Salesforce did not hire a single new software engineer in fiscal year 2026. Not one. CEO Marc Benioff announced the freeze while simultaneously cutting the company’s support headcount from 9,000 to roughly 5,000. Across the industry, the SkillSyncer Layoffs Tracker has recorded more than 205,000 tech job cuts so far this year — a 66% increase over the same period in 2025.

The same week those numbers circulated, a blog post titled “Software development with AI is starting to feel like cooking steak” hit the top of Hacker News. The argument, stripped to its essence: AI has done to coding what the modern supply chain did to steak. Anyone with a pan and a YouTube tutorial can produce a decent ribeye. Anyone with Cursor or Copilot and a prompt can produce a working web app. The skill floor has collapsed.

The post is charming and, in its narrow frame, correct. But it misses the more interesting question. When anyone can cook a steak, what happens to the restaurant business?

The Thing That Was Never Scarce

For two decades, the software industry convinced itself — and its investors — that the scarce resource was the person who could type the code. Salaries inflated accordingly. Bootcamps multiplied. The median software engineer at a major tech company earned more than the median lawyer or accountant, often for work that consisted largely of wiring together APIs and styling form elements.

The cooking-steak analogy is useful here, but not for the reason its author intends. A decent steak is easy to produce because the hard parts — breeding, feeding, slaughtering, aging, distributing — were solved by industrial systems long before the home cook touches the pan. The cook is performing the final, easiest step in a chain that required enormous capital and coordination.

Software is now in the same position. The hard parts — infrastructure, frameworks, libraries, cloud orchestration, security protocols — have been abstracted and automated to the point where the final assembly step, the thing we used to call “writing code,” is increasingly the least valuable link in the chain. Microsoft reported earlier this year that 20 to 30 percent of all code written in its ecosystem is now generated by AI. That number will only rise.

The Menu, Not the Kitchen

If the code-typing was never the scarce resource, what was? Two things: knowing what to build, and having someone to sell it to.

The Salesforce freeze is instructive here. The company did not stop building software. It stopped hiring the people who type it. The product roadmap, the customer relationships, the distribution channels, the pricing strategy — none of that got automated. What got automated was the translation of those decisions into functioning code.

A product manager at a mid-size SaaS company put it to me in a Slack DM this week: “We shipped more features in Q2 with 40% fewer engineers than we had in Q4. The bottleneck was never the coding. It was deciding which three features actually mattered.”

That is the real story the cooking-steak analogy obscures. When production costs collapse, the value does not disappear — it migrates upstream, to the people who decide what gets produced, and downstream, to the people who own the customer relationship. The kitchen becomes a commodity. The menu and the dining room do not.

The Split Nobody Planned For

The labor market is already reflecting this shift, though not in the way the doomsayers predicted. The 205,000 layoffs are real, but so is the hiring data. Employers report significant difficulty filling roles that require production AI experience. The market has not shrunk; it has bifurcated.

On one side: engineers who spent their careers as high-end typists — skilled at translating specifications into syntax — are discovering that their skill is now worth roughly what a short-order cook’s is. The work still exists, but the premium has evaporated.

On the other side: engineers who can design systems, evaluate tradeoffs, and — crucially — direct AI toolchains to execute on those designs are in higher demand than ever. The job posting data from Indeed Hiring Lab shows software development listings down 36.4 percent from pre-pandemic levels, but the listings that remain are qualitatively different. They ask for architecture, not syntax.

The cooking analogy holds here too. The person who can cook a steak is replaceable. The person who can design a menu, manage a kitchen, and keep the dining room full is not.

What the Freeze Actually Means

The instinct, when confronted with numbers like Salesforce’s zero-hire year, is to panic. Two hundred thousand layoffs feels like an industry in collapse. But the more interesting read is that the industry is finally being forced to distinguish between activity and output.

For years, engineering headcount functioned as a proxy for ambition. More engineers meant more features, which meant more growth — or so the logic went. AI-assisted development has broken that equation. A team of five can now produce what a team of twenty produced two years ago. The companies that recognize this first will not be the ones that cut deepest; they will be the ones that redirect the savings into the things AI cannot yet do: understand a customer’s problem before the customer can articulate it, identify a market that does not yet exist, build a brand that people trust.

The steak is easy now. The restaurant is the hard part. That is not a crisis for software. It is a correction — and, for the people who spent their careers learning to cook rather than learning to feed, a long-overdue one.

Sources