A new /wayfinder skill described by Latent Space focuses on one of the messier parts of AI-assisted development: deciding what to do when the path is unclear. Matt Pocock presents the skill as a tool for greenfield projects and planning situations with many possible next steps.
The idea is useful because agentic coding tools often perform best after the work has been narrowed into concrete tasks. Planning, by contrast, requires exploring constraints, choosing an approach, and making uncertainty explicit before code is written.
For developers, the practical promise is a more structured way to navigate early project decisions rather than asking a model to jump straight into implementation. The source frames it as a planning aid, not as a replacement for technical judgment or project ownership.