Your billable hour is being repriced in a room you're not in.
If you're a senior consultant charging $400/hour, a domain analyst pulling $180K, or a specialist whose expertise took fifteen years to build — the floor under your rate is moving. Not in five years. In the next eighteen months. The Pragmatic Engineer's latest breakdown shows engineers already being compensated against AI leverage, not headcount. Staff engineers who ship with agent stacks are getting repriced upward. Engineers who can't are getting repriced down, or out.
What happens to your book of business when a Series B startup decides one operator plus a stack of agents replaces the three-person advisory retainer you've held for four years?
I'm Matt Cretzman. I build AI systems for a living, and I've spent the last two years studying the exact pattern that's coming for every knowledge worker on the planet. The engineering repricing of 2026 is not the story. It's the preview.
The Engineering Preview, In Plain Numbers
Here's what the data is showing. Companies are no longer hiring engineers to scale output linearly. They're hiring engineers to direct agent fleets. One staff engineer with Claude Code, Cursor, a custom toolchain, and three sub-agents now ships what a five-person team shipped in 2023.
The market noticed. Comp packages are bifurcating. Engineers who operate as orchestrators are getting paid like they're managing teams — because functionally, they are. Engineers who still ship line-by-line are being benchmarked against the agent floor, which keeps dropping.
This isn't a layoff story. It's a repricing story. The role didn't disappear. The value of the role got redistributed to whoever owns the leverage layer.
Now apply that template to every other expert profession.
The Same Restructuring Is Already at Your Door
Consultants: your deliverables — the deck, the framework, the discovery interview synthesis — are now being produced by GPT-5-class models in twenty minutes. The client doesn't care that yours is better. They care that yours is six times more expensive and four days slower.
Analysts: your reports are being generated by agents reading the same public filings, the same earnings calls, the same trade press. Your edge was synthesis speed. The synthesis layer just got commoditized.
Domain experts: the tacit knowledge in your head — the thing you assumed was unautomatable — is exactly what platforms like Mercor are paying $80/hour to extract from people like you, by the thousands, into training sets that will compete with you next quarter.
That is the second extraction economy. I wrote about it in Chapter 4 of my book. Experts dumping fifteen years of judgment into someone else's model for hourly wages. Jacob selling his birthright for a bowl of stew, except the stew is a 1099 and the birthright is your entire professional moat.
Why "Just Use AI Yourself" Is the Wrong Answer
The common advice is: experts should just adopt the tools. Use Claude. Use ChatGPT. Get faster.
This misses the structural point entirely. Speed alone doesn't defend you. If every consultant in your category gets 5x faster, the floor on consulting rates drops 5x. You ran on a treadmill and ended up in the same place, at a lower price.
The engineers who are getting repriced upward aren't the ones who just use AI. They're the ones who own systems — repos, internal tools, agent stacks, evaluation frameworks — that compound. Their leverage is artifact-based, not effort-based. Each week they ship something that keeps producing value next week without their hands on it.
That's the gap. Most experts are still trading hours for dollars while the engineering world has already moved to trading artifacts for equity-style leverage.
Knowledge Debt: The Liability Most Experts Don't See on the Balance Sheet
I call the thing you owe to your future self knowledge debt. Every year you operate as a senior expert without converting your tacit judgment into owned artifacts, you accrue more of it.
You know the framework you use to diagnose a struggling pipeline. It's never been written down. You know the seven questions that surface a misaligned executive team in under an hour. They live in your head. You know the eighteen failure modes of a bad acquisition integration. They've never been codified.
That's the debt. And the interest payment is coming due now, because the platform layer — the LLMs, the agent frameworks, the vertical SaaS — is racing to codify generic versions of your expertise from public data and Mercor-style extraction labor. When they finish, your unwritten edge becomes worthless. Not because it was wrong. Because someone else shipped the artifact version first.
The engineers winning in 2026 paid down their knowledge debt early. They wrote the internal docs. They built the tools. They turned what they knew into what they owned. That's the entire game.
What an Expert-Owned Stack Actually Looks Like
This is what I built Skill Refinery to do. It's a Knowledge Delivery System — the operating layer between an expert's tacit judgment and a market that can buy artifacts instead of hours.
The stack has four layers:
Capture. Structured extraction of your frameworks, decision rules, failure modes, and judgment patterns. Not transcription. Codification. The output is a knowledge graph you own.
Refinement. The raw capture gets processed into reusable artifacts — diagnostic flows, decision trees, evaluation rubrics, scenario libraries. These are the things that compound.
Productization. The artifacts become deliverables. Audits, assessments, custom-trained agents, productized engagements, licensed playbooks. Each one is a card you can play repeatedly without re-extracting from your brain.
Distribution. You own the customer relationship, the pricing, the IP, the revenue. Not a platform. Not a marketplace that's quietly training a model on your output.
That last part is the whole thesis of On Whose Terms. There are three extraction economies operating right now. LLMs took the knowledge already — without consent, on their terms. Mercor and its peers are buying the next layer hourly — on their terms. The third economy is the one where the expert owns the cards, the revenue, and the distribution. On your terms.
You're already in one of these economies. The only question is which.
The Math an Expert Needs to Run This Week
Here are three numbers worth pulling up before next Monday.
Your hourly leverage ratio. Take your annual revenue. Divide by hours worked. Now subtract the percentage of those hours that produced something reusable. If less than 20% of your time generates compounding artifacts, you're operating like a 2022 engineer in a 2026 market.
Your codification gap. List the ten frameworks, diagnostics, or judgment calls you make most often. How many exist as artifacts someone else could be trained to run? That's your knowledge debt principal.
Your platform exposure. Where does your revenue actually live? If it's on a platform that can change its take rate, kill your category, or train a competitor on your output, you don't own a business. You own a tenancy.
Fix the worst of the three first. The engineers who got repriced upward in 2026 didn't fix all three at once. They picked the binding constraint and shipped against it for ninety days.
The Window Is Narrower Than It Looks
The engineering repricing happened over roughly eighteen months from inflection to new normal. Consulting, analysis, and domain expertise are next, and the cycle will be faster because the tooling is more mature and the playbooks are now public.
This is a stewardship question more than a strategy question. The expertise you carry is a gift compounded over a career. Letting it leak into someone else's training set for hourly wages is not a neutral choice. It's a decision about whose terms your life's work serves.
I'm writing a book about exactly this — On Whose Terms: The New Expert Economy and the Fight for What You Know. If the thesis resonates, join the launch list and you'll get the first chapters before they're public.
The full system, the case studies, and the build notes live at mattcretzman.com. Engineers got eighteen months of warning. You're getting yours now.
Keep Building,
— Matt