I used to treat AI like a vending machine. Punch in a request, grab whatever drops out, walk away a little disappointed. Turns out I was hiring a private chef and asking for a peanut butter sandwich. This whole reframe comes from Dan Martell, the founder behind Martell Ventures, who builds AI products daily and just dropped a full Q&A breaking down how he actually uses these tools. I was genuinely hooked watching it, because he doesn’t stop at “use ChatGPT.” He hands you the exact move for each problem.
So here’s the analogy I want you to hold onto: AI is a full kitchen staff. You’ve got a chef, a prep cook, a sous chef, and a manager. Most people only ever talk to the front-door greeter. Let’s walk through the kitchen the way this expert lays it out.
🔪 The prep work: give the kitchen your ingredients
A chef can’t cook without ingredients, and AI can’t help without context. The problem is typing takes forever. Martell’s fix is simple: talk, don’t type. He uses a voice tool (Wispr Flow) that lives on his phone and laptop so he can dump context three times faster than typing. His second trick flips the whole thing around. Just add “ask me questions if you need more context” to the end of any prompt, and the AI pulls what it needs instead of guessing.
📋 The recipe card: the MAPS framework
Generic answers happen because AI averages out everything ever written. To get something sharp, the creator uses a recipe card he calls MAPS:
- Mission: the real goal behind the task (“I need more customers”).
- Ask: the specific thing you want built (a lead list, a landing page).
- Parameters: everything about your situation. Old scripts, past clients, whole books of context. Most folks give a paragraph.
- Shape: the format you want back. Spreadsheet, bullet list, short and tight.
He points out you can feed it two or three full books’ worth of info, and that’s where the voice tool earns its keep again.
🧠 The head chef who remembers you: the master prompt
Every new chat feels like meeting a stranger. Martell’s five-minute fix changed the game for him. Instead of filling out a giant questionnaire, he connects the AI to tools that already hold his info: email, Google Drive, Notion, Slack. Then he tells it to “build the most detailed master prompt of who I am from all this information.” It nails about 90%, you fix the hallucinated bits, save it, and paste it into every other AI you use. He leans on Claude here for its memory system and because he trusts Anthropic with sensitive data.
🍳 The line cooks: putting the staff to work
This is where the video really opens up. A few of my favorite stations:
- Learn anything fast: He tells AI to find only recent resources, then build a learning plan and teach the relevant parts. His daily driver is NotebookLM, which spins your sources into a podcast, quizzes, even a video overview.
- Set and hit goals: Turn vague wishes into SMART goals, answer questions out loud, then have it generate a hyper-realistic vision board. Pro tip from the creator: resize it to 9×6 so it becomes your phone wallpaper.
- Build a daily dashboard: Feed it your goals, get a written plan, then a red/yellow/green tracker. He builds these with Codex and even turns them into a little mobile app.
- Audit your calendar: Connect it, hand over your goals, and tell it to act like a “ruthless mentor” who argues with you about where your time goes.
- Write like you: Connect Gmail or Outlook, have it study your emails, and build a tone document so future drafts sound human.
- Automate repetitive tasks: His rule is DRY, don’t repeat yourself. Record yourself doing a task while narrating, then have the AI turn it into a reusable “skill.”
🗂️ The pantry and the manager: organizing and running everything
Messy notes everywhere? The creator picks one home base, points every tool at it, and lets an AI agent clean it up nightly. He personally uses Obsidian for private stuff and Notion for team work. For finances, he built HelloFrank (one of his own ventures) to act like a CFO that connects to bank accounts and emails a daily cash position. And for the assistant everyone wants, he splits it into “single player mode” (his agent Apex, aka Kai) and “multiplayer mode” for teams (Precision). Full transparency, those last three are his companies, and he says so plainly.
🛠️ When the kitchen catches fire: fixing code errors
Even non-coders will love this one. His big idea: show the problem to the AI. Ask your code to create a visible error window, screenshot it, drag it into the AI, and say “fix it.” He mostly uses Claude Code, and when it’s stuck, he pits Grok or Codex against it so two models fix each other.
How to apply this today
- Pick one station, not all of them.
- Grab a voice tool so context stops being a chore.
- Run your next request through MAPS.
- Build your master prompt once and reuse it everywhere.
The honest kicker Martell lands on is that tools change every 90 days, but the principles don’t. Most people watch and do nothing. I’d love for you to be the exception, so go watch the full video for the exact prompts and walkthroughs, then pick one thing and actually run it.