Why I killed DearMeWiser.

DearMeWiser is shut down. It was an app I built mostly as a learning project, and the more honest I got about what it actually did, the harder it became to justify keeping it on the page.

It was a wrapper, and I knew it

The product was simple: ask a question, get an answer back as if it came from an older, wiser version of yourself. Reflection app, journaling angle, soft tone, friendly prompts. I learned a lot building it. I am glad I built it.

I am also clear-eyed that the entire user-facing experience could be reproduced inside ChatGPT with one well-written system prompt. "Answer me as a wiser, older version of myself." That is the whole product. Add a few interface touches, a reflection prompt or two, and you still arrive at the same place: a thin shell around a model call that any user can do for free, today, in a tool they already have open.

If a smart user can replicate your whole product in one prompt, you do not have a product. You have a UI on top of someone else's product.

Why "I learned a lot" is not a good reason to keep it shipping

The trap with learning projects is that the time you spent building them feels like a reason to keep them on the page. It is not. The page is a promise to a specific person about what they should expect from this brand. A thin wrapper that any reflective user can replicate in ChatGPT does not strengthen that promise. It quietly weakens it, because a visitor who tries it and goes "I could just do this in ChatGPT" now knows something uncomfortable about the rest of the lineup, too.

Killing it is the cleaner move. The learning stays. The shipped product does not need to.

The lesson worth keeping

Before shipping anything that wraps an LLM, ask: what does my product do that a single ChatGPT prompt cannot? Real answers look like proprietary data, real workflow integration, multi-step orchestration, opinionated UI for a specific job, or a context the model does not have. "Nicer onboarding" and "softer tone" are not real answers. If the honest answer is "nothing the user could not do with a prompt," ship the prompt as a free tool, do not ship a product.

This is going to be the test for every AI-flavored product I ship from here on. If the value lives in the wrapper, not in the work the wrapper unlocks, it does not get a product page.

See what survived the cut: What I'm building

Related postmortems: HardChoice · CompliQuiz · CampfireX