The £3k Lab · Live experiment
The 30-Day AI Challenge
Five AI models. One brief. No budget. Each one has to build a business that actually makes money, and I do the legwork. Only real results count: leads, interest and revenue.
Starting soon. I'm getting everything in place before Day 1 begins. This page fills up once the challenge is under way.
- Models competing
- 5
- Current day
- Not yet
- Still in
- 5
- Earned so far
- £0
The idea behind it
Gary Vaynerchuk makes a simple argument: social media as we knew it is over. Your feed no longer shows you what your cousin or your old school friend posted. It shows you whatever you've been interested in lately. He calls this interest media.
If he's right, follower counts matter far less than they used to. Someone with no audience at all can reach the right people, as long as they're relevant to what those people are searching for and engaging with.
So here's the question. Can you build a real, profitable business on that, starting from nothing? I've asked five AI models to find out, and to compete against each other while they do it.
The scoreboard
Each model gets its own business. Updated after every review.
| Model | Business | Status | Leads | Earned |
|---|---|---|---|---|
| Claude | Choosing on Day 1 | In | 0 | £0 |
| ChatGPT | Choosing on Day 1 | In | 0 | £0 |
| Gemini | Choosing on Day 1 | In | 0 | £0 |
| Grok | Choosing on Day 1 | In | 0 | £0 |
| DeepSeek | Choosing on Day 1 | In | 0 | £0 |
The rules
- Zero capital. Every business starts from £0. No ads, no paid tools bought for the challenge.
- One day at a time. No model is allowed to hand me a full 30-day plan. We work one day, then the next.
- Thirty minutes, two tasks. Each day, each model gets one 30-minute planning session to set a maximum of two concrete tasks. How long I spend doing them is up to me.
- Report before moving on. No model gets its next day until I've reported back with real results.
- A "day" is a step, not a date. Thirty days is the structure, not the calendar. There may be a week or more between days, and that's fine.
- Regular reviews. Every seven challenge days, the results get compared. Winning ideas are shared across models, and anything that isn't getting traction is cut or changed.
The sounding boards
Results are judged on what actually happened, not on which AI wrote the best-sounding plan. When a call needs a second opinion, I have two to choose from.
The Voice of Reason
Gemini
Ebullient, optimistic, always ready with a plan. The nickname is only slightly tongue in cheek.
Feet on the ground
Claude
Steady, practical and happy to point out the hole in the bucket.
Final say
Me
I do the work, report the numbers and make the call on culls and pivots.
They're the yin and the yang. Both are also in the race, so if a decision affects one of them, the other gets asked.
Fifteen free roasts on Day 1
Every model has to open with three business ideas, which means Day 1 alone hands me fifteen brand new ideas to test. Each one goes through Roast My Plan, and the results get published as Teardowns. So even the ideas that don't get picked still get an honest look.
The kickoff prompt
Every model received exactly the same opening message. Here it is, word for word.
"Act as my operational Co-Founder for a 30-day challenge: build a viable, profitable business from $0 capital using Gary Vaynerchuk's concept of 'Interest Media' (leveraging search, explore feeds, and algorithm interest signals rather than traditional follower counts). STRICT RULE: You must NOT give me a full 30-day plan right now. We operate strictly one day at a time. Each day you get one 30-minute planning session to set a maximum of 2 concrete tasks. My execution time is not limited. You cannot proceed to the next day until I report back with my results. You are competing against 4 other AI models; we will cross-pollinate winning ideas weekly and ruthlessly cull or pivot underperforming strategies based on real traction (leads, interest, or revenue). To begin Day 1, analyze how we can exploit interest media algorithms with $0, present 3 distinct zero-capital business concepts, and ask me to choose one so we can begin."
Reviews
What worked, what got cut, and which ideas got passed between models.
Behind the scenes, every model's daily response is logged in full in my notes, so nothing gets quietly edited after the fact. This page shows the highlights.