Why "Pretty Good at Things" No Longer Pays
Someone asked: “Is AI quietly killing the value of being pretty good at things?”
And the answer, if you strip away the optimism, is yes. Completely. Systematically. In real time.
But not in the way people think.
What “Pretty Good” Actually Was
Let’s be clear about what “pretty good at things” meant economically before AI:
You were a competent software engineer. Not elite. Not a 10X founder. Just solid. You knew Python, could ship features, understood databases, made reasonable architectural choices.
You made $150K/year. You had job security. You could freelance and charge $150-200/hour.
Or: You were a decent writer. Not Hemingway. Not even top-tier journalist. But you could turn in clean copy, meet deadlines, understand structure. You made $60-80K as a staff writer or contractor. You had a moat.
Or: You were a solid product designer. Not exceptional. But you knew your craft, could critique other work, had taste. You commanded $100-150K. You had leverage.
That middle ground; the “solid B+ professional”..that was the economic majority. And it worked.
Why did it work?
Because scarcity + distribution + trust = pricing power.
You were one of 10,000 competent Python engineers in the world. Getting a competent one to your specific project required friction, cost, and time. That friction = your salary.
Scarcity meant you could charge a premium over “just okay” (commodities) without being elite-level expensive.
What AI Did to That Math
Here’s what changed:
Scarcity collapsed overnight.
Now every founder, every PM, every analyst can generate “pretty good” output instantly. A coding task that took a competent engineer 40 hours takes Claude 20 minutes.
The output isn’t perfect. It’s not always right. But it’s good enough.
And “good enough” that was $150K/year work, or $10K/month freelance work, is now a $20/month subscription away.
The supply curve didn’t shift. It inverted. From “rare” to “infinite.”
What happens to pricing when supply goes from scarce to infinite?
It collapses.
Not to zero (yet). But to the commodity rate. And pretty good at things is now competing in the commodity market.
The Split: Elite vs. Commodity
Here’s what actually happened:
The middle ground didn’t just compress. It disappeared.
You now have two categories:
Category 1: You Understand the Fundamentals
You’re not “pretty good at coding.” You understand why code works the way it does. You can audit what Claude produces and catch the 3-5% of outputs that are subtly wrong. You understand the tradeoffs. You can guide the AI toward the right solution.
You are now a force multiplier. You do 10X the output with the same effort. You’re more valuable than ever.
You go from $150K to $300K+ because you’re not competing on output speed anymore. You’re competing on judgment.
But this requires deep expertise. Real expertise. The kind that takes 10+ years to build.
Category 2: You’re Competing on Speed with Everyone Else
You can use AI. Great. So can 500,000 other people with a $20/month subscription.
Your “pretty good at Python” is now competing directly against:
The high schooler in Lagos using Claude
The bootcamp grad in India using Cursor
The laid-off engineer trying to get rehired using Gemini
You’re in a global commodity market for output. And you’re not the cheapest, not the fastest, and not the most reliable.
Your pricing power: destroyed.
Why “You’ll Just Learn to Use AI Better” Is Cope
There’s a comforting narrative: “AI is just a tool. The people who get good at using AI will thrive.”
This is technically true and economically false.
Yes, you can learn to use AI better than the average person. But so can everyone else. The skill of “prompt engineering” or “using Claude well” is:
Easy to learn (2-4 weeks)
Widely available (thousands of YouTube tutorials)
Quickly commoditized (everyone catches up)
Learning to use AI well is like learning to use Excel well in 2010. It’s a minimum viable skill. Not a competitive advantage.
The people who “thrive with AI” are the ones who already had deep expertise before AI existed. The people who can:
Understand whether the AI output is correct (requires domain knowledge)
Identify what the AI got wrong (requires pattern recognition)
Redirect the AI toward the right solution (requires intuition + experience)
That’s not “learning AI.” That’s being good enough at your field that AI becomes a speed multiplier instead of a replacement.
And that group is maybe 10-15% of the “pretty good” professionals who existed before.
The Brutal Economics
Let me show you the math:
Pre-AI:
100 competent engineers worldwide: $150K each
Scarcity creates pricing power
Supply constraint = $15M total payroll
Post-AI:
Same work output achievable by: 10 elite engineers + 50 generalist operators using AI
Elite engineer productivity: 10X output, $300K salary = $3M
Generalist operator (AI-assisted): Same output, $40K salary = $2M
Total payroll: $5M (down 67%)
But wait..there’s more work now. AI made it cheaper to build things, so more things get built.
So instead of 100 engineers doing 1 project each, you have:
10 elite engineers doing 10 projects each (directing AI)
50 generalists doing 10 projects each (executing with AI)
Now you need 600 engineers (not 100) to do all the work.
The jobs didn’t disappear. But:
90% of the jobs are now “generalist with AI” ($40K) instead of “competent professional” ($150K)
10% of the jobs are “expert directing AI” ($300K)
That’s not job destruction. That’s class destruction.
The broad middle of solid professionals doesn’t exist anymore.
Where This Really Breaks Down
The optimists say: “But you need domain expertise to validate AI output. That’s still valuable.”
True. But here’s the catch:
Expertise without scarcity is eventually worthless.
Right now, in 2026, if you’re a real expert; someone who actually understands their field deeply..AI makes you more powerful. You can validate output, guide AI, catch mistakes.
But this advantage lasts maybe 3-5 years.
After that, the AI gets good enough that it catches its own mistakes. Or good enough that the “validation” work also gets automated. Or good enough that the judgment calls the expert was making become embedded in the model.
We’re already seeing this:
Medical imaging: Radiologists who understood the subtleties of interpretation were irreplaceable. Now the AI catches things the radiologist misses.
Tax preparation: Experts who knew every loophole were gold. Now AI knows every loophole better than the expert.
Software architecture: Architects who made good judgment calls were rare. Now Claude can make reasonable architectural calls in seconds.
The expert advantage compresses as the AI gets better.
The Real Casualty: Optionality
Here’s what actually worries me:
Being “pretty good at things” used to give you optionality.
You were a solid engineer? You could freelance, consult, start a company, jump between jobs, negotiate for raises. You had leverage because you were relatively rare and useful.
That optionality is gone now.
You’re either:
A. Expert enough that you’re running the AI (rare, hard to build, requires 10+ years of real work)
B. Competent enough to use AI effectively (extremely common, easily replaceable, commodity wage)
There’s no middle ground where you’re stable, reasonably compensated, and have leverage.
The median software engineer salary has already started falling (not in nominal terms, but in purchasing power and negotiating position). Mid-level design jobs are disappearing. Freelance writing markets are collapsing.
This isn’t because the jobs disappeared. It’s because the bargaining position of “good but not great” professionals evaporated.
Who Actually Wins in This
Let me be clear about the distribution:
Winners:
People with significant pre-AI expertise (engineers with 10+ years, specialized domain experts)
People with capital to own the AI tools or direct AI output
People who can aggregate operators + AI into higher-level services
People in fields where AI can’t work (trades requiring hands, high-stakes judgment, relationship-dependent work)
Losers:
The 80% of “competent professionals” whose entire value proposition was being scarce + reliable
Junior professionals (who would have learned by doing) now learning by watching AI
Career switchers (who relied on “pretty good at things” to restart)
Anyone whose expertise is 5-10 years old
The Hidden Cost Nobody Talks About
Here’s what gets no air time:
Using AI for your work destroys your expertise development.
The engineer who builds features by directing Claude learns 10% of what the engineer who builds them from scratch learns. They don’t develop intuition. They don’t hit the walls that teach you how systems actually work. They don’t build the deep knowledge.
This creates a hidden time bomb:
In 5 years, you’ll have a generation of “engineers” who can direct AI but don’t actually understand engineering. They can’t debug novel problems. They can’t architect systems that need to scale to new constraints. They can’t innovate.
And the few people with real expertise will command extraordinary premiums because they can do what the AI-directed generalists can’t: think.
The Uncomfortable Question
If AI can do the work of a “pretty good” professional, and the output is acceptable 95% of the time, then what was the value of that professional actually about?
Was it the work? Or was it the scarcity?
I think it was the scarcity. And now the scarcity is gone.
The hard truth: being pretty good at things never had much inherent value. It only had market value because good people were rare. Scarcity created the pricing power, not competence.
Now that AI can create infinite supply of “pretty good,” the market value is zero.
What Happens Next
There are a few scenarios:
Scenario A: The Optimistic One
AI becomes good enough that almost everything gets cheaper, more abundant, and better. New jobs emerge in areas we can’t predict. Society adapts. The median standard of living goes up.
This happened after tractors (agricultural jobs disappeared, new jobs emerged). This happened after the internet (typing pool jobs disappeared, new jobs emerged).
It could happen again.
But the transition period is brutal. And the new jobs probably require different skills, in different places, for different compensation.
Scenario B: The Realistic One
The middle class compresses. Wages in “AI-competed” fields fall 30-50% over the next decade. The premium for genuine expertise skyrockets. Inequality increases sharply.
Society figures out some redistribution mechanism (UBI, taxes on AI, wealth caps) or it doesn’t. If it doesn’t, social stability becomes a problem.
Most likely outcome: we muddle through, some redistribution happens, and people adapt by moving into non-AI fields or by building expertise deep enough to stay valuable.
Scenario C: The Dystopian One
AI gets good enough that it can do almost everything. Genuine human expertise becomes genuinely scarce (because there’s no incentive to build it). Capital owners consolidate power. The majority of humanity becomes economically unnecessary.
I think this is 15-20% likely by 2040. Not 50%. But not negligible.
What You Actually Need to Do
If you’re “pretty good at something” right now, here’s the unvarnished truth:
Get much better, or get into a non-AI field.
If you stay in your field: Develop the expertise depth that AI augments instead of replaces. This means understanding the fundamentals so well that you can validate AI output, catch subtle mistakes, and guide solutions. It means going from “competent” to “expert.” It takes work.
If you can’t do that: Plan a transition. To trades (plumbing, electrical, HVAC—AI can’t do these yet, and they’ll be in demand). To relationship work (therapy, coaching, sales). To AI-direction (learning to manage AI systems and aggregate output into higher-level products).
Understand your timeline. You probably have 3-5 years before the market fully adjusts. Use that time to either:
Deepen expertise
Build something (product, brand, audience) that has defensibility beyond just your output
Develop skills in a field AI can’t touch
Build your own thing. The one thing that hasn’t been commoditized: ownership. If you can direct AI to build something and own the output (product, company, brand), you have leverage. Being a contractor offering “AI-assisted services” pays commodity wages. Being a founder using AI to scale a product keeps optionality.
The Honest Version
Being “pretty good at things” was never a stable economic position. It only felt stable because scarcity propped it up.
AI didn’t create this problem. It just revealed it.
And now that it’s revealed, the people trying to stay in the middle ground—using AI to keep doing what they’ve always done, at the same level they’ve always done it—are going to discover that their job security was always an illusion.
The scarcity is gone. The pricing power is gone. The optionality is gone.
What’s left is a binary: either you’re expert enough to stay valuable in an AI world, or you’re competing in the commodity market and losing to infinite supply.
Most people won’t make that transition. They’ll get squeezed. And the economy will look very different in 10 years.
The winners will be the people who understood this transition was coming and acted before it was obvious.
If you want to dig into defensible skills, AI-proof career transitions, or how to build expertise that compounds instead of compresses, that’s a separate conversation. But the time to have it is now, not in 2 years when your market rate has already adjusted downward.
The good news: the people reading and thinking about this are already ahead of the curve. The bad news: most people won’t think about it until it’s too late.