AI First Workshop - Vienna, 14-15 Oct 2026

Your cheat sheet for Claude Code on this desktop.

claude, claude+, claude++

  • claude = plain Claude Code, asks before every action
  • claude+ = claude --dangerously-skip-permissions
  • Auto-approve, start with: claude+ "your task"
  • claude++ = claude --dangerously-skip-permissions --continue
  • Reopens your last conversation, e.g. after /exit to add an MCP server
  • Auto-approve is fine here: a disposable cloud computer

/model: brain and effort

  • /model = opens the model picker
  • Up/Down = pick a model, Left/Right = set the effort
  • Effort = how hard it thinks: low, medium, high, xhigh, max
  • Enter = switch and keep it for new sessions
  • s = this session only, Esc = cancel
  • Shortcut: /model opus or /model sonnet switches at once

Which model when

  • Sonnet 5.5 = your default here, medium effort, everyday work
  • Opus 5.5 = the hard tasks, twice the Sonnet price
  • Haiku 5.5 = quick bulk work, 1/20 of the Sonnet price
  • Fable 5.1 = never today, five times the Sonnet price
  • Stuck on a hard problem: /model opus, later /model sonnet
  • More thinking, same model: /effort high

Esc, Shift+Tab, Ctrl+C

  • Esc = stop Claude mid-answer, the work so far stays
  • Esc Esc = rewind to an earlier point, same as /rewind
  • Shift+Tab = next mode, e.g. Plan: it plans, you approve
  • Ctrl+C = stop; on an empty prompt press twice to quit
  • /exit = close Claude, claude++ brings it back
  • Wrong direction? Esc, then just say what you want