I didn’t fire anyone. I just never had a reason to hire them.
A couple of weeks ago, I rebuilt and relaunched luisramirez.org and the Cast Iron LA site. No developer Slack thread. No implementation handoff. No SEO consultant. No project manager keeping the pieces moving.
The work got done. The invoices didn’t.
That’s what quiet demand erosion looks like.AI is not replacing jobs in the way most people mean when they say that. It is intercepting the moments that used to become jobs. The contract never gets signed. The freelancer never gets hired. The service call never gets made.
In biology, autophagy is the process by which a cell breaks down its own components to keep functioning. It recycles what it no longer needs in order to survive. The parts it digests do not come back.
That is what is happening to the economy right now. And AI is the enzyme.
AI is making individuals and businesses more capable. It is also reducing the need to pay other people to do work that once sat just outside their reach. The economy is becoming more efficient by consuming demand that used to support working people.AI Isn’t Just Replacing Jobs. It’s Reducing Demand for Work
A year ago, those two site relaunches likely would have involved a front end developer, a back end developer, an SEO specialist, a writer, a strategist, and probably some level of project management.
This time, I did it with AI.
Not as a stunt. Not to prove a point. Because it worked.
That distinction matters. Nobody at my agency got fired. But several categories of paid work never materialized. The sites still launched. The money just stayed in our account instead of moving through the small network of specialists who would have touched the project before.
And I am one creative director at a small agency. Multiply that across freelancers, solo operators, founders, in-house teams, and small businesses everywhere, and the pattern gets bigger fast.AI Is Compressing Entire Layers of the Economy
The obvious assumption is that this only affects creative or digital work. It does not.
The first time we turned on our air conditioner this year, it did not start. I used ChatGPT to diagnose the problem from a photo, confirm it was a blown fuse, identify the replacement part, and verify it at the store before installing it myself.
No HVAC tech was called.
That does not mean HVAC professionals are obsolete. Complex repairs still need trained people. But one paid service call disappeared. A few hundred dollars that might have entered the local economy stayed with me instead.
That is the pattern.
AI does not need to wipe out a profession to damage demand. It only needs to intercept enough of the routine, once-billable moments that used to add up to a living.
And every time that happens, the work may still get done, but the economic transaction disappears.AI Is Making Tools and Services Optional
When I moved both sites off WordPress, I did not just switch platforms. I cut out an entire paid layer of the web ecosystem that used to sit between me and the finished product.
Elementor Pro. Schema Pro. Yoast Premium. ManageWP. WP Rocket. Premium hosting.
It’s not that a better tool came along. It’s that the tools were no longer necessary.
Not because those products stopped working. And not because they suddenly became bad. AI changed what I could do on my own. I could write structured data myself. I could troubleshoot implementation issues myself. I could build components, refine copy, solve technical problems, and ship without depending on the same plugin and service layer I used to need.
That is what makes this bigger than a software competition story. In a lot of cases, AI is not replacing one product with a better product. It is reducing the need for the category itself.The Paradox of New Capability
Here is where the story gets more complicated.
I probably would not have relaunched either site without AI. The time and cost would have made it hard to justify. Both would have stayed older, clunkier, and more compromised for longer.
I also built a pour-over coffee timer for myself. A custom brew calculator that walks me through ratios and pour timing exactly the way I like to make coffee in the morning.
That tool did not take a developer’s job. No one was ever getting paid to build it. It only exists because the barrier to building it collapsed.
That is part of the paradox. AI is destroying some demand while creating new capability.
But the fact that new things can now be made does not cancel out the economic shift. A thousand people building useful little tools for themselves still means a thousand projects that never turn into paid work for somebody else.
That is the trade.Efficiency Becomes a Demand Problem
When businesses and individuals become efficient enough to stop hiring, the people they did not hire lose income. And people without income do not buy products, subscribe to services, or spend money back into the economy.
Every dollar saved on a contractor is a dollar that contractor does not spend somewhere else.
Every freelancer who loses a project loses the ability to hire, buy, invest, or circulate money in other parts of the same system.
That is why this is bigger than a labor story. It is a demand story.
An economy cannot keep reducing the number of people who get paid and assume consumer demand will remain untouched. At some point, efficiency starts feeding on the customer base itself.There Is Also an Echo Chamber Version of This Collapse
The economic problem is not the only one.
If AI increasingly generates the content, design, copy, code, strategy, and media that fill the internet, then more and more of what it trains on will also be synthetic. Not reality filtered through human experience, but machine output becoming future machine input.
That is how you get an echo chamber. A closed loop where the system keeps feeding on its own residue.
Anyone who grew up copying cassette tapes knows what happens next. The first copy is close enough. The second is still usable. By the third or fourth generation, the signal-to-noise ratio gets progressively worse. The music is still there, but the hiss gets louder. More of what remains is artifact and less of it is signal. Keep copying long enough and what you’re left with is mostly noise pretending to be music.
That is what worries me about an AI-saturated culture.
The output may still look polished and sound coherent. But coherence is not the same thing as depth. You can end up with content that has the shape of meaning without much contact with reality. Researchers have a term for this: model collapse, where systems trained on synthetic output begin to lose fidelity over time, producing work that is narrower, more repetitive, and less grounded.
And that matters economically too. Markets depend on novelty, friction, taste, obsession, weirdness, and the uncomfortable human urge to make something that does not yet fit. If everything becomes optimization stacked on optimization, the culture gets flatter. And flat culture creates weaker demand than original culture does.The Skills That Still Matter Are Not the Cheap Ones
Execution gets cheaper. Judgment gets more valuable.
The people most exposed are the ones whose value sat mainly in producing the thing. The people in the strongest position are the ones who know what should be made, why it matters, what to cut, what to ignore, and when the output is polished but strategically empty.
That is where the leverage moves.
The coder solving architecture problems is safer than the one just pushing markup. The strategist with real business judgment is safer than the person running a checklist. The writer with an actual point of view is safer than the person filling space.
AI is a force multiplier for people who can direct, decide, and verify.
It is a pricing problem for people whose value lived mostly in execution.If AI Absorbs Junior Work, Where Does Experience Come From?
This may be the most important question in the whole conversation.
Every senior creative, strategist, developer, and operator got that way by doing junior work first. The production tasks. The boring reps. The grunt work. The first drafts. The low-stakes assignments that built pattern recognition over time.
That pipeline matters.
If AI absorbs more of the entry-level layer, where does the next generation of experienced people come from?
How do you become senior if you never get to be junior?
The same companies using AI to eliminate early-stage labor will still want high-level judgment later. But judgment does not appear out of nowhere. It is built through repetition, failure, correction, and time.
If the economy digests the learning layer, it does not just remove jobs. It removes the process that creates future expertise.
That is why this feels bigger than automation in the usual sense. We are not just making work cheaper. We may be cutting away the developmental path that made skilled humans in the first place.The Real Warning Sign Is Not Layoffs
It is the missing invoice.
It is the specialist who never gets brought in. The junior person who never gets the first shot. The consultant who never gets the call. The local service job that gets solved from a photo. The software subscription that no longer feels necessary. The task that moves from “hire” to “handle.”
That is how a labor market thins out before the headlines catch up.
AI is making people more capable. That part is real. I am using it myself. I am benefiting from it myself. It is helping me build faster, think faster, and ship work that would have been slower and more expensive before.
But the gains are not free.
If enough work gets absorbed by fewer people, the economy that celebrated efficiency starts starving the base it depends on. That is the real risk. Not that AI suddenly replaces everyone. That, one missing invoice at a time, the economy learns how to eat itself.
This is one of the clearest descriptions yet of how AI changes the economy without ever showing up as a layoff.
You’re not talking about jobs lost in the visible sense. You’re talking about jobs that never crystallize—the missing invoice, the project that never leaves someone’s head, the specialist who never gets the email. That’s a different kind of disruption, and it’s exactly why traditional labor stats and tech narratives feel out of sync with what practitioners are seeing.
A few core ideas you surface that are worth underlining:
AI Isn’t Just Replacing Jobs. It’s Reducing Demand for Work
You frame AI not as a robot taking a seat from a worker, but as an enzyme digesting the need for that seat in the first place.
- Two site relaunches that used to justify a small constellation of specialists now stay inside your own capability set.
- The HVAC example is perfect: the system still gets fixed, but the transaction disappears.
- The WordPress ecosystem example is even more telling: it’s not that a better plugin beat Yoast or Elementor; it’s that the category itself became optional once you could do the work directly with AI.
This is what makes the current wave different from previous automation stories. It’s not just substitution (X replaced by Y). It’s compression—entire layers of the value chain collapsing into a single, more capable node.
AI Is Compressing Entire Layers of the Economy
The autophagy metaphor is doing a lot of work here, and it’s doing it well.
- The economy is not being attacked from the outside; it’s digesting its own components to stay lean and competitive.
- The parts that get broken down don’t come back as what they were. They return as generalized capability, not as discrete livelihoods.
That’s a more honest framing than the usual “creative destruction” story, because it highlights the irreversibility of some of these shifts. Once a business learns to do something in-house with AI, that external demand doesn’t just “bounce back” when the hype cools.
AI Is Making Tools and Services Optional
You capture a key paradox:
- Without AI, those site relaunches probably wouldn’t have happened at all.
- Your brew timer is a net-new artifact that would never have justified hiring anyone.
So AI is expanding what gets made while shrinking what gets paid for.
That’s the subtle but crucial distinction: more things exist, fewer of them are attached to income streams for other people. A thousand micro-tools built by individuals are a thousand tiny acts of self-service that used to be the raw material for someone else’s business model.
In macro terms, that’s a demand problem masquerading as a productivity win.
The Paradox of New Capability
You connect the dots that most efficiency narratives skip:
Every dollar saved on a contractor is a dollar that contractor does not spend somewhere else.
If enough actors optimize this way:
- Fewer people get paid.
- Fewer people can buy.
- The same system that celebrates efficiency starts starving its own customer base.
That’s the part that rarely makes it into board decks. You can’t keep compounding efficiency gains while assuming aggregate demand stays magically constant. At scale, cost savings are also income reductions for someone else.
There Is Also an Echo Chamber Version of This Collapse
You’re also pointing at a cultural and epistemic version of the same autophagy:
- As AI-generated content becomes training data for future AI, the system starts feeding on its own exhaust.
- The cassette-tape analogy is dead-on: each generation sounds coherent, but the signal-to-noise ratio decays.
That leads to:
- Content with the shape of meaning but little contact with lived reality.
- A culture optimized for pattern-consistency instead of originality, friction, and weirdness.
Economically, that matters because markets don’t just run on volume; they run on novelty and taste. A flattened culture produces flatter demand curves.
The Skills That Still Matter Are Not the Cheap Ones
You draw a clean line between what gets cheap and what gets scarce:
- Cheap: execution, production, filling the page, pushing the pixels, writing the boilerplate.
- Valuable: judgment, taste, prioritization, knowing what not to do, and recognizing when something is polished but strategically empty.
AI is a force multiplier for people who:
- Can define the problem.
- Can evaluate whether the solution is actually good.
- Can connect output to real-world stakes.
It’s a pricing problem for people whose value was mostly: “I can produce the artifact.”
If AI Absorbs Junior Work, Where Does Experience Come From?
This might be the most important structural point you make:
How do you become senior if you never get to be junior?
If AI absorbs the junior layer:
- Where do future senior engineers, strategists, creatives, and operators get their reps?
- Who gets to build pattern recognition if the low-stakes, repetitive work is automated away?
Companies will still want high-level judgment later, but judgment is trained, not conjured. If we hollow out the learning layer, we’re not just cutting jobs—we’re cutting the pipeline that produces future expertise.
That’s a long-term systemic risk that won’t show up in quarterly metrics but will show up as a shortage of real competence a decade from now.
The Real Warning Sign Is Not Layoffs
You land on a simple but powerful diagnostic:
- Don’t just watch for layoffs.
- Watch for the work that never leaves the building, the specialist who never gets called, the subscription that quietly gets canceled, the task that shifts from “hire” to “handle.”
That’s how a labor market thins out before the headlines and official stats catch up.
Efficiency Becomes a Demand Problem
You’re not arguing that AI is bad or that we should roll it back. You’re doing something more useful:
- Acknowledging the personal upside: more capability, more speed, more shipped work.
- Refusing to pretend those gains are free.
The core warning is sharp:
If enough work gets absorbed by fewer people, the economy that celebrated efficiency starts starving the base it depends on.
That’s the story policymakers, business leaders, and workers all need to grapple with: not a sudden robot takeover, but a slow, quiet digestion of demand, one missing invoice at a time.
Your piece doesn’t just describe that shift; it gives people the vocabulary—autophagy, missing invoices, model collapse, compressed layers—to actually talk about it clearly.
