Claude Code creator: How a self-taught programmer reshaped the AI programming track
```
A self-taught programmer without a computer science degree is reshaping the competitive landscape of the AI programming track.
Boris Cherny, head of Anthropic's AI programming tool Claude Code, has transformed the product over the past year from a "unremarkable launch" into Anthropic's core growth engine, helping Anthropic catch up to and even surpass OpenAI in revenue.
Boris Cherny video screenshot
From $1 Billion to $30 Billion: A Steep Growth Curve
Numbers speak for themselves.
According to The Information, Claude Code's annualized revenue was $1 billion last December, rising to $2.5 billion in February of this year, with further acceleration recently. Cherny stated in an interview that this growth has driven Anthropic's overall annualized revenue to $30 billion.
This growth curve wasn't steep from the start. Claude Code was officially launched in February 2025, with mediocre initial reception. Cherny admitted: “It didn't take off immediately; it stagnated for a while.” The real breakthrough happened when a more powerful underlying model went online.
In May last year, Anthropic held its first developer conference. Cherny sat backstage between stage crew and lighting technicians, testing Claude Code with the newly released Opus 4 and Sonnet 4 models.
"All I remember was the feeling: 'Oh my god, this thing, it really works.'" he said. Afterward, the proportion of code generated by Claude in his work jumped from about 10% to nearly 30%.
Last November, with the release of Claude Opus 4.5, the product took another leap forward. Venture capitalist Tian Yu (Micron Ventures) said that it was from then that AI reached human-level abilities in software engineering—the new model enabled Claude Code to run longer, reason more efficiently, and make fewer coding errors.
A Grassroots Path of a Self-Taught Programmer
Cherny's personal journey itself is a winding tale.
Cherny is 34 years old this year. He started programming in secondary school by writing web code for eBay product listings, later entertaining himself by writing programs on calculators.
He attended the University of California, San Diego, majoring in economics, but ultimately did not obtain a degree—after dropping out, he founded a startup.
In September 2024, after nearly seven years at Meta, he joined Anthropic. At Meta, he led work on improving code quality. "I was drawn by Anthropic's mission to develop advanced AI for human benefit," he said.
In September 2024, he joined Anthropic as the first engineer for Claude Code.
Leaving Cursor, Returning Again
Before Claude Code's real breakout, Cherny had an episode.
Last July, Cherny and Claude Code product lead Catherine Wu both left the company, jumping ship to the AI programming tool startup Cursor—he became Chief Architect and Head of Engineering, Wu became Product Lead. Cursor was both a customer and a competitor of Anthropic, and the industry regarded this poaching as a major win for Cursor.
But the outcome was unexpected. Cherny said he realized it was a mistake just a few weeks later, proactively contacted Anthropic, and asked to return. The two soon resumed their original positions.
"Returning was 100% about missing that mission. Besides directly doing alignment research, I can't think of anything more urgent or important than what I'm doing now," Cherny said.
"Alignment research" refers to the field of aligning AI model goals with developers' intentions.
Reshaping the Market, Impacting Traditional Software Stocks
The commercial value of Claude Code has already been validated by leading companies. According to Anthropic, Netflix, Accenture, and Uber have all become its customers.
Cherny estimates that the software engineering market targeted by Claude Code exceeds $1 trillion.
The rise of this product has also cast a shadow on the capital markets. Salesforce, ServiceNow, Workday, and other traditional software companies' stock prices are under pressure, prompting the market to reassess their long-term business models. At the same time, concerns about software engineers being "replaced" are spreading throughout the industry.
On this, Cherny believes programming is still "a highly practical skill," but he also predicts: "In a year's time, we'll see a world where anyone can do software engineering. Just as programming is solved for me, it will be solved for everyone."
In actual application, Claude Code's capabilities have also impressed engineers. Rob Sun, founding engineer at robotics company Dexterity, completed a computer vision project using Claude Code in just a few days.
"This was a series of problems I never thought I could solve alone," he said, "if it were just me, it might take three continuous years."
Next Step: From Coding to “Doing Everything”
Cherny and his team are extending Claude Code's automation capabilities beyond programming to domains like legal and financial tasks, with the corresponding product being Claude Cowork.
The team is itself a deep user of Cowork. Cherny has Cowork monitor employees' weekly progress reports; if someone misses a report, Cowork automatically sends a reminder on Slack. "I was just getting myself a cup of coffee," he said.
He's also using Claude for booking flights and handling parking tickets. Last month in Seattle, he had Cowork autonomously navigate the local Fish & Wildlife Department’s website, paid $7, and successfully applied for a shellfishing permit.
The Claude Code team remains lean. Cherny estimates the team's per capita revenue is the highest in the world, but refuses to disclose specific numbers. The final version of Claude Cowork was developed by the team using Claude Code in just 10 days.
Competition and Risks: The Moat Is Not Impenetrable
The leading position of Claude Code is facing pressure.
OpenAI has shifted its resources to its programming tool Codex; some developers report Codex is more efficient than Claude Code in certain situations, and some even use both together.
Last month, Anthropic had an unexpected leak—Claude Code's underlying software was accidentally made public, giving competitors a rare glimpse into how it works and plans for unreleased features, including making Claude Code more proactive.
Additionally, Anthropic’s latest model Mythos, which has shown outstanding performance in programming tests, is not publicly released due to its perceived potential for network attacks.
Risk Warning and DisclaimerThe market has risks; investment needs caution. This article does not constitute personal investment advice, nor does it take into account the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular situation. Investing based on this article is done at your own risk. ```