There’s a particular kind of whiplash that comes from reading two well-known tech figures discuss AI’s effect on programming and then looking at what is happening inside actual engineering organizations.
The conversation involved Boris Cherny, an Anthropic technical executive and the founder and head of Claude Code, and Lenny Rachitsky, a well-known Silicon Valley tech writer and product person. Cherny’s prediction was blunt: within a year or two, you really won’t need to learn programming. He reached for a grand historical analogy—Gutenberg’s printing press. Printers, he argued, freed scribes from tedious copying so they could focus on more creative work like bookbinding and illustration; over 200 years, literacy rose from about 1% to 70%. The implication was that AI would do the same for programmers: remove Git operations and dependency management, and let us focus on “genuinely interesting product thinking.”
Rachitsky was excited too. People without technical backgrounds can now build projects, he said. Instead of spending hours on Stack Overflow, you ask an AI.
I understood the appeal. I also felt something was off.
From where I sit—as a software engineer—the story doesn’t match the reality.
What I see isn’t liberation. It’s removal.
Cherny’s future is attractive: senior programmers combine multidisciplinary knowledge and do more valuable work. But what happens to the people who are not yet “senior”?
Look at the industry’s actual numbers.
Amazon, between October 2025 and January 2026, cut roughly 30,000 corporate roles across two rounds. This was not a minor trimming of fringe businesses. About 15% of HR roles were cut. Parts of AWS’s mid- and back-office teams were slimmed down. In China, the export seller management team was cut by as much as 50%. It was the largest corporate layoff in Amazon’s 30-year history, affecting nearly 10% of its 350,000 corporate employees.
Block—the U.S. payments company behind Cash App and Square—cut 40% of its staff in February 2026, shrinking from more than 10,000 people to fewer than 6,000. CEO Jack Dorsey was direct in his public letter: AI tools are fundamentally changing how companies operate. “Smaller teams, using the tools we’re building, can do more and do it better.” After the layoff announcement, Block’s stock rose nearly 25%.
Executives have been equally direct on earnings calls: AI lets us produce the same output with fewer people. Translated: basic programming work can be automated, so the people who do basic programming are no longer needed.
Cherny’s future—senior programmers applying cross-disciplinary knowledge—rests on a precondition: you first have to become a senior programmer. But junior roles are disappearing in swaths. Young people are finding the entrance to the “level up” path nearly blocked.
Was the printing press really so gentle?
The printing-press analogy leaves out a few important facts.
The story of scribes happily liberated is survivorship bias. Historical records show that most scribes did not gracefully transition into “interesting illustration.” Their skills lost value directly. More importantly, the printing press took 200 years to spread. Society had time to adjust.
Today’s AI shift is different. Amazon cut 30,000 people in half a year. Block cut 4,000 in a few months. That is not “transition.” It is a generation’s career path being closed at speed.
Then there is the idea that you no longer need to understand the low-level details. On a podcast, Cherny mentioned that 80% of Anthropic’s code is now written by Claude. He shared an example: an engineer wrote a service in Go for a month but “still doesn’t really understand Go.” He considers this a good thing—if the result is correct, you don’t need to know the details.
But as someone who writes code, I know that “tedious work” is exactly the path to understanding systems. When AI removes that path, the surface looks like liberation. In practice, it is deskilling. Your bargaining power shifts to whoever controls the AI black box. The person who “doesn’t understand Go but can write a service” may look efficient, but is actually easier to replace.
The cognitive split
This is the split I mean: in the same industry, executives describe one future and frontline engineers face another reality.
<table> <thead> <tr> <th>What the tech commentary describes</th> <th>What I see as an engineer</th> </tr> </thead> <tbody> <tr> <td>AI lets programmers focus on “interesting work”</td> <td>Basic roles are vanishing: Amazon cut 30,000; Block cut 40%</td> </tr> <tr> <td>Cross-disciplinary work creates more value</td> <td>Young people can barely get a foot in the door</td> </tr> <tr> <td>A slow, gentle change like the printing press</td> <td>Layoffs that used to take years happen in months</td> </tr> <tr> <td>“You don’t need to understand the low-level details” is liberation</td> <td>Deskilling makes people easier to replace</td> </tr> </tbody> </table>There is an implicit premise in Cherny’s narrative: technological change is always good, and those who adapt will benefit. But technology is never neutral. When he says “you can do more interesting work,” he assumes everyone has the standing to enter that decision-making layer and that the market will automatically compensate those who are displaced.
The reality is harsher. A Harvard study covering 62 million workers found that after companies introduced generative AI, junior developer roles fell 9–10% within six quarters. Big tech hiring of new graduates has dropped by 50% over the past three years. Some companies say it outright: if an AI coding agent costs less, why pay $90,000 for a junior engineer?
This is not “transition pain.” It is a structural door closing.
The printing press did not automatically produce the Renaissance. Without institutional buffers—real retraining, social protection, antitrust enforcement—it may mainly produce faster unemployment.
And this time, the people being optimized out are not blue-collar workers. They are software engineers, once treated as a safe harbor for the educated elite.