We built and released four remixes with FADR because I wanted to test something for myself: how would distributors, streaming platforms, copyright systems, playlist curators, and listeners treat music made partially with A.I.?

The experiment involved four specific releases:

All four remixes have now been taken down. The original songs were not the A.I. experiment. These four remix versions were.

What we were testing

The vocals and lyrics came from our existing human-made songs. FADR generated new musical backing, and I edited the beats it supplied. That made the remixes partially A.I.-generated: the source performances remained ours, but the tool handled a major part of the new arrangement and composition.

Would a distributor flag reused or copyrighted material? Would streaming services treat a release differently after it was labeled partially A.I.? Would listeners notice? Because many DJs and producers use the same tools, would generated material collide with music somebody else had already uploaded?

Easy to use—and that was part of the problem

FADR made the process extremely easy. Drop in a song, choose a direction, and receive remix options. But as I tried it with different songs, I began hearing the exact same loops again. The results felt less like completely original generation and more like a system serving a pool of prepared musical pieces across different keys and styles.

I cannot prove how FADR constructs every result from the outside. I can only describe what I heard: distinct inputs sometimes produced unmistakably familiar loops. That raises an important question. How much is created uniquely for the user, and how much is selected, rearranged, or transformed from a shared library?

What happened at distribution

I disclosed the remixes as partially made with A.I. when distributing them through LANDR. I expected at least one to be flagged. LANDR’s review systems have seemed sensitive to copyrighted work, stock loops, and potential conflicts, and I thought another producer might already have released music containing the same FADR material.

None of the four was rejected for a detected conflict.

That could mean the material had not appeared in another scanned release, the systems did not identify an overlap, or disclosed partial-A.I. material is handled differently. This experiment cannot tell us which explanation is correct. It does show that checking a “partially A.I.” box did not create a meaningful visible barrier in this case.

Streaming treated the songs normally

The streaming platforms appeared to treat the remixes like any other release. I submitted some to playlists, and “Blueberries (WUUYUH Remix)” received a placement fairly quickly. From my side, the disclosure felt like metadata collected during upload rather than a classification that changed how the music was presented or evaluated.

A few listeners did notice and asked, “Now you are using A.I. in your music?” I was glad they caught it. Their questions proved transparency can matter when the information is visible. But a vague partial-A.I. label still leaves people guessing about what the machine actually did.

“Partially made with A.I.” is too broad

Nearly every producer I know uses tools that include machine learning or A.I.-assisted processing. Mastering software, intelligent EQ, low-end balancing, compression, noise removal, vocal cleanup, source separation, and frequency analysis can all use A.I. These tools may identify a problem or suggest a starting point, but they do not magically finish a record. A trained ear still decides what the song needs and takes responsibility for the result.

Sound design creates another gray area. A producer might ask a tool connected to a synthesizer for a deep growling bass or a lush pad, then reshape that starting point. That can resemble browsing a huge preset library. The musician still writes the part, develops the melody, performs instruments, programs drums, arranges the song, records vocals, and makes thousands of creative decisions.

That is fundamentally different from placing a completed song into a service, requesting a remix, and letting the service generate the creative foundation. In our experiment, A.I. was not merely detecting frequencies or cleaning noise. It performed a substantial share of the compositional and arrangement work.

We need better disclosure

Distributors and streaming services should distinguish fully A.I.-generated music, A.I.-generated vocals, composition, backing music or remixes, A.I.-assisted sound design, and A.I.-assisted editing, restoration, mixing or mastering. Those categories would stop a producer using an intelligent EQ from being placed in the same bucket as someone who typed a prompt and accepted a completed song.

Why I ran the experiment

I am relentlessly curious. I do not want to form my opinion entirely from headlines, marketing, or people who have never touched the tools they discuss. I work in technology and use A.I. professionally. I understand how useful it can be, and I understand why we must be honest about its risks. I wanted firsthand knowledge of the process, disclosure, checks, platform response, and listener reaction.

Why we took the remixes down

I have spent decades learning to produce, record, DJ, play instruments, write songs, and engineer audio. The part I love is making the music. Letting a generator take over the central creative work did not feel like an extension of that practice. It felt like stepping away from it.

That is why “Blueberries (WUUYUH Remix),” “Rearview (WUUYUH Remix),” “Working (Bear Cole Remix),” and “Forget Me (Turntable Kachina Remix)” are no longer available. We learned what we wanted to learn, and keeping them online no longer represented what 1st Drop Music stands for.

Our position is straightforward: we do not use A.I. to generate our music. We may encounter A.I.-assisted technology inside modern production tools, but we do not hand the creative center of a song to a generator and present the result as if we composed it.

The artwork became part of the lesson

It is fitting that the artwork for these remixes was also made with A.I. I have used generated imagery in parts of the 1st Drop Music project, but I do not claim that makes me a visual artist or illustrator. The tool made the images.

This experiment made me slow down there too. A.I. makes endless content possible, but endless output is not automatically meaningful. My takeaway is to use it sparingly, disclose it honestly, and protect the parts of the process where human expression is the point. If a generated image helps tell one story, make the next thing yourself. Choose intention over volume.

The Trojan horse

The danger is not one novelty song or one experimental remix. The danger is allowing so much generated material onto platforms that human-made music becomes harder to find, value, and sustain.

A prompt can produce an audio file. That does not make the act equivalent to writing lyrics, developing harmony, building rhythm, playing instruments, singing, arranging, producing, and revising a song until it says what the artist needs it to say.

What remains

After this project, we published our A.I. stance across the 1st Drop Music websites. We want music to live forever as human-made emotion through sound—not as an ocean of generated noise.

The best response is still to create. If you are a musician, keep making music. If you are an artist, keep making art. Learn the tools, question the claims, test what you need to test, and be honest about what you used. Do not surrender the creative work that made you care in the first place.

That is what this experiment clarified for me. I do not want A.I. to make music for me. I love making it myself.