Most people use AI like a magic 8-ball
They type, they pray, they paste, they ship something mediocre, and then they wonder why their output reads exactly like every other person's. The whole thing looks productive from the outside and goes nowhere.
A breakdown that came across the feed reframed the whole problem. Seven habits separate the people getting an unfair edge from the people writing the same generic copy as everyone else. It isn't about prompts. It isn't about tools. It's about the order you do things in, what you commit to, and what you refuse to delegate.
Three of these matter way more than the others. Here's the lift.
*Ad
Have you been noticing how Claude is the most talked about tool on the internet right now? CLaude has been launching new models & features like claude co-work, claude design, skills, connectors literally every week.
The World’s first Claude-a-thon, A 2 day deep dive into Claude, its user-cases and 10+ other AI tools happening this weekend from 10 am-7pm EST – and they have just 1000 free seats for a limited time.
In the workshop, you'll do deep research on Claude. Build your own artifacts and dashboards. Create full presentations using Claude. Set up Claude Connectors like Indeed to automate your job search.
Register NOW! (free for the next 48 hrs)
🧠Live sessions- Saturday and Sunday
🕜10 AM EST to 7PM EST
Brief the model before you prompt it
Most people open a fresh chat and start typing. That's the amateur move. The fix is boring and unsexy: a folder called AI Files. First file inside, who you are, your tone, your audience, the kind of work you do. Upload it. Define the task. Define what success looks like. Then prompt.
You're not prompting anymore at that point. You're briefing. There's a huge gap between the two and the output makes the gap obvious within one reply. A briefed model writes like someone who already knows your voice. A cold-prompted model writes like a Medium intro from 2019.
If you only change one thing this month, change the order. Set up first. Prompt second.
Master one tool. Delete the rest from your bookmarks.
The chase is the problem. Forty tabs open, three free trials running, a half-finished Notion doc comparing them all, and zero deep skill in any of them. The fix is ugly: pick one. Use it for thirty days. Only that one. Touch the boring features nobody bothers with, like Projects, memory, file uploads, search, custom instructions.
Going deep beats going wide by a wide margin. The people who look like wizards with one tool didn't find a hack. They just stopped chasing the next thing for long enough to actually learn the current one. The next shiny tool on Twitter will still be there in thirty days. It will not change your output if you can't run the one you already have.
In partnership with
ChatGPT gives you generic answers because you give it generic prompts.
You know the fix: longer prompts, more context, clearer constraints. But typing all that takes five minutes per prompt, so you shortcut it. Every time.
Wispr Flow lets you speak your prompts instead of typing them. Talk through your thinking naturally — include context, constraints, examples — and get clean text ready to paste. No filler words. No cleanup.
Works inside ChatGPT, Claude, Cursor, Windsurf, and every other AI tool. System-level, so there's nothing to install per app. Tap and talk.
Millions of users worldwide. Teams at OpenAI, Vercel, and Clay use Flow daily. Free on Mac, Windows, and iPhone.
Teach the model what you already know
This one rarely gets articulated this cleanly. Most folks try to write out a system prompt by hand and run out of patience by paragraph two. There's a much easier move.
Open a chat. Tell it: "Ask me questions about my expertise." Let it interview you. Your rules, your no's, your audience, the things you won't say, the angles you always take. Twenty minutes later you have a knowledge base built by being asked, not by trying to remember everything from scratch. Export it to one .md file. Reuse it for months.
You can build a personal AI advisor without ever writing a system prompt. The model knows how to interview you better than you know how to write about yourself.
Talk to it like a colleague, not a vending machine
The vending machine model is: drop in a prompt, expect a finished thing to fall out. That's where the bad output comes from. The colleague model is conversation.
Three moves change the game here. Open with "Don't start yet. Ask me questions." Read v1 carefully and name every specific thing that's wrong, by name, not "make it better." Then push harder: "Argue against this draft." That last one is the underrated move. Asking the model to attack its own work surfaces the soft spots you'd otherwise miss and ship.
Three exchanges with the model beat one perfect prompt every time. Stop trying to one-shot it.
Ship rough. Lead the model. Don't follow it.
Two skills braided together at the end of the breakdown.
The first: build the rough draft with AI in twenty minutes. Show it. Let people react to a real thing instead of polishing something nobody has seen yet. Rough in public beats perfection in private. Every single time. The feedback loop is the product.
The second: split every task into two buckets. What does the model do, what do you do. Give it the 80%. Keep the 20%. The 20% is judgment, taste, and the parts where being wrong actually costs you something. The line that stuck most:
"If you can't spot the mistake, don't delegate it."
Better quality filter than any prompt framework. If you can't tell whether the output is right or wrong, you're not leading the model. You're being dragged behind it.
Three things to actually do this week
🔹 Create an AI Files folder. One file inside: who you are, your tone, your audience, the work you do, what success looks like. Upload it to your tool of choice. Reference it on every new chat for a week. Watch the cold-start problem disappear.
🔹 Pick one tool. Close the rest of the tabs. Whichever you already use most. Thirty days, only that one. Touch the boring features: Projects, memory, file uploads, custom instructions. The next shiny tool on Twitter will still be there in a month.
🔹 Add "Argue against this" to your next draft. Whatever you're about to ship, paste it back into the model with that one instruction. Read what it surfaces. Most of it will be right. The output you ship after that conversation is the one that doesn't read like everyone else's.
Your ads ran overnight. Nobody was watching. Except Viktor.
One brand built 30+ landing pages through Viktor without a single developer.
Each page mapped to a specific ad group. All deployed within hours. Viktor wrote the code and shipped every one from a Slack message.
That same team has Viktor monitoring ad accounts across the portfolio and posting performance briefs before the day starts. One colleague. Always on. Across every account.
*Ad
The skills that compound
The reason this list works is that it isn't a list of prompts or hacks. It's a list of habits. Hacks decay the moment the model updates. Habits compound across every model, every tool, every workflow you'll touch for the next five years.
The middle gets automated. Speed and volume aren't where the edge lives. Taste, judgment, original thinking, and the ability to lead the model live in the 20% you refuse to delegate. That's the part nobody can copy off you.
Pick one of these and start this week. Not all seven. One. That's the whole game.
Which skill are you working on this week?
