The biggest lie in the "AI replacing jobs" conversation is that it hasn't happened yet. It happened years ago. You just didn't notice, because the people who did it aren't the ones arguing about it online.
The Debate Is the Signal
Last week, I posted a tasteful little ragebait:
Obviously, it stirred up a good controversy, especially among designers. I don't blame them.
But it became clear that, like most things, people are split down the middle.
Group 1: The people debating whether AI is "good enough" to replace them. Reading the headlines. Block laying off their staff. Microsoft's AI CEO saying people have 18 months left. Citrini's doom report getting 25M+ views.
Copywriters arguing AI copy isn't good enough. Designers insisting AI can't match human creativity. Developers saying vibe coding produces nothing but garbage.
Group 2: The people who already decided, already integrated, already shipping. They see where the market is going and want to be ahead of it. These people are producing 3-5x the output they did two years ago.
Everyone in Group 2 is watching Group 1 argue in the comments thinking "oh boy, they're screwed." Everyone in Group 1 is looking at Group 2 thinking "these guys are delusional."
So let me be plain:
If you're still arguing about it, the people who aren't arguing are taking your jobs.
You're Asking The Wrong Question
"Will AI replace X?" generalizes too much, and misses what this shift is actually unwinding into.
The right question is:
Were you the expertise? The judgment? The taste? Or were you just the execution layer?
Because AI only replaces one of those. The same one that got replaced by the printing press, the steam engine, the automated assembly line.
There was a line from Citrini's article that captured this well:
"A lot of what people called relationships is simply friction with a friendly face."
AI kills friction. If your job was navigating complexity for other people, you're exposed. That's who the headlines keep reporting on.
Take logo design - since that's what got me the most hate:
Say you're a highly paid logo designer. Multi five-figures per project.
You assigned your value to something. A combination of expertise, skill, social proof, and so on.
But here's the important question:
How much of that value - which equals price - is determined by the actual output the client receives?
And how much is determined by how hard the job is? How much time it requires? How much effort?
There's your answer.
Sure, some people will value the artisan. The craftsmanship of a proper, hand-designed logo. But those will be the same people who listen to vinyl and hand-make their own pasta. It won't be the vast majority of clients, who simply care about how well that vector file represents their business.
AI didn't create this problem. It's just accelerating it.
A Faster Beginner Is Still a Beginner
A beginner with AI doesn't become an expert. They become a faster beginner.
This is true in every domain:
Coding: Karpathy calls AI "an over-eager junior intern savant who bullshits you all the time." His approach: tight leash, slow, paranoid. It only works if you can evaluate every line.
Senior dev + AI = 10x output. Junior dev + AI = 10x technical debt.
Design: Average designers use average prompts and get average outputs. The best designers pivoted to strategic roles years ago. The ones charging what they charged five years ago are charging for recognition and taste - not pushing pixels.
Media buying: Pushing buttons in Ads Manager should not be a unique advantage, but it has been for years. Good marketing ideas can now rise without navigating archaic dashboards. If your moat was the dashboard, your moat is drying up.
Business: A cardiologist placed 3rd at Anthropic's hackathon out of 13,000 apps. Built it in seven days while working at the hospital. Domain expertise + AI tools = winning products.
The tools amplify the input. If the input is good - if the person using them is an actual domain expert - then as AI improves, they produce 3x more work, then 4x, then 5x. Their usefulness scales exponentially, directly parallel with the tools they use.
That's where the secret is.
When Figma ships a legitimate code ↔ canvas tool, there will be 100x more bad designers shipping bad things. And the good designers will be shipping 100x more good design. They'll get paid more, land more projects, and - this part matters - they'll become better designers. Because they'll be exposed to 100x more opportunities to come up with ideas and materialize them.
The Slop Flood Is Your Advantage
AI made it trivially easy to produce something. So the world is drowning in it. AI-written articles. Vibe-coded apps that look good in demos and produce zero useful output. Copy-pasted purple gradient landing pages. Generic designs. Formulaic copy. Mediocre everything, at scale.
2026 is the year of technical debt.
But when the floor rises and everyone can produce slop, quality stands out more, not less.
All rap music started to sound the same. People got bored and started seeking original music. Walk into any young-people-oriented club in NYC right now - they're playing Pitbull and Ke$ha, not anything that came out this year.
The people who can't tell if output is good or bad were always going to get replaced. In every field. AI accelerating this should excite you.
The people who understand why something works are doing 3-5x the output, and the gap widens every single month.
The People Pulling Ahead Already Decided
I've curated a network that spans serious tech builders and cutthroat business operators. I'm seeing the best technology as it's released, and seeing what the real operators are actually using day-to-day to drive growth.
Half are building for the sake of building. Buying Mac Minis to spin up the newest agent setup, demoing how autonomous everything is - while the agents are just checking email and messaging them good morning. Legitimate progress is being made, but it's been contaminated by the race to get funded with empty promises and overpriced launch videos.
The other half are using what many would consider already outdated tools. Most of the legitimate business operators I know have been using Claude Desktop for 90% of their work. They think Claude Code is a literal coding app. They can't connect APIs. They don't know what an MCP is.
And they're loving it. Genuinely 10xing their output. They feel like they're using Jarvis.
Because the technology doesn't matter. The output does. How does this make me more money, today?
And the very small group doing both - pushing serious numbers, fully integrated, designing end-to-end workflows that research, design, build, launch, and optimize - chains of AI doing what used to require a team -
They're not posting about it.
Because they don't have time to. Because AI hasn't replaced them.
They still need to make the decisions. Pick the angles. Understand the market. Apply thousands of hours of boots-on-the-ground expertise to actually produce results.
An AI agent won't launch a more profitable offer than I would manually. That's not the point.
The point is it enables me to launch more profitable offers. To focus on the actual needle-moving work.
So stop debating whether AI is good enough to replace you.
Identify what you're actually good at. What you should be improving. How to amplify the output you can produce.
"You have 18 months to become the person who provides context, not the person who executes without it." – Forbes