Everything in this pillar so far describes things that have already happened. This one doesn't, and it would be dishonest to present it in the same voice.
So: this is an argument, with its premises stated and its failure conditions named. Not a forecast, and definitely not a set of predictions with confident numbers attached: the whole point of this library is that we don't do that.
Read it as "here's what follows if these things are true", and check the premises yourself.
The premises
Four, all of them observable today rather than projected:
- Producing a competent asset keeps getting cheaper. Direction of travel over the past few years, no sign of reversing.
- Attention stays fixed. Twenty-four hours in a day, no mechanism by which that changes.
- Self-asserted credentials keep losing value as they get cheaper to produce.
- Accountability can't be automated. It requires someone who can be held to a thing.
If you disagree with any of those, the rest doesn't follow, and that's a legitimate place to get off.
What follows
1 · Selection infrastructure becomes the valuable layer
If supply is unlimited and attention is fixed, whatever filters becomes the powerful thing. That's already true of platform algorithms; the argument is that it becomes true one level up: in the tools brands use to decide what to make at all.
The value moves from tools that help you produce to tools that help you choose and verify. Production tooling trends towards commodity pricing because it's competing with something approaching free.
What would falsify this: if selection turns out to be automatable to a high standard, the value moves again: to whoever owns the data the selection runs on. Watch whether "which of these forty should we make" becomes a solved problem or stays a judgement call.
2 · Verified reputation becomes portable and standard
If self-asserted credentials keep devaluing, the market needs a substitute, and the only candidate is evidence a third party generated: completed, paid, rated work.
The argument is that this gets standardised and portable, the way credit ratings did. Not that any particular platform wins, but that "show me verifiable delivery history" becomes a normal thing to ask a freelancer, in the way that "show me your portfolio" is now.
What would falsify this: if verification systems get gamed at scale, they'd devalue just like the thing they replaced. The defence is gating on real money moving. But gating is only as good as its weakest link.
3 · The middle of the market thins
Two ends stay healthy: commodity (cheap, fast, automated or near-automated) and specialist (access, judgement, accountability, hard to substitute).
The middle (competent generalist execution at a mid rate) gets squeezed from both sides. This is a familiar pattern in markets where a production input collapses in cost, and there's no obvious reason content would be exempt.
What would falsify this: a durable quality gap that keeps competent-human execution clearly above automated output for long enough that the middle re-forms around it.
This is the argument behind specialise or disappear and pricing when production got cheaper.
4 · Provenance becomes a feature people ask for
If generated content is discounted on sight, being able to demonstrate what was made how becomes commercially useful: not for ethical reasons, for pricing reasons.
Expect this to appear first where the stakes are highest: regulated claims, journalism, anything where being wrong is expensive.
What would falsify this: audiences habituating. Discounts fade: people once distrusted stock photography and now barely notice it. It's entirely possible generated content simply becomes the wallpaper and stops triggering anything.
That last one is the premise I'd bet against most readily, and I'd rather say so than write it as though it were settled.
What this doesn't predict
Being explicit, because vagueness here is how forecasting articles avoid ever being wrong:
Not a timeline. "Eventually" is doing real work in all four. Structural shifts in how work gets found and priced take years and move unevenly by market.
Not that any particular tool wins. Including ours. Arguments about market structure are not arguments about vendors.
Not that human creative work declines. The argument is that its composition shifts: towards access, judgement and accountability, away from execution hours. That's a different claim and a more defensible one.
Not with numbers. Anyone giving you a percentage for how much of content will be AI-generated by a given year is producing a plausible-sounding figure. That's exactly the failure mode this library exists to avoid.
What to do with an argument like this
Not act on it directly. Arguments about where things are going are worth roughly what they cost, and this one cost you eight minutes.
What they're useful for is checking your assumptions against your own evidence. Look at your own last two years:
- Has the mix of work you're asked for shifted towards or away from execution hours?
- Are clients asking for verification they didn't ask for before?
- Has the middle of your own market got harder?
Your answers are better data than anyone's argument, including this one. If they contradict the premises above, trust yours.
Where Acumin fits, and the honest disclosure
Acumin is a bet on premises 1, 2 and 4, and I built it, so treat this section accordingly.
The product is pointed at selection and measurement rather than generation: a bet that when production is cheap the scarce thing is knowing what to make and whether it worked. It can't invent a figure about your performance: a bet that verifiable analysis outlasts abundant analysis. And the Creator Network gates ratings on real, paid, delivered bookings: a bet that verified reputation beats self-asserted reputation.
Those bets could be wrong. If selection becomes fully automatable, if audiences stop discounting generated work, if verification gets gamed, the design decisions look over-cautious rather than principled.
I'd rather state that plainly than write a piece implying the future is settled and we happen to be standing in it. The premises are checkable, the falsification conditions are named, and you can watch them yourself.
How to use this tomorrow
Pick the premise you're least sure about and write down what you'd expect to observe over the next year if it were true, and what you'd expect if it weren't.
Then actually check in a year. That's the only way anyone finds out whether an argument like this was worth reading.
Related: What doesn't change is the durable half, and the safer thing to build on. How brands find creators now is the part of this that's already happened.