Codex to maintain a Korea trip AI brain
Yuta used a first overseas trip to Korea to try a travel workflow built around local files. Before leaving, Codex organized an AI brain containing an index, goals, current status, next actions, open questions, entities, outputs, logs, attachments and raw material. It also produced a PDF guide that could be consulted from a phone.
Original prompt
Shared by the source author. See source notes for context.
今日はかなり疲れたので、ゆっくり過ごしたい
Reach people exploring AI agents
See placements →What happened
A completed Korea trip supported by structured local records, a PDF guide and on-trip plan updates.
Reported by ゆーた (@yuta_mizukami)How the agent did it
- Create indexed trip files for goals, status, actions and unresolved questions.
- Export a phone-readable PDF guide.
- Update the plan with actual fatigue and transport-card questions.
Where a person stepped in
Yuta made every payment and final decision, and asked hotel staff about missing practical details.
Yuta used a first overseas trip to Korea to try a travel workflow built around local files. Before leaving, Codex organized an AI brain containing an index, goals, current status, next actions, open questions, entities, outputs, logs, attachments and raw material. It also produced a PDF guide that could be consulted from a phone.
Those files gave later questions a place to start. Instead of rebuilding the trip context in each conversation, the author could update the recorded plan and ask about the next decision. During the completed trip, the workflow supported T-money balance and top-up calculations. When Yuta was tired, a direct request for a slower day changed the plan.
Codex handled the structured preparation and records, while ChatGPT supported questions and image interpretation on location. The author deliberately kept all payments and final choices with the human traveller. This was a context and planning system, rather than an account of autonomous ticket buying.
The experiment also exposed omissions. Plug compatibility and toilet customs had not been covered well enough before departure, and hotel staff supplied missing practical information. A reusable version would keep unresolved questions visible alongside the itinerary. The source's useful contribution is its file structure and continuing updates, with the completed journey showing where local human help still mattered.
Prompt
Original prompt
Source: note.com · ゆーた (@yuta_mizukami)
Frequently asked questions
What was the AI brain?
A local folder structure separating goals, status, actions, questions, outputs, logs and raw material.
What changed during travel?
The author used T-money calculations and requested a slower day when tired.
Did it handle all practical questions?
No. Plug and toilet information was missing, and hotel staff helped.
More details
About this account
Codex organized local trip context and helped revise plans and calculations.
Firsthand source opened and read. Prompt is a verbatim excerpt of the published original instruction. The author's August 15 follow-up documents the AI-brain structure used for this same journey.
Credit: ゆーた (@yuta_mizukami)
Source published 14 Aug 2026
Source checked 14 Oct 2026
Tools and topics
Local Markdown files · PDF · ChatGPT
Korea trip · Travel notes · T-money
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.