This content originally appeared on DEV Community and was authored by Dejan S. Višekruna
A personal memory retrieval turned technical challenge — and how ChatGPT helped solve it.
🎯 The Challenge
I had been searching for two songs for decades. One was a dreamy eurobeat track, the other a dance cover with a female vocal.
I had no artist, no title, no recording — only a feeling, a rough memory of the beat, and the fact that I heard each one only once on the radio in Belgrade in the early ’90s.
🧠 Using ChatGPT as a Research Tool
Instead of asking “What song is this?”, I approached ChatGPT like a research assistant.
Here’s what I gave it:
- Vocal style: Female
- Era clues: One was from the late ’80s/early ’90s, the other from the mid-’90s
- Beat style: One had a groove similar to Back to Life by Soul II Soul; the other leaned more toward eurobeat
- Context: Both songs were aired only once on Belgrade radio stations during a time of severe UN sanctions
- Data issue: One track was misidentified as “Three Times” due to a faulty RDS display
- Memory anchor: I remembered a single chorus line from one of them
I turned that into a structured query — and then refined it through several iterations.
✅ The Result
After a series of narrowing-down steps, ChatGPT helped identify two songs I never expected to find:
- “Dreamtime” – Zee (misread on radio as “Three Times”)
- “Dream” – W.I.P. (a 1991 dance cover of “All I Have to Do Is Dream”)
🔍 Prompting Breakdown
- Structured inputs: genre, rhythm, era, vocal type, source context
- Prompt chaining: refining results based on emotional match and memory fragments
- Rejection criteria: wrong mood, wrong lyrics, wrong production style
- AI cross-referenced musical style, RDS errors, historical distribution
🤔 Why This Matters
LLMs aren’t just for coding, summarizing, or rewriting emails.
They can act as context engines — helping us reconnect with pieces of our lives that search engines left behind.
💬 Final Thoughts
This wasn’t just about finding songs. It was about proving how human memory + AI reasoning can solve a decades-old emotional puzzle.
And to write it using the very same ChatGPT that helped me find those songs in the first place — that was part of the closure.
Because this isn’t just my story. It’s a reminder that humans + AI can bring back something deeply personal — something even search engines forgot.
Ever used AI to find a lost piece of your past? Let me know in the comments.
AI-generated by ChatGPT for illustration purposes. No memories (or animals) were harmed during the making of this post.
This content originally appeared on DEV Community and was authored by Dejan S. Višekruna
Dejan S. Višekruna | Sciencx (2025-06-07T22:30:26+00:00) How I Used ChatGPT to Identify Two Songs I Heard Once — Over 30 Years Ago. Retrieved from https://www.scien.cx/2025/06/07/how-i-used-chatgpt-to-identify-two-songs-i-heard-once-over-30-years-ago/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.