Remixed AI Is Still Transformational
“AI just remixes.” Good. Most valuable work is skilled remix.
There’s a TikTok critique out there which refers to a new article that basically says that generative AI only recombines what it’s seen, so it can’t be truly creative.
This is true in a narrow sense but irrelevant in practice. Google “only” indexed pages. Spreadsheets “only” recombined cells. Git “only” tracked diffs.
None of them invented knowledge from scratch; they changed how fast, how widely, and how consistently we could use what already existed. LLMs do that for language, logic, and routine decision-making. Dismiss it if you want; real operators will use it and ship more.
By the way this article is the first in a series where we will cover research papers and writing about AI itself, rather than its applications.
The complaint misunderstands the job
Most work isn’t a blank-page act of genius. It’s recombining known moves under constraints: customers, stack, timing, and taste. LLMs are combinatorial engines. That’s not a flaw; that’s a feature.
They bring scale (recall across patterns no single person carries), speed (drafts and options in seconds, which buys more iteration), and a better interface (natural language as an API that unlocks non-experts and multiplies experts).
It’s throughput and consistency at exactly the layer most teams are bottlenecked.
Where remixing already moves numbers
Engineering: Great engineers are artists, but most of the job isn’t art. It’s glue code, migrations, tests, and “teach me this library fast.” This is where LLMs shine. They spin up scaffolds and test harnesses in minutes instead of days, then suggest safer refactors while flagging edge cases and compatibility traps you’d otherwise discover the hard way. The payoff is focus: strong engineers spend their time on architecture and the gnarly bugs where originality actually matters.
GTM & Support: Marketing runs on recognizable beats. Models draft the baseline and tailor by segment; humans supply judgment and taste. In support, the move is to get a human on exceptions fast. Those exceptions are product truth in disguise, and they’re where loyalty forms.
Creative & Research: The win here isn’t divine inspiration; it’s structured divergence at scale. You can explore dozens of viable frames, tones, and constraints in minutes, then converge with human taste. What used to take a team days now lands in an afternoon and is done better because you saw more of the possibility space before you chose.
“But is it creative?”
If creativity means recombining known elements into something apt, timely, and useful, LLMs do that every day. If it means Einstein in a box, no. But we don’t need that to change outcomes. The practical split is simple: humans set direction, choose constraints, judge quality, and accept trade-offs; models provide high-throughput, high-recall recombination inside those fences. Once you work that way, quality climbs fast because you can iterate toward taste at a pace you couldn’t afford before.
You don’t need Einstein to come up with very, very powerful innovation. What made innovation happen in the time of the caveman wasn’t deep contemplation about the future, but instead was recognizing innovation in nature and adapting it. Is the wheel that different from a log rolling down a hill? What about the universe? Atoms and then molecules combining through random interaction ultimately, through natural selection, led to our world today.
Remixing and refining is creativity.
The price
Warning: The power of remixing comes with a price that isn’t measured in $ or electricity. When AI does your iterations for you, at mega scale and speed, part of your brain atrophies. The recent MIT study makes this obvious and follows the pattern where googling reduced the ability to recall facts. However, we believe this price is worth paying because it will enable transformations that our humble meat brains could never do alone.
The hype cycle
Six months ago, the loudest voices insisted GenAI would replace everyone. Now the backlash says it’s “just a remix,” therefore trivial.
That whiplash is classic Hype Cycle behavior: the rush to overpromise, then the equal rush to dismiss which skips the operational middle where value compounds. “Just a remix” is this season’s deflationary slogan. As quickly as people thought GenAI would replace everyone, they now want to be on the other side.
So human.
So wrong.
🎥 Check out our latest podcast on the ProductMind YouTube channel.
🎵 We are excited to announce we have expanded our podcast 🎙️to Spotify. Please give us a listen and if you like what you hear share with a friend, follow us, and (or) rate us 5 stars. ⭐️⭐️⭐️⭐️⭐️
Want to dive deeper?
Check out our book BUILDING ROCKETSHIPS 🚀 and continue this and other conversations in our 💬 ProductMind Slack community and our LinkedIn community.