How do you decide which projects to keep and which to let go?
I keep seeing builders end up with several things in motion at once.
Side projects, experiments, internal tools, ideas that are half built.
It’s rarely about motivation or skill.
Deciding what to stop often feels harder than deciding what to start.
For those who’ve been through this,
how do you decide when to double down versus walk away?
I recognize this pattern way too well. For me the hardest part wasn’t starting things, it was admitting when something hadn’t earned more time yet.
What eventually helped was separating progress from movement. Lots of activity can look like momentum, but unless something external is changing, users behaving differently, clearer demand, actual constraints, it’s hard to justify doubling down.
That tension is part of why I started thinking about validation as an ongoing signal instead of a one-time decision.
How do you personally recognize the moment when continuing stops being the rational choice?
What helped me was separating progress from signal.
If a project shows real external signal (users responding, feedback changing direction, even small commitments), I double down.
If it’s just internal momentum (shipping, tweaking, polishing) with no pull, I pause or cut it.
I try to time-box validation and let evidence decide, not attachment.
This framing is super helpful — especially the idea of time-boxing validation and letting evidence decide. Curious if you’ve ever regretted cutting something too early?
Yes — a few times. In hindsight, the regret wasn’t about killing the project, but about why I killed it.
The ones I regretted cutting early usually had weak demand but strong learning loops — users were few, but their behavior was changing how I thought about the problem.
What helped later was adding a second filter to time-boxing:
not just “is there pull?” but “is new information still emerging?”
Curious — when you look back at cuts you regret (if any), was it lack of signal… or misreading the type of signal you were getting?