Seven Habits for Using AI Like a Top CEO

📷 AI illustration · Hakolah

The creator of this video argues that the people who get the most out of AI mostly differ in how they talk to it, more than in which model or tool they use. He lays out seven simple habits, which we summarize here in our own words, with examples of how to apply them.

Source: the video "How to Use AI Like A Top 0.1% CEO" by creator Nick Saraev. This article is a summary and read in our own words, not a literal translation.

1. Talk, Don't Type

The average person types around 30–50 words per minute but speaks around 150–200. Not every spoken word will be perfect, but you still get far more across to the AI by speaking than by typing.

Typing30–50 words per minuteSpeaking150–200 words per minute (3–4× faster)
Your voice is three to four times faster than your fingers.

How to apply it: use any free voice dictation tool, ideally one you can train on your own words and terms so it gets more accurate. (We covered Wispr Flow as one example in an earlier article.)

2. An Always-On Voice Assistant in Meetings

Picture an executive assistant who's always with you: listening to the meeting, taking notes, and turning what's said into tasks. Big companies use real-time transcription for exactly this, and the most extreme version is a system that picks up the to-dos mentioned in a meeting ("so-and-so needs to do this…") and starts doing them before the meeting is even over.

🎙️ Meetingvoice + screen📝 Live transcriptfull text✅ Taskswho does what⚙️ Executionduring the meeting
From talking to done, without anyone typing a line.

How to apply it: ask the strongest model you have to build a tool that listens to your mic and system audio, transcribes the whole meeting, and then creates tasks in the systems you use every day. The video's creator says running something like this can cost as little as a dollar or two a day.

3. Show, Don't Tell

A picture is worth a thousand words, and a video is 24 or more pictures every second. So the best users ask the AI itself to make visuals that explain the idea: interactive charts, 3D models, or an explainer video.

Example: if you're moving into a new office, go past the floor plan and ask for a 3D model with realistic lighting that you can walk around in, then furnish it with real products to try before you buy. The same goes for turning dense reports into clear visuals. (The illustrations in this very article were made in code this way.)

4. Have the AI Interview You

Telling it what you want often isn't enough, because you may not know enough about the subject, or may struggle to put exactly what you want into words. Relying only on what you know is like trying to pipe an ocean through the straw of your own experience.

The ocean: all it knowsThe straw: your experience alone
Let it ask you, and it reaches what you didn't know you needed.

Tell it: "I want to get X done. Ask me anything you might need to know to suggest the best way of doing it."

5. Ask for Flaws, Not Opinions

Models are trained with reinforcement learning from human feedback (RLHF), so they're sometimes rewarded more for agreeing with you than for being right. Ask for their opinion and they'll usually lean toward praise, even when you're wrong.

✗ Instead of: "What do you think of this?" · "Is this email okay?" · "What do you think of my project?"
✓ Ask: "What's wrong with my idea?" · "Poke holes in my thinking." · "Roast my essay."

Once you've heard the criticism, you're the judge of whether the bad outweighs the good.

6. Try It Yourself

If you doubt a model can do something, ask it to do it instead of asking people. Tools change so fast that, according to the video, a score on a benchmark once seen as near-impossible can climb past half within a single month.

53%A hard benchmark over 30 days
What's true today may not be true tomorrow, so try it.

The video notes that Nvidia's CEO runs AI coding tools on his own computer and tries them himself, on the idea that the best answer to "is this possible?" is to try it at least once.

7. Keep an Open Mind

AI tools will take many forms: digital avatars that chat with you like a coworker, small wearable devices, ways of working we aren't used to yet. Try them. Whether it works or not, you'll have learned what the tool can do, and the video argues that not trying costs more these days.

Not tryingTrying
"Not trying" is the heavier cost.

Summary