I keep starting projects and never finishing them

Early on I had to decide whether Masterji should let you quit. It refuses to open the next phase until the evidence is in, so the obvious next move is to make it hard to walk away from a goal too — some friction, a confirmation, something that costs you.

I went the other way. Closing is never blocked, in either direction: a builder whose idea died has to be able to bury it, and one who genuinely finished has to be able to say so from whatever phase they were in. What the server keeps is not the power to stop you. It is the reading.

Quitting is not the failure. Quitting untested is

When a goal closes, the record says one of four things, and none of them is a label you get to pick:

• ACHIEVED — you finished it and there is banked work behind it.
• UNVERIFIED — you say you finished, with nothing accepted on the record. Not an accusation. Just what there is to point at.
• INVALIDATED — you made real contact and it said no.
• UNTESTED — dropped before the world got a vote.

That last pair is the whole essay. Two abandoned projects, closed the same afternoon with the same shrug, and the record separates them: one is a finding and the other is a blank.

If you have a pile of unfinished things, my guess is that almost all of them read UNTESTED. Not because you are lazy — because the point at which a project stops being fun arrives reliably before the point at which anybody outside your own head has seen it. You lost interest during the part that was still private. So the pile is not a record of ten things that didn't work. It is ten blanks, and it has taught you nothing about which of your ideas were any good.

Which is also why the pile hurts more than it should. Ten genuine rejections would be a market education. Ten blanks are just evidence about your own follow-through, and that is the least useful thing a graveyard can contain.

Two conversations is the whole difference

The threshold is a constant in the code called INVALIDATED_AT, and it is set to two. Two accepted proofs of real-world contact — from VALIDATION onward, so a written problem statement doesn't count — and a dead idea reads INVALIDATED instead of UNTESTED.

Two is deliberately low, and it is low for the same reason the VALIDATION gate is three rather than fifteen. This is not the number at which you understand your market. It is the number at which "this idea didn't survive testing" is a fact about the record rather than a flattering thing you said about yourself.

And it is computed rather than claimed, which is the part I care about. The builder writes whatever prose they like about why they stopped. Whether the idea was actually tested comes from proofs they had to earn.

A flaker asked to classify their own exit will always pick the flattering label.

That sentence is a comment in the model file, and it is the reason the reading is not a dropdown. Every one of us would tick pivoted.

Why the board won't call it a failure

There is a cohort board in Masterji that shows what a group of builders actually did, and a member between ideas shows up dimmed. Not struck through. Not drawn in the red the product reserves for a gate refusing. Closing an idea that did not survive contact is the product working, and a board that scored it as a failure would contradict everything else on the screen.

An INVALIDATED close is filed next to the wins, because it is one. You spent two conversations and bought a real answer, which is a better return than most weeks produce.

The closed record keeps everything too: the day-by-day check-ins, every proof you banked, how many came from talking to people. It is a page you can send someone. That is a deliberate inversion of how this usually feels — the abandoned project as something with a record attached rather than a folder you avoid opening.

What to do with the pile you already have

You cannot retrofit contact onto a dead project, and there is no point trying. But the pile does have one use, and it takes about twenty minutes.

Write the list. For each one, answer a single question: did anyone outside my head give this a verdict? Not did I show someone — did somebody who was not being kind to me act, pay, return, or refuse. Mark each project tested or untested. Do not grade them, do not write the reasons, and do not open the folders.

The count is the finding. If it comes back nine untested out of ten, then your problem was never idea selection and it was never discipline in the ordinary sense. It is that you have never once run the experiment, so every idea still looks equally plausible — which is exactly the state that makes the next new idea feel more attractive than the current one.

Then, on the next one, the only rule that matters: two conversations with people who do not owe you a favour, before the part where it stops being fun. Get the idea to a verdict that is not yours. Quit afterwards if the verdict says so — that is a good day's work, and the record will say what it was.

Masterji is free, and the walk-through needs no sign-in if you want to see what the record looks like before you keep one.