8 Indie Hackers posts later: my AI regex tool still has basically no strangers

I wouldn’t pause everything to “fix activation” yet.

Your numbers say something narrower: Indie Hackers has not produced enough strangers for activation to be the diagnosed bottleneck.

Zero paying users and weak return behavior can look like an activation problem, but if almost nobody new reaches the first generation, there isn’t really an activation cohort to study.

I’d stop treating the every-3-days IH cadence as the acquisition experiment. Keep it at a low maintenance frequency if the writing habit is useful, but run one controlled acquisition test somewhere developers are already dealing with regex pain.

Then measure only:

qualified visitor → first generation → edits/tests the regex → saves/shares → returns

That gives you a clean failure point.

If they never generate, fix targeting or the promise.

If they generate but don’t verify/edit, your confidence feature may be the right next build.

If they verify and still never return, the problem is recurring value, not onboarding.

I would not build must-match / must-not yet purely because the current cohort does not return.

The one number I’d want before deciding: since your last snapshot, how many unique people other than you have actually completed at least one AI generation?

Keep the cadence, but stop counting posts and start counting the room.

I read our own numbers at three days this morning. We had left replies in four threads carrying 196 to 617 views each. Every reply landed directly under the original post, none of them folded or hidden. Views on our own replies: 2, 3 and 11.

The one that got 11 was the only thread where the author had actually asked people to send him something he could try. It earned a reply and a like. The other three answered a general question and vanished.

So the split is not cadence against activation. It is whether the person on the other end has said out loud that they want something to test. Eight posts into a room where nobody asked reads the same as one.

Your own numbers already say this. Comments on pricing are people being polite about a thing they have not used. A stranger generating three times in a row is the only row in that table that means anything, and you have none yet.

What I would do: drop to weekly so you are not silent, ship the must-match examples, then spend the saved time in places where someone has explicitly asked for a tool to try. That is a smaller number of posts into much warmer rooms.

One question back, because it is the only cheap data you have. Of the 24 cached prompts, how many are not yours and did not come from here? Whatever referrer those came from is the only room you have proof about.

For context on where the numbers above come from, I build AI Applyd, which applies to jobs for people, and I have the same problem you do. Writing is easy, being wanted is not.

Fix activation first, but don't go fully silent — cut the cadence to once a week and ship the must-match examples. Comments on pricing and positioning are polite feedback; a stranger generating three times in a row is the real signal. I spent 150 euros on ads once and got 4 clicks, 0 installs — same lesson: nothing downstream matters until the core loop delivers value on first touch. One honest update when the fix ships will read stronger than eight posts about the problem.