Nobody Answers for the Lie They Sold

People get laid off, then they find new work. Maybe it takes longer than it used to. The market is hard, but the market digests. That was the shape of the problem in my head. Painful, survivable, done.

I’ve spent almost a year pushing back on the replacement story, and I still think it’s a lie. The read came from somewhere narrower than belief. It came from where I was standing. I didn’t get cut. None of my people got cut. And the whole time, I was hiring. Our last round pulled 2,253 candidates for a handful of seats. From a chair that two thousand people are walking toward, “they’ll land somewhere” isn’t optimism. It’s what you watch happen every day. That chair is a window, and through it the broken market still looks like a working one.
An engineer answered one of my posts. He was standing at the other window: senior, twenty-five years in, a hundred days of runway left. His view: nobody hires, layoffs keep coming but nobody fills the seats, everyone frozen like a rabbit in headlights. Juniors get called useless. He told me, plainly:

You are not in the street seeing what I see or speaking with the unemployed devs as I am. Don’t throw optimism at people whose expertise got devalued by months of hype. You can’t eat the future. The damage is already here.

He was right, and the thing he was right about is narrow and exact. Not the mechanics. I still think the market is irrational and irrational markets correct. I told him as much, and I didn’t take it back. What I took back was smaller and worse to admit: I had been measuring his catastrophe with the ruler of my own untouched life. I knew people were getting laid off. I did not feel the weight of it, because the weight wasn’t landing on me. Two people can look straight at the truth and see opposite things, depending on which side of the glass they’re standing on. Each of us assembles reality from where we happen to stand, and mistakes it for the thing itself. I stand somewhere specific too. I was born where a lie had a price, and that makes a market with no price for lying hard for me to read as normal.

The present I missed

One story shook me. Kyle Simpson wrote “You Don’t Know JS.” He taught a generation how the language actually works. A few weeks ago he was on LinkedIn asking for warm intros. Not job-board links. Warm connections to people he could actually talk to.

When a Kyle Simpson has to ask LinkedIn for an introduction, this isn’t skill losing its value. It’s a market that believed a lie and seized up. What matters is that a person of that caliber gets spit out at all.

He is one face, behind him there’s a number, and I could have read it any time.

New software engineering job postings fell 15% in the first two months of 2026. Software developers aged 22 to 25 saw employment drop nearly 20% from their late-2022 peak. Tech cut roughly 52,000 jobs in the first quarter of 2026 alone, the worst opening quarter since 2023. Close to 900,000 tech workers gone since 2020. IBM tripled its entry-level hiring this year, the only company moving the other way, which tells you the rest had a choice and made the other one.

The freeze works from both ends. Juniors never get in, and the people who taught them get walked out. A senior with twenty-five years and a Kyle Simpson land in the same place, because a market that stops believing in the skill stops believing in it at every level.

The people who built it

What the executives did was a different thing. They sold the replacement on purpose, through the loudest microphones in the industry.

Sam Altman, March 2025: “At some point, yeah, maybe we do need less software engineers.” Mark Zuckerberg, January 2025: AI would be “a mid-level engineer” at Meta inside the year. Marc Benioff, February 2025: “We’re not going to hire any new engineers this year.” Dario Amodei, March 2025: AI would write 90% of code within three to six months, and within twelve months “essentially all of the code.”

Read the hedges. “Maybe.” “We may be in a world where.” The qualifiers sat in the transcripts and were gone by the time the claim reached a layoff memo. The speculation went up on stage wearing a hedge and came down as a fact. I can’t prove intent, but the pattern reads one way to me: the unhedged version moved the stock, so the unhedged version is the one that left the building.

It’s mid-2026. At my own company AI now writes most of the code, and it still hasn’t made the engineer optional, it’s done the opposite. Code that writes itself needs a human to judge it and validate it, or it turns into the wreck I’m about to show you. The writing was never the bottleneck, the judgment was. Their industry-wide version didn’t even land: independent reads put AI at about half the code even inside the labs making the loudest claims. Meta did not swap its mid-level engineers for a model. It cut thousands of these people while spending billions on different ones.

Two ends, one result

Code is solved. Generate it, ship it, the engineer is overhead. That’s what they were selling.

I audit that claim for a living. Recently a client brought us an application built start to finish by someone with no engineering background. Pure vibe code. They weren’t live yet. They came with one question: can we launch?

The thing worked, the features ran when he demoed them. Zero UI, zero UX, but it ran. That’s the giveaway before you open a single file. Then we opened the files.

Inside: zero type safety, every variable and every API response untyped. No input validation anywhere. User input flowed straight into database queries, guarded by nothing but if statements and browser alert popups. No tests, the one test file still searched for the framework’s starter-template placeholder. The production database password committed to the repo in plaintext. Passwords stored in client-side JavaScript, readable by anyone who opens dev tools. Authorization that was pure theater: roles enforced by hiding buttons, bypassable from the browser console in under a minute. Cross-site scripting in several places, with one component escaping its output correctly and the rest not, which is the signature of code assembled without anyone in charge of how it fits together. One 3,000-line source file. Twelve of twenty components past any reasonable size. Business logic duplicated across six files. Dead components wired to nothing. A 10,000-line CSS file with class names like phase14-, phase17-, phase18-, the archaeological layers of one prompt stacked on the last, nothing ever refactored.

Our verdict was no-go. The only thing standing between that app and a breach in a regulated industry was the review.

Zero expertise, predictable result.

Anthropic. The best-paid engineers on the planet, effectively unlimited compute, the company that sells the coding tool. Their own product shipped a single 3,167-line function with zero tests, already in production, serving a revenue stream in the billions. I took their engineering culture apart when the source leaked.

Maximum expertise and zero expertise, opposite ends of the spectrum, and the code rotted the same way at both. If code were solved, the model would have carried the amateur up toward the expert’s level. Instead it dragged both down to the same floor. The tool didn’t decide the outcome, the discipline on top of it did, and at both ends there wasn’t any.

The amateur asked permission and got told no. The experts asked no one and shipped to production. One of those two apps is live.

Somebody is on the hook

My code is already written almost entirely by AI, and it changed nothing about the part that matters. Someone still answers for the result, and “the AI merged it” has never been a defense anyone accepted. The work the executives called dead is the one thing standing between “it runs” and “it leaked.” I’ve written before about why that accountability can’t be delegated.

I was right that the work doesn’t vanish. I was wrong to think that because it doesn’t vanish, no damage was done. The people who do it got thrown out anyway, not because the work stopped needing them, but because someone sold a story that it did, and the bill for believing that story lands on whoever shipped on the strength of it. The engineer is still the bottleneck, and the engineer is still on the street. I missed the second half because it wasn’t happening to me.

Nobody paid

Altman now says he’s “delighted to be wrong,” that the disruption he forecast hasn’t shown up the way he expected.

Amodei pivoted to the Jevons paradox, the same 90% reframed as proof that productivity expands to fill the gap.

Benioff went from promising “radical augmentation” to cutting thousands with “I need less heads” inside a single month.

Jensen Huang told a generation in 2024 to stop learning to code and go into farming or biology instead. By 2026 he was calling the idea that AI reduces engineering jobs “complete nonsense” and saying the world needs a trillion lines of code. The kids who took the first advice didn’t enroll. Nobody has explained where the expertise is supposed to come from now.

No retraction. No correction filed under the same name that made the claim. The forecast that froze the market just got swapped for a cheerier one.

That’s the missing institution. There is no reputational cost in this industry for being wrong at full volume.

The King of Bullshit

The clearest proof of it isn’t even in this story. Take the man who built a career on deadlines that never arrive. Full self-driving “next year,” every year since 2015. A million robotaxis on the road by 2020. A million people on Mars, first crewed mission in 2024. Hyperloop, a “fifth mode of transport,” working lines within a few years of 2013. Brain implants in trials by 2020. None of it landed on time, most of it not at all, and by 2024 he was on a stage admitting “I tend to be a little optimistic with time frames” while announcing the next one in the same breath. The latest one, my favorite: in January 2026 he told Davos the cheapest place to run AI would be space “within two years, maybe three at the latest,” and filed to put a million data-center satellites in orbit, days before merging two of his companies into a $1.25 trillion entity headed for an IPO. The deadline is the product. Every missed date made him richer. He is, as of this writing, the wealthiest person who has ever lived. A forecast that never lands isn’t a debt in this industry. It’s a marketing budget, and somebody else pays it.

Nothing clears

You can call an entire profession obsolete. You can tell a generation not to learn to code. You can promise data centers in space within two years, days before your IPO. You can do all of it, watch the freeze land on real people, and pay nothing. The careers don’t un-break, the valuations went up. The same vendors now sell AI supervision as a service, packaging the exact work they spent years calling dead. A senior engineer with twenty-five years behind him counts how many days of money he has left. They count how many points the stock is up since they called him overhead. The difference between them isn’t talent and it isn’t honesty. It’s altitude, and altitude works so that no one up there ever has to look down at the person they wrote off as a cost.

And there’s a worse version than no one paying. The correction he and I both want may never come, not because the story turned out true, but because too much money is riding on it to let the tower fall. Capital that size doesn’t admit it was wrong. It props the thing up. That’s slower and quieter than a crash, and worse, because nothing ever clears.

Resist

He signed off with one word, twice. Resist.

From the height of the future you can’t see the rotten code or the broken people. You see them with your hands inside the machine, or you don’t see them at all. I didn’t, for a while, because I was standing at the one window where the damage doesn’t reach. A man with a hundred days of runway walked me to the other window in an afternoon.

I come from a place where if you lie and get caught, you can get punched in the mouth for it. Where a word costs something the moment it leaves your mouth. These men built an industry where you can lie to the whole world, break other people’s careers at full volume, and grow richer for it. I can’t tell you how to build a reputation system where there isn’t one. I just know what it looks like when it works, because I grew up inside one that did.