Product
Flynge
An AI thinking partner for hard decisions, built to help you understand how to solve the problem, not to solve it for you.
I founded it, and I designed it, set its product strategy and built it myself, with AI.
An AI chat session · For the consequential calls · 2026–present · Live in alpha
The problem
Hard decisions rarely fail for lack of an answer
They fail because the person facing them committed to a solution before they understood the problem.
- The assumption nobody tested
- The option nobody argued for
- Personal decisions sit directly beside domains that require a licensed professional
- Nobody brings a genuinely hard decision to something that learns from it
The solution
A thinking partner, not an oracle
Flynge is an AI chat session for exactly those moments: it works the decision, the problem or the task through with you until you understand how to solve it yourself.
- It names the assumptions you did not know you were making
- It argues the strongest version of the option you are about to reject
- A dedicated section states plainly that Flynge is not a substitute for a therapist, doctor, lawyer or financial advisor
- Sessions are private: never used to train models, never shared, never sold
Takeaway: The only useful contribution is the argument they have been avoiding.
Insight
The obvious build is quietly the wrong product
An answer machine demos brilliantly, but the person walks away holding an answer they do not understand and cannot defend the moment it is questioned.
- Understand the decision
- Make the solution yours
- Explain how you got there
Takeaway: For a consequential decision the understanding is the deliverable.
Decisions
What I decided, and what it cost
Help them understand, don’t hand them the answer
Instead of: Just give the answer. It is faster, it demos far better, and it is what most people say they want when they arrive.
Cost: Slower, and a weaker first impression than a tool that simply answers. Some people want the answer and leave. That is the intended filter, but it is still a cost.
Steel-man the position the user is about to reject
Instead of: Help people build the case they walked in with, which is what they ask for, and what makes a session feel productive.
Cost: It is unwelcome, and it makes some sessions feel worse to sit through than a supportive one would. That is the intended cost, but it is still a cost.
Publish what the product is not
Instead of: Stay quiet on the boundary. The apparent surface stays larger and nobody is turned away at the door.
Cost: It shrinks the perceived scope at exactly the moment someone is deciding whether to try it. Twenty years in regulated healthcare made that trade easy to accept.
Never train on user sessions
Instead of: Use the sessions. It is standard, it improves the product, and it is the single most valuable asset the product generates.
Cost: Forgoes the obvious data flywheel. I would make the same call again, and I understand exactly why almost nobody does.
Takeaway: Thinking it through properly is worth the friction.
Outcome
An invitation-only alpha, early by design
- Conceived, designed and built solo in 2026, with AI as the entire engineering capability
- The number that matters is not sessions started, but whether people come back with a second decision
Takeaway: A second decision is the signal that the thinking actually held.