World Travel Memories

Inspiration

I have lived as a perpetual traveler for the last six years, usually spending about 84 days in each city. During that time, I accumulated thousands of photos and videos organized across 32 city folders in Google Drive.

Posting those memories manually became exhausting. Every post required me to choose a file, remember the story behind it, write a caption, and schedule it. Eventually, the effort caused me to stop posting altogether.

I wanted my Instagram account to work more like an automatic travel slideshow, allowing me to share my complete archive without having to make a new creative decision five times every day.

What it does

World Travel Memories automatically publishes five travel photos or videos per day.

The automation cycles through all 32 active city folders, giving every city one turn before returning to the beginning. Within each city, it prioritizes media that has never been posted. A city only begins reusing older media after all of its unique eligible files have been used.

The name of the Google Drive folder automatically becomes the Instagram caption, showing the location and the dates when I lived there.

How I built it

I am a nontechnical builder, so I worked with GPT 5.6 to design the workflow and write the original Google Apps Script.

The workflow is:

Google Drive → Google Apps Script → Cloudinary → Buffer → Instagram

Google Apps Script selects the next eligible file from Google Drive. Cloudinary makes the media accessible to Buffer. Buffer schedules and publishes the final Instagram post.

A Google Sheet records every attempt, including the city, file name, media type, Buffer post identification number, status, and scheduling details.

The Build Week extension

During OpenAI Build Week, the automation unexpectedly stopped posting.

I used Codex to inspect the existing live project rather than rebuilding it. Codex discovered that an oversized video exceeded Google Apps Script’s file processing limit. Because the original error was not caught, the scheduler retried the same file every five minutes and never advanced to the next city.

Codex repaired the automation by adding file size checks, failure logging, and recovery logic. It also preserved the existing credentials, folder structure, posting history, schedule, captions, and rotation state.

After the repair, Codex tested both photo and video drafts, verified the active trigger, and restored live posting.

Challenges

The most difficult challenge was that the automation appeared to remain enabled, but no useful failure was recorded. One oversized video silently prevented the entire system from progressing.

The repair required identifying the exact failed file, protecting the scheduler from unsupported files, and ensuring that one failure could never stop all 32 cities again.

What I learned

I learned that a working automation also needs recovery behavior. External services, file limits, and unexpected media can fail, so the system must record errors and continue safely.

I also learned that a nontechnical person can use GPT 5.6 and Codex to create, diagnose, and maintain a real automation that solves a personal problem.

What is next

A future version could automatically compress oversized videos instead of skipping them. I would also like to add a simple dashboard showing folder progress, unused media totals, failed files, and upcoming cities.

Built With

• automation
• buffer
• cloudinary
• google
• gpt-5.6
• instagram
• javascript
• openai