9
That transfer has no name yet. No legal category. No movement.
That is the problem.
The language comes before the legislation.
"Your way of working is yours" is the next sentence.
Knowledge workers need to start saying it. Out loud. Together. Now.
I hope this is a horror story. The kind that feels real enough to disturb you but turns out to be fiction.
I'm not sure it is.
AI accounted for a quarter of all US layoffs last month.
And every time someone is let go, something happens nobody is talking about.
8
Knowledge workers are at the same moment.
But the thing being taken is harder to see.
Not your output. Output always belonged to the employer.
How you think. Your methodology. Your judgment, now running in a system you no longer have access to.
7
We have solved asymmetries like this before.
GDPR came because people named what was being taken.
The WGA just signed a new four-year deal last week, expanding AI protections. Training data as property.
Writers won something real by saying out loud what was being stolen.
6
This isn't about career ceiling.
It's about whether the floor holds.
As layoffs rise, supply of workers rises. Company-side intelligence gets better. The individual has less to differentiate with.
The floor doesn't just stop rising. It drops.
4
Your methodology doesn't sit in a vault.
It runs. Across hundreds of users. Compounding with everyone who came before you. Refined by volume and diversity no individual can match.
3
We've seen something like this before.
Musicians built entire genres. The labels kept the masters.
But musicians kept their instrument. They could still perform. Still compete.
What's happening to knowledge workers is more severe.
2
The company keeps more than the work.
When a knowledge worker operates inside a company's AI infrastructure, their prompts sit in that system. Their judgment calls. How they break down a problem.
When they're laid off, the output leaves.
The method stays.
3/3
Some things should not be skills at all. Skills are always-on. If you only need something at a specific moment, use a custom command instead.
A map of your skills is worth more than a library of them.
1/3
More AI skills is not better.
Before I even started a task, the AI was loading context through dozens of skills. Tokens burning. Wrong things firing.
What I learned after mapping 16 skills:
2/3
Conversation skills belong in Claude chat. Pressure-testing ideas, planning strategy. Thinking before building.
Build skills belong in the codebase. Review, architecture, test. Cursor or Claude Code.
We're not replacing the old way of building software. We're not just vibing with AI either. We're shapeshifters β learning to navigate the loop, not just run it.
I used to think AI would make everything linear.
Brainstorm β stress test β build β launch.
Clean. Fast. Efficient.
But that's not how it actually works.
What I've realised is that ideas behave like water: they find their shape based on what they encounter. Building in this new age is less like executing a plan and more like sculpting. You respond to what the material does. The form starts to suggest itself.
AI speeds up execution, but it can't tell you:
β When an incoming idea is worth disrupting your build for
β When to go visual instead of staying in code
β When a new feature request means revisiting an assumption you buried three weeks ago
I'm mid-build in Cursor and a new idea comes in. So I go back to Figma to think visually. Then back to code. Then a customer request shifts something I assumed was settled. Then inspiration from somewhere else reframes how I'd implement a detail.
4/ None of this stops you from building.
But knowing the list before you choose your path changes what you build toward.
The itch is the same starting point. What comes after it isn't.
You vibe coded it. It works. It fits your brain perfectly.
Selling it is a different size of world.
Auth. Onboarding. Billing. Rate limits. Legal. Unit economics.
None of this stops you. But knowing the list before you decide which path you're on changes what you build toward.
3/ π¦ Feature flags by plan or user π‘οΈ Rate limiting β someone hammers your AI API, that's your bill π° Unit economics β does money in actually cover what goes out? π Activation and retention, not vanity metrics π Error monitoring, uptime, a support channel βοΈ Terms, privacy
2. The gap isn't technical skill. It's the invisible layer every real product needs:
π Auth and account management π Onboarding for people who don't think like you π³ Billing, upgrades, cancellation flows π Data isolation (obvious until you get it wrong)