Sabotaging projects by overthinking, scope creep, and structural diffing

3 hours ago

The author will attend Babashka Conf and Dutch Clojure Days in Amsterdam and invites connections.
Projects tend to go in one of two directions: immediate execution with minor revisions or overthinking due to prior art, leading to no progress.
A successful woodworking project with a friend focused on enjoyment and a simple shelf, avoiding overcomplication.
Research into semantic diff tools took 4 hours before realizing the goal was a personal Emacs workflow, prompting a return to minimal scope.
Long-term interests like language design and CAD tools remain stuck in overthinking, lacking clear success criteria.
The author advocates embracing action over excessive consideration, learning from past experiences like YAGNI principles.
An attempt to build a fuzzy path search in Emacs with an LLM led to unnecessary feature creep, but was eventually simplified.
A review of structural diffing tools highlights limitations in line-based diffs and tools like difftastic, with notes on various alternatives.
The desired workflow involves semantic tree diffing for reviewing LLM-generated code changes in Emacs, akin to Magit's entity-level staging.
Plans include building a prototype using treesitter, diffast, and dynamic programming, with no pressure to commercialize.

Hasty Briefsbeta