The problems worth building for right now

A problem being unsolved is not the same as a problem being worth solving. Some things stay unsolved because nobody profitable actually has them. Others stay unsolved because every attempt so far made the same avoidable mistake. Telling those apart is most of the work.

Ask who was already paying

The strongest version of "this is a real problem" isn’t a Reddit thread of people complaining — it’s a dead product that had paying customers before it shut down. Complaints are free. A credit card is a much harder signal to fake, and it means the problem cleared a bar that pure opinion never has to clear.

Look for the same failure mode repeating

When you can find more than one attempt at roughly the same problem, and they died for similar-sounding reasons — too expensive to acquire customers, too manual to deliver, too narrow a wedge — that repetition is informative. It usually means the problem is real, but the standard approach to it is broken. That’s a much better starting point than a problem nobody has attempted at all, which is often unattempted for a reason nobody’s told you yet.

Cheap to validate, expensive to fake

The best problems to build for right now are the ones you can test small — a landing page, a manual concierge version, a waitlist with a real reason to join — before you build the automated, scaled version. If a problem can only be validated by building the whole thing first, that’s usually a sign to shrink the bet, not skip the validation.