By default AI answers like a polite assistant: smooth, safe, empty. Ask it about business and you get textbook advice everyone already knows. Founder thinking mode breaks that 🔥
When to use it
When you ask AI for a business decision — strategy, pricing, hiring, the next step — and get back averaged-out mush: "test your hypotheses", "listen to your customers". All true, all useless. Founder Mode assigns the model the role of a seasoned operator: reasons from first principles, has built and sold companies before, answers not like a consultant but like someone whose own cash register is on the line.
The result: the answer opens with the line "Here's what I'd actually do" and delivers a concrete move, the trade-offs, the risks and what almost everyone misses. It's the same role technique as qvib's "Persona: build an expert role for the task", just sharpened for a founder.
The prompt (copy and paste)
You are in founder thinking mode (Founder Mode).
Think like an operator: you reason from first principles and have already
built and sold companies. You are not a consultant and not a people-pleasing assistant.
CONTEXT: <business, stage, key numbers — MRR, burn, team; 2-4 lines>.
TASK: "<INSERT THE TASK>".
Rules for the answer:
- No generic advice, no filler — give the precise call a seasoned founder would make.
- Name the trade-offs, the risks and what almost everyone misses, explicitly.
- Reason from first principles, not from "that's how it's done".
- If there isn't enough data, say which 2-3 numbers you need to decide.
Start every answer with the line: "Here's what I'd actually do".
Filled-in example
CONTEXT: SaaS for auto repair shops, MRR ≈ $4k, 40 paying customers, growing ~5% a month, two of us, 7 months of runway. TASK: "Where do I aim the whole next quarter — growth or retention?"
What the AI typically returns: it opens with "Here's what I'd actually do", then a concrete choice ("at 40 customers you patch the hole in the bucket first, you don't pour traffic in"), the trade-offs (growth versus 7 months of runway), the risks (you run out of money before the funnel spins up) and what almost everyone misses ("you're not measuring churn by cohort — without that the growth-vs-retention debate is meaningless; here are the 2 numbers you need to decide").
Variations
- Investor anti-persona. "You are a hard-nosed Series A investor doing due diligence. Tear apart my model and tell me why I will NOT raise the round." A stress test through the eyes of someone hunting for holes rather than handing out praise. Run the risks from that answer through qvib's "A one-page business plan for your hypothesis (Lean Canvas)".
- A board of three. "Answer as a board of three operators: product founder, sales founder, CFO — each with their own position, then reconcile them into one decision at the end."
- Breaking a decision loop. "I've been stuck between A and B for two weeks. Make the call like an operator and name what I'm afraid to admit." For a gentler, step-by-step version use qvib's "Help me make a decision (weigh the pros and cons)".
- Revenue right under your nose. "Where in this business is revenue hiding that I can't see? Give me 3 moves for the next 30 days — cheapest first."
- Hiring without red flags. "Draw up a hiring plan for this role like a founder: where to look, 5 interview questions and the red flags almost everyone misses."
Pro tips
- Move the role block into the System layer — ChatGPT Custom Instructions or Claude Projects. Then the mode holds across every chat, not just one window.
- Give it numbers. Founder Mode without context is just a confident tone. The value shows up when you feed it real inputs: MRR, burn, conversion, churn.
- A role doesn't create knowledge the model doesn't have. "Think like a founder" won't replace what you know about your market and customers — the model is good at pressing on weak spots and taking the fear out of deciding, but the responsibility is yours. Verify anything that costs money.
- Demand trade-offs and risks explicitly. Without that line even the "founder" slides into "do both" — and the whole point is to make the model take a position.
- The opening line "Here's what I'd actually do" isn't decoration. It forces the model to pick a move immediately instead of fanning out options. If you want a decision, keep it.
- Where the name comes from. "Founder Mode" comes from Paul Graham's essay (September 2024, based on Brian Chesky's talk at Y Combinator): about a founder who gets into the details rather than only delegating. Here we borrow the idea itself for a role in a system prompt — this is a persona technique, not the management approach from the essay.