# OpenAI dot vs Claude Code routine (notes)

A companion to Build Alone Episode 06. Both run without you watching. They're built for different kinds of work.

## The test
Same job for both: every morning at 6 AM, find the AI stories that matter to people building on their own.
- **dot (named Scout):** posted 3 short stories at 6:00 sharp, each with a source. It asked two questions before starting (time zone, and whether it may post without asking). Two of its three stories on day two repeated day one.
- **Claude Code routine:** finished at 6:10 with 18 items added to a Notion database, and listed what broke (dead sources, blocked transcripts).

## Five differences
| | OpenAI dot | Claude Code routine |
|---|---|---|
| Starts | on its own (proactive research, read-only while it looks) | when you ask, or on a schedule / API call / GitHub event |
| Memory | remembers you; can't edit it line by line yet | blank every run; reads the files you write |
| Works in | its own cloud computer; ChatGPT, Slack, Teams | your project and files; other apps via connectors (MCP) |
| Price | from $100/mo (ChatGPT Pro); not in UK / EU / Switzerland yet | from $20/mo (Claude Pro); uses your normal limits |
| Control | Custom Rules per action; always asks for passwords and money | doesn't stop to ask; trim connectors before it runs |

## Set one up
- **dot:** ChatGPT desktop or web → create your dot → give it a name and a job. Check which plugins are already connected first: a new dot inherits them.
- **Claude Code routine:** in Claude Code type `/schedule` and describe the job, or create one at claude.ai/code/routines. Remove every connector the job doesn't need.

## Smallest key
- Check what's already connected before you start.
- One project folder, not your whole drive.
- Read-only if it only needs to read.
- Money and passwords stay with you.