Future topics

Speculative or deferred work. None of this requires reinventing Nushell.

Committed product plan (not speculative): stale PATH/env refresh via planned env-refresh + shell-host. See Shell Host and Environment Refresh. The sections below remain research / deferred topics.

Have your stdout and eat it too

TUIs often abandon stdout for layout control. Research directions:

  • Shell-owned layout with structured payloads (Nushell path)

  • Hidden machine-readable stream alongside human view (agent/automation)

  • Terminal protocols that reflow without the child process redrawing

Interactive TUIs and machine-readable pipelines often fight for the same fd. Research patterns where tools can offer both: structured data for scripts, rich UI when a human is present — without lying about which mode is active.

Display vs transport decoupling

Push further on separating "what the pipeline carries" from "what the terminal shows." Structured shells already help; terminal and protocol work can reinforce the split.

Terminal protocol notes

Modern terminals (Ghostty, Kitty, WezTerm, and peers) expose richer protocols than classic VT. Document how those interact with structured stdout, images, and progress without collapsing display back into the data channel.

Mixed fonts (speculative)

Explore whether shell or terminal protocol can mark spans with font roles (for example monospace vs sans) in one output stream for readability—without breaking pipes. Likely needs terminal-emulator cooperation; fish/Nushell hacks alone are insufficient long-term. Exploratory only — no implementation planned here.

Upstream contribution

Prefer patches and plugins for Nushell / terminals over greenfield shells unless a concrete gap cannot be fixed upstream.