Customers see your tunnel vision before you do

You cannot detect tunnel vision from inside the tunnel. The light at the end is right there, but you stopped looking up from the rails. I learned this last week when an early user caught two failures in my product that I had built, reviewed, and shipped.

What I shipped

An early user opened my product last week. Loud Camel is a tool that helps researchers get cited and recognized. The first paper it surfaced was attributed to the wrong author. The named researcher had not written that paper.

Then they flagged something heavier. The cold-email drafts the product writes for users imply the sender read the recipient’s paper. The sender did not. I built that flow. I reviewed those drafts. I shipped them anyway.

How I lost the star

When I started Loud Camel I told myself integrity was the north star. Every recommendation honest. Every email truthful. Then I spent four months deep in OpenAlex joins, email parsing, and pipeline plumbing. The star drifted out of my field of view. I was looking at the code.

This is the bug in founder cognition that scares me most. The thing I cared about the most became the thing I stopped checking. Not because I stopped caring. Because I stopped looking.

Why founders cannot audit themselves

I have tried the standard remedies. Weekly review of priorities. A pinned list of values on the wall. Asking myself whether I am building what I said I would build. None of it pulled me out. The frame you use to evaluate the work is the same frame that built the work. You cannot audit yourself from inside the tunnel.

What customers see that you cannot

The user who writes to say ‘this looks wrong’ pulls you out. The teammate who says ‘wait, are we sure?’ pulls you out. They see the product the way you wanted it seen. You see it the way you currently see it. The customer sees what you stopped seeing.

If you are building something, schedule the conversations that yank you back to the surface. Treat them as a check on whether you still recognize the product you wanted to make.

What I am changing

I am adding a validation step that confirms the attributed author actually appears in the paper’s author list before any recommendation is surfaced. I am rewriting the cold-email drafts so they do not pretend the sender read what the sender did not read. I am writing back to every user who flagged something and thanking them.

The next time the north star drifts, I want a user to notice before me. I would rather hear it from them at month four than ship past it for another four months alone.