Overthinking as sabotage: when scope creep kills personal projects

Kevin Lynagh argues that personal projects fall into two camps: ones he just builds and finishes, and ones he sabotages by surveying prior art, expanding scope, and never shipping. The decisive variable is how clearly he has internalized his own success criteria. A weekend kitchen shelf worked because the goal was simply to do woodworking with a friend; long-running interests like a Clojure/Rust hybrid language or a CAD tool stall because the criteria stay fuzzy and the inner critic silences the generative impulse to just build something.

He extends the same lesson to LLM-assisted coding. Rebuilding a Finda-style fuzzy path search, he let an agent pull in the Nucleo library and chase anchor-matching semantics for path segments, burning hours before realizing he had never actually wanted that feature. He posits a possible conservation law: gains in coding speed from LLMs get spent on unnecessary features and rabbit holes. The piece closes by pivoting into notes on structural/semantic diff tools, a four-hour weekend research detour he caught himself on by recommitting to the original minimal scope — a nicer Emacs diff workflow he could just write himself.