We had decided to split into smaller groups of four to six. The purpose was simple. To work towards our goals, share our experiences and support one another.
The next day, I started wondering about something.
How would we keep track of all this?
Without letting the administration take over the work. Without yet another project management system. Without long task lists, notifications and dashboards that nobody wants to open after three weeks.
Just something small.
A system that helped each person record one goal, a few important actions and what happened between meetings.
A system that supported the conversation without becoming the conversation.
What do I actually need to learn?
I have long been interested in MVPs.
The term is often understood as a small or incomplete version of the product we intend to build one day.
But I don't think that is quite right.
An MVP is not necessarily the smallest version of the product.
It is the smallest credible experiment that can teach us something that matters.
So before deciding what to build, we need to ask:
What do I actually need to learn?
And what decision will I make once I have the answer?
In this case, the question was not whether I could build a system for mastermind groups.
I knew I could probably build something.
The question was whether a small, clearly defined framework could help a real group keep track of goals, actions and mutual support.
Would people create a group?
Would they invite others in?
Would they record real goals?
Would they update their progress?
Would the group use the system at its next meeting?
And would people come back once the novelty had worn off?
I could spend a few months thinking about these questions.
Or I could start looking for answers.
Scope is more than a document
On 17 September, I started building.
Before that, I had tried to define what needed to be in the first version and, just as importantly, what should not be there.
One goal per person.
One to three actions.
A private group.
A shared view of progress.
Monthly meetings where the group celebrates wins, discusses lessons learned and decides on the next steps.
Not a project management system.
Not a social network.
Not a messaging platform.
Not everything that might possibly become useful later.
Experience has taught me that scope is more than a document. It needs to be clear enough to make it difficult to negotiate with.
I once tried the same task, the same prompt and the same scope document on two different AI models.
One returned 11,295 words.
The other returned 2,747.
The same task. The same definition. Completely different interpretations of what “within scope” meant.
Sometimes you don't know whether to laugh or cry.
When AI lets us build more in less time, defining boundaries does not become less important.
It becomes more important.
The question is no longer just what we can build.
It is also what we have the judgment to leave unbuilt.
If a feature does not reduce an important uncertainty, help us learn or make the experiment credible, why should it be included?
Maybe later.
But not yet.
Four days later
On 21 September, Mastermind.is went live.
Four days after building began.
I think the product is great.
But that is simply not what matters most.
I built it. I know the thinking behind it. I know why each part is there, and of course I can see the value I hope it will create.
But my own opinion is not meaningful validation of the hypothesis.
The people the product is intended for need to see value in it.
They need to use it.
Or not use it.
They need to run into things I did not anticipate, misunderstand something I thought was obvious and ask for something I did not consider important.
Maybe the system helps the groups.
Maybe it gets in their way.
Maybe this is exactly enough.
Maybe it is still too much.
I don't know.
But I am not going to wait a year for the first real evidence.
Now the experiment can begin
It is easy to confuse launching a product with validating an idea.
They are not the same thing.
The fact that Mastermind.is is live does not prove the hypothesis. It mainly proves that it was possible to build a version with a clearly defined scope and put it in people's hands.
Only now can the experiment begin.
Now reality can answer.
That is really the biggest advantage of being able to build quickly with AI.
Not that we can produce more features.
Not that we can launch more products.
But that we can shorten the time we spend believing in our own ideas without letting reality test them.
If I had spent the next few months adding features the experiment did not need, I would not necessarily have been improving the product.
I might have been postponing the answer.
An idea can look beautiful inside a folder.
It can be logical, well thought out and backed by a convincing business model in Excel.
Nobody uses it.
Nobody misunderstands it.
Nobody rejects it.
That is why it can keep looking great.
A hypothesis learns nothing when its address is:
Drive → Folder 17 → Ideas.
Where does yours live?