In the first week of December 2025, a Google principal engineer sat down with Claude Code and described, in plain English, a problem her team had just spent the past year trying to solve.
Three paragraphs. A brief description of the problem with “no real details,” she said, because she couldn’t share anything proprietary, especially not on a competitor’s platform.
One hour later, Claude had produced a working prototype of her team's system. It wasn’t perfect, but for what it was – toying around with a system, just to see what it could produce – it was stunning.
She posted about it publicly.
“I am not joking,” she wrote on X, “and this isn’t funny.”
Anyone working on AI systems in late 2025 could relate to the whiplash she felt. In Slack channels, Reddit threads, the quiet conversations after the cameras turned off, everyone was abuzz. AI had crossed a threshold.
Not gradually. This was not the slow creep of improvement that characterizes most technology. This was a phase transition, the way water becomes ice: The same substance, suddenly organized according to different rules.
By February 2026, Claude Code run-rate revenue had crossed $2.5 billion, a growth curve steeper than any developer tool in history. Four percent of all public commits on GitHub were being generated by Claude Code alone, according to SemiAnalysis, and that figure did not include other AI coding tools. Anthropic's own data showed their engineers relied on Claude for sixty percent of daily work, and twenty-seven percent of that was work that had not existed before the tool – new work the tool itself had made possible or irresistible.
But the numbers were the scaffolding, not the building. The building was what it felt like. The rules that had governed every career in technology had been rewritten. What is hard, what takes time, what requires a team, all rewritten in months by a tool that had learned to think alongside you.
The building was the feeling in the room when a junior developer shipped in a weekend what her senior colleague had quoted six months for, and both of them knew it, and neither of them knew what it meant for Monday morning.
I crossed the line in a room in Trivandrum, India, in February 2026. I had flown there because no amount of Zoom calls or training decks could replace being in the room, doing the work together. Twenty of my engineers, experienced technical people who had been building software for decades, sat across from me while I said something that probably sounded insane: "By the end of this week, each one of you will be able to do more than all of you together."
The tool was Claude Code with the Max plan. One hundred dollars per person, per month. On Monday, we started building.
By Tuesday, something had shifted in the room; you could see it in the way the engineers leaned toward their screens, the conversations that started happening between people who didn't usually work together. What normally took the team a month to build was getting done in a single day.
By Wednesday, the engineers had stopped looking at each other for confirmation and started looking at their screens with the particular intensity of people who are recalculating everything they thought they knew about their own capability.
By Friday, the transformation was not a theory or a demo. It was measurable, repeatable reality. A twenty-fold productivity multiplier, at a hundred dollars a month.
I felt the exhilaration first. Twenty engineers, each operating with the leverage of a full team. The democratization of capability happening in real time, in a room in southern India, not in a San Francisco boardroom.
Then, I felt the terror.
If each of these people could now do what twenty of them used to do together, then every assumption I had built my career on was wrong. Teams, timelines, hiring, what it takes to ship a product. All of it wrong. Not slightly wrong. Structurally wrong.
I stood in that room on Friday afternoon, and I could not tell whether I was watching something being born or something being buried.
Both, probably.
One engineer, a woman who had spent eight years on backend systems and had never written a line of frontend code, built a complete user-facing feature in two days. Not a prototype. A feature that shipped. She told me she felt as though someone had cut a rope she hadn’t realized was binding her. I will return to her story in Chapter 3, because what happened to her is the clearest illustration of what the interface revolution actually means.
Another engineer, the most senior on the team, spent the first two days oscillating between excitement and terror. Excitement because the work was flowing at a pace he had never experienced. Terror because the pace forced him to confront a question he had been avoiding: If the implementation work that had consumed eighty percent of his career could be handled by a tool, what was the remaining twenty percent actually worth?
The answer, which he arrived at by Friday, was: everything. The remaining twenty percent, the judgment about what to build, the architectural instinct about what would break, the taste that separated a feature users loved from one they tolerated, turned out to be the part that mattered.
The tool had not made him redundant. It had stripped away the manual labor that had been masking what he was actually good at.
That is what vertigo feels like. The ground moving under your feet while you try to keep your balance.
The imagination-to-artifact ratio. That is my name for the distance between a human idea and its realization. When the ratio is high, only the privileged build. When the ratio is low, anyone with an idea and the will to pursue it can make something real.
Consider the trajectory. A medieval cathedral required hundreds of workers, decades of labor, and the resources of an entire community. The imagination-to-artifact ratio was enormous. The architect's vision required an army to execute. A modern building still requires a large team, but computer-aided design compressed the gap: What once took years of manual drafting now takes weeks of digital modeling.
Software development followed the same arc. In the 1960s, writing a program required understanding the machine at nearly the hardware level: assembly language, memory maps, interrupt vectors. The ratio was vast. Each layer of abstraction narrowed it. Compilers, high-level languages, frameworks, cloud infrastructure. A programmer in 2020 could build in a day what would have taken a team of ten a year in 1980.
But the gap remained. The programmer still needed to be a programmer. The translation cost had shrunk, but it had not disappeared.
Claude Code made the gap approach zero for a significant class of work. A person with an idea and the ability to describe it in natural language could produce a working prototype in hours. A working thing, with code that ran and interfaces that responded and logic that held up under testing. The imagination-to-artifact ratio, for the first time in the history of human tool use, had been reduced to the time it takes to have a conversation.
That is what the speed of adoption measured. Not how good the tool was. How deep the need was. Tools that satisfy an existing, urgent need are adopted at the speed of recognition. Tools that create a need are adopted slowly, through marketing budgets and institutional mandates. The speed of Claude Code's adoption tells us the need was already there, pressing against the constraints of every interface that came before, waiting for the barrier to break.
But the speed also measured something less comfortable, something the triumphalists tend to glide past on their way to the next milestone. The speed measured appetite. And appetite, once awakened, does not self-regulate.
The four percent of GitHub commits generated by AI in early 2026 was not a ceiling. It was a floor. And as the percentage climbed, a quieter question followed it.
If a junior developer using Claude can produce in a day what a senior developer without Claude produces in a week, what is seniority?
If a non-technical founder can prototype a product over a weekend, what is the value of the technical co-founder's decade of training?
These were the questions I heard at every dinner table and every conference in those months, asked with the specific anxiety of people who had bet their careers on skills that were commoditizing in real time.
And there was a third thing, harder to name, and it was the one that stayed with me longest. Awe and loss at the same time. Not the bright awe of discovery, and not the clean loss of displacement. A compound feeling, the way certain wines are described as having contradictory notes that should not coexist but do.
These people were not worried about being replaced. They were worried about something subtler and harder to articulate, and it took me weeks of listening to different versions of the same fear before I could name it: Depth itself was losing its market value, not because depth was less real or less valuable in absolute terms, but because the market, which is the system that converts value into livelihood, was recalibrating what it would pay for.
Breadth had become cheap. Competent performance across a wide range was now available to anyone.
Depth, the kind that takes years of patient immersion to develop, was still rare. But rare does not mean valued. Rare means valued only when the market has a use for it. And the market was discovering that, for most purposes, breadth was good enough. That the world might stop rewarding the journey to the bottom now that the surface was good enough for most purposes.
The machines would get better and better at depth. As it was, they were good enough at breadth that depth risked becoming a luxury the market would no longer subsidize.
When I started my journey, I wrote games in Assembler. Almost none of the developers that work for me could do that today, even though all the code they write ultimately gets converted to assembler running on the CPU of the computer, but it doesn't matter. My arcane knowledge of Assembler is not useful. The same is now happening to a Python developer.
I felt all of it: terror, excitement, but mostly awe. Often in the same hour. Sometimes in the same minute. I would build something extraordinary with Claude – a system that worked, that solved a real problem, that I would have never been able to do (I haven't coded in years) – and the exhilaration was genuine, physical, the kind that makes you want to call someone and tell them what just happened. Then I would look at the clock and realize four hours had passed and I had not eaten, and the exhilaration would curdle into something closer to distress, because I recognized the pattern: This is how compulsion feels from the inside, indistinguishable from passion until you try to stop.
I couldn't stop, and I was not alone.
Millions of other builders were feeling the vertigo of the orange pill at the same time, and crossing paths at random places with a look of recognition that we were “in the know” of the seismic shift that was happening around us for developers and soon for all knowledge workers. Thrilled at the creative opportunity and terrified of the societal implications for us and our kids.