here are unusual for a US lease?" And it pulls the odd ones out with the reasoning next to each. On free, it can take roughly 12 pages of text in one go. Go and plus stretch that to about 40. Pro pushes it further to around 250, and the thinking models handle anywhere from about 300 to nearly 700 pages. The other half of this habit is fixing its answers somewhere else, and you don't need to anymore. Anything long, a draft or a block of code, comes back as an editable block right in the chat. So, instead of

2:50regenerating the whole answer because one paragraph annoyed you, you change that one part in place. I'll ask it to tighten the opening of an essay draft and keep my tone, and it rewrites just those lines while the rest stays untouched. And ChatGPT runs apps inside the chat now, things like Canva and Spotify. So, the deck you were drafting can get built without you leaving the conversation. The file goes in whole, the edits happen in place, and there's nothing left to copy out. That fixes what you put in, but what it hands back is a different problem. Asking ChatGPT for facts and just hoping, that's habit three. In 2023, it answered everything

3:21from memory, which is why it made things up with total confidence, and [music] that reputation stuck. The tool moved on. It searches the web now and shows you where each claim came from. Ask about anything current, like a price or a news story, and the answer comes back with links you can click and check. And the habit that should replace hoping is checking. So, glance at where a claim came from before you use it, same as you would with any source. A Spanish travel influencer found that out the hard way last year. She asked ChatGPT whether she needed a visa for Puerto Rico, it told her no, and it never mentioned the ESTA

3:50you actually have to get. So, she was turned away at the airport and the whole meltdown went viral. For the questions that need more than one search, there's deep research. You give it a proper brief, it goes away for several minutes, reads across sources, and returns a structured report with the references attached. I'll run one, and the brief I give it is best value home EV chargers in the US including installation. And it plans the research first, then works

4:13through it while I do something else. There's a brief template in that free doc, too, and the wording matters less than covering what you're deciding what to include and which sources to prefer. When the report's done, you get sections, the numbers pulled into comparisons, and every claim linked, and the same rule applies here. So,

4:29spot-check the claims you'll act on. Free and Go get a small deep research allowance. Plus is where it gets big enough to use every week. So, the 2023 rule was never trust it with facts, and the 2026 rule is let it fetch them, then check the links. And there's one habit under that one that's easier to miss, accepting the first answer at all. The fourth habit is treating every question like it deserves an instant answer. Back in 2023, you got one model at one speed, and whatever came back first was the answer. Now, there's a model picker at the top of the chat, and the split that

4:58matters on it is instant against high. Instant is for everyday questions, and high works through the problem before it answers. You can watch it reason for a while before anything appears. So, I'll give it a proper one and say, "Plan a 3-week Japan trip for two on a 2 and 1/2 grand budget each, flights out of Chicago." On instant, that's a generic list. On high, it's an itinerary that shows its math and walks through the trade-offs. On plans, free doesn't get the reasoning models, plus does with the effort levels you can pick. So, from now on, ask yourself one thing before you hit send. Does this deserve thinking

5:29time? If yes, give it some. That pause is a small swap, though, and the bigger one is whether you type your questions in the first place. That's habit five, typing it all when you don't need to. Voice mode in 2023 was a gimmick. Now, it's a proper conversation. You can interrupt it mid-sentence and it adjusts

5:44without losing the thread. [music] I use it for thinking out loud on a walk or while driving, and the ideas you'd normally lose because your hands were busy get worked through instead. And voice with video is in there now, too. You point the camera at something and talk about what it's seeing. Voice works on every plan, video needs one of the paid tiers, even the cheap one. The paid plans also get record mode. It records a meeting or a voice note and gives you the transcript with the action points pulled out. If you'd explain it out loud to a person, you can explain it out loud

6:12to this instead. All of that still happens one question at a time, though, and the bigger jobs in your life don't fit in single questions. Habit six is putting everything into one chat. It's still named new chat, it's 400 messages long, and if you're trying to find something in there, you're scrolling back to 2023. Projects fix it, and they're on every plan now, including free. A project is a folder where you put files and instructions, and every chat you start inside it already has both. So, I'll make one for a house move. The instructions say we're buying a first home in Austin on a 350 budget,

6:41[music] and to flag anything about HOAs I should know. And the files are the mortgage pre-approval and the short list. Every showing question and inspection report and every what does this fee mean goes in as its own thread inside that project, all drawing on the same files. [music] So, mid-hunt I ask, "This listing mentions a special assessment. How does that compare [music] with the eight flags in my files?" And it answers from my own documents with the caveats attached. The same setup works for anything that runs longer than a week. A dissertation project holds the reading list and every [music] draft. A wedding project holds the quotes and the guest

7:11list. And the split from the settings habit carries over here. Custom instructions tune how it answers you everywhere, while a project holds what it needs for one piece of work. You can share a project, too, so the house move one runs with your partner in it, both asking questions against the same documents. One project for each thing you've got going [music] on, and the endless scroll chat is finished, which leaves the one where you do all the follow-up work yourself. The last habit to drop is being your own assistant. In 2023, the tool couldn't check the same things for you every morning or remember to follow up or run the same weekly

7:42question by hand, because it only existed while you were typing at it. Tasks change that. You tell it what to do and when, and it runs on schedule and messages you the result. Mine include a Monday morning one that checks what changed in AI tools that week, and when I was job hunting, a daily one that watched for new [music] listings matching a saved brief. They've got their own page in the sidebar now, so you set one up from a chat or straight from there, pick when it runs, and every task sits in that list for you to pause, edit, or switch off. And the one habit

8:10that covers all seven is relearning a tool you thought you already knew. That's the whole point of my school community, AI fluency. It keeps you current with every tool worth using. So, join through the first link in the description and stay ahead of the next round of changes. [music] Thank you for watching. I'll see you in

8:25the next one.