Experiments

An AI Tool That Summarizes Complex Documents

Wrapped an LLM API into a simple web interface. Had a real use case in mind. Discovered that the use case was mostly solved by ChatGPT already.

What I Built

A web app that accepted a document upload and returned a plain-language explanation. The target user was someone who receives a lot of dense professional documents and doesn’t want to read them.

The Approach

Used a serverless platform with an AI API on the backend. The frontend was minimal — just a drag-and-drop zone and a text output area. Total build time was maybe two weekends.

What Happened

Abandoned. The core problem: by the time I shipped it, ChatGPT’s file upload feature worked fine for this exact use case. My version had no meaningful differentiation.

What I Learned

  • “I could build that” and “I should build that” are different questions
  • Competing with ChatGPT on general tasks is a losing proposition
  • The underlying idea (plain-language document explanation) is still good — the vertical matters
  • Distribution was never thought through; I was building a solution without a clear path to users