killdate.dev8 minute read
Why AI anxiety is different
The anxiety about AI isn’t irrational. It’s accurate.
Humans are evolutionarily wired for loss aversion — costs register at roughly twice the psychological weight of equivalent gains. AI makes that asymmetry visceral: the skills you spent years building are suddenly worth less, the job role that was already precarious is now more so, and the identity you went into debt to build is replicated by a model in seconds. The upside is abstract, somewhere in the future. The problems land now — in your income, your stability, your future employment prospects.
Public companies have a fiduciary obligation to shareholders. Private companies have no choice but to compete in ruthlessly competitive markets. When AI can execute a function cheaper than a salaried employee, replacing that employee isn’t a choice — it’s the mandate. The uncomfortable truth is that many corporate roles in westernized economies were always process-following, information-routing, template-filling. Not jobs requiring judgment. Organizational overhead waiting for better software. If your role can be replicated by a few prompts, you weren’t doing a job — you were executing a function. Feeling entitled to a paycheck for work that required no real human judgment is a negotiation the market is no longer willing to have. This isn’t cruelty. It’s correction.
AI adds a dynamic that older technologies didn’t. Every previous wave — railroads, electricity, the internet — was at least comprehensible in principle. You could understand how it worked, even if you couldn’t build it. AI’s decision-making is opaque to the heavyweight intellectuals training the models, let alone the people using it.
What history suggests, though, is stranger and more interesting than the fear implies.
Bubbles are a feature, not a bug
Every major technological revolution has followed the same arc. Speculative capital floods into new infrastructure — more than the economics can justify. The bubble inflates. It bursts. Companies go bankrupt. And then, quietly, the infrastructure stays.
Carlota Perez calls this the Installation Period. Financial capital builds what production capital cannot justify. The crash isn’t the failure of the innovation — it’s a transfer mechanism. Infrastructure moves from those who overpaid to those who get it cheap. And that’s when things get interesting.
The railroad bubble of the 1840s left 70,000 miles of track. The companies mostly went under. Freight costs dropped 90%. Chicago became the food system of the world — grain markets, stockyards, commodity futures, none of it viable without cheap freight. Sears launched a mail-order catalogue that was a software layer on the rail network. Carnegie negotiated rates to near-zero and built a steel empire nobody could touch. The bubble built the tracks. The tracks rewired commerce.
The telecom bubble is the cleaner example. By 2001, 95% of the fibre laid during the boom was dark — the companies bankrupt, the cable unused. YouTube launched in 2005. Netflix in 2007. AWS in 2006. Each was a bet on bandwidth staying near-free. It did — infrastructure acquired at bankruptcy-auction prices. Streaming video at 1999 pricing was economically insane. At 2005 pricing it was inevitable. The substrate made the application layer possible.
What I’ve come to love about capitalism — not uncritically, but specifically — is this: it has a mechanism for funding infrastructure that no rational actor would build alone. Bubbles are wasteful and transfer wealth in ugly ways. The people who lose in the bubble are never the people who win in the deployment phase. Different actors, different timescales, same infrastructure.
The AI version is straightforward. The LLM and GPU buildout of 2022–2025 is bubble-scale financing. Whether the builders profit is beside the point — I don’t care. The bubble leaves cheap inference and models that can build alongside you. The question is what you make with it.
Taste is the differentiator AI can’t close
For most of the history of building things, taste was bottlenecked by production. A good or bad idea was hard to ship. The friction of making something — writing the code, designing the interface, producing the copy — acted as a natural filter. Slow execution bought time to course-correct.
None of that means the skill floor dropped. Working across Claude Code, Cursor, Codex, AWS infrastructure, complex CI/CD pipelines, Python, Node.js, and Remix requires broader technical depth than someone banging away at Python syntax on a bad-idea startup. You need to understand how the layers connect, where the seams are, and what breaks under pressure — not just what to prompt. AI handles the gruntier production work. It demands you be more technical, not less — due to the complexity of the build.
AI removes that friction. Execution is becoming cheap, fast, and available to everyone. Which means the differentiator is no longer who can produce — it’s who can direct. Taste has always mattered. Now it’s the only thing that compounds.
There’s a line from Michael Polanyi I keep coming back to: we know more than we can tell.
Tacit knowledge — expertise internalized past the point of articulation. The surgeon who can’t explain the cut. The designer who knows something’s wrong before they can say what. The engineer who feels a system about to break before the logs confirm it.
That’s taste. Not soft. Not decoration. Analytical judgment compressed into something faster than language.
The emotional response is the computation — not a distraction from it.
The bureaucratic roles being automated out of existence required none of this. These functions weren’t just replaceable — they were a productivity drag embedded in the price of everything built on top of them. As that overhead compresses, prices follow. Business free cash flow expands. Consumer discretionary income goes further. That capital flows into new products, new services, new companies — most of which, I’d wager, get built by small teams running on AI.
What AI can execute — and what it can’t originate
AI can build the thing. It cannot decide what the thing should be, or notice when it’s becoming the wrong thing. Those two gaps are where your real leverage lives.
The canvas is yours. Your taste — what you notice, what draws you, what feels unmet — determines what gets built. The product vision, the design direction, the decision about what matters to a user: these come from something inside you that isn’t transferable. Your instincts about what to build are expressions of your personality, not just your skill.
AI executes against the canvas. It can’t define step one of the thousand steps a great product requires.
Drift detection is yours. The person who can feel when something is getting off track — before it’s visible in the output, before the tests fail — that’s intuition built from experience. A process without that human instinct behind it is just bureaucracy. With it, it’s leverage.
Neither of these works without real depth. Genuine UI/UX judgment, product management instincts, engineering knowledge that goes past the surface — these aren’t nice-to-haves, they’re the prerequisites. AI amplifies your existing craft. The shallower that craft, the shallower the output. A mediocre designer with AI produces mediocre things faster. A deep practitioner builds things that previously required a team.
Both compound with time. The canvas gets richer as your taste develops. The drift detection gets sharper as you accumulate scar tissue. Neither can be downloaded.
The difficulty scales with how far you can see
The feedback loop is compressed — you can ship fast and see the consequences fast. That part is true. But it would be dishonest not to say what the loop actually feels like from the inside.
This way of working is genuinely brutal. You are holding the product vision, managing an AI that will confidently go wrong without noticing, reading output for drift, making architecture calls, feeling the seams of a system you’re building in real time — on a small team, moving fast.
The further you can see into the consequences of each decision, the more you’re holding at once. An architecture call isn’t just the immediate output — it’s the refactor it will or won’t require in six weeks, the seam it will or won’t expose in production, the drift it will or won’t hide from you. That’s the actual cognitive weight of this work.
If you find it demanding, that’s probably the right signal. Most people fire a few prompts, get something mediocre back, and decide the tool isn’t ready. The ones who stay — who hold the vision, feel the drift, catch the seams before they compound — were always going to build things that matter.