Descriptions:
Dylan Davis, who runs an AI consultancy, walks through one of the more underused features of both Claude and OpenAI Codex: the ability to schedule automated tasks that run without user involvement. The video covers Claude’s Co-work feature (available in the desktop app) and Codex’s Automations side by side, showing how nearly identical the two implementations are and why switching costs are low for anyone already using either platform.
The tutorial covers practical setup details — keeping your computer awake so scheduled runs don’t fail, how both apps self-heal by retroactively catching missed schedules when restarted, and how to configure cadence, model selection, and permission levels. Davis also demonstrates a manually-triggered variant: a task configured to run only when you press a button, which suits users who want AI to execute heavy lifting without granting full autonomy.
A key highlight is Davis’s three-question framework for deciding whether a task is worth automating at all: Does it need to run on a clock? Will the output drive a real action? Can it run autonomously without meaningful financial, legal, or reputational risk? He argues most people skip this evaluation entirely and build automations that execute reliably but deliver no value. The video concludes with guidance on scoping prompts tightly to reduce failure rates — practical advice for anyone setting up Claude or Codex workflows that need to run reliably in the background.
📺 Source: Dylan Davis · Published May 11, 2026
🏷️ Format: Tutorial Demo
