Why Anthropic’s Bun Acquisition Proves Software Engineering Isn’t Dead¶
“Software engineering is done.”
— Adam Wolff, Anthropic Engineer, November 2025“We’re acquiring 14 engineers and paying hundreds of millions to maintain a JavaScript runtime.”
— Anthropic, December 2025
One of these statements is a lie. The other is a check that cleared.
The Acquisition That Exposed Everything¶
On December 2, 2025, Anthropic made its first-ever acquisition: Bun, a JavaScript runtime with zero revenue, for a reported price in the low hundreds of millions of dollars. The same day, they announced Claude Code had crossed $1 billion in run-rate revenue.
Let that sink in.
A company whose employees publicly declare software engineering “done” just made its largest financial bet ever on… software engineering infrastructure. They didn’t fork Bun’s MIT-licensed codebase and let Claude rewrite it. They didn’t use their supposedly all-powerful AI to build their own runtime. They wrote a massive check to acquire 14 human engineers and the expertise they represent.
If AI could truly replace software engineers in “three to six months” (as CEO Dario Amodei predicted in March 2025) why would Anthropic spend hundreds of millions on human talent they’ll supposedly never need?
Because they know it’s not true.
Follow the Money, Not the Tweets¶
The tech industry has a pattern: executives predict automation will eliminate jobs while simultaneously hiring for those exact roles. Watch what they do, not what they say.
Anthropic’s current job board: 200+ open positions, including Software Engineers for Claude Code, infrastructure engineers, and developer tools specialists.
Anthropic’s job applications: Include a checkbox requiring candidates to certify they “will not use AI in the application process”, to assess “non-AI-assisted communication skills.”
Anthropic’s internal research (December 2025): Engineers can “fully delegate” only 20% of work to Claude. Not 90%. Twenty percent.
Jarred Sumner, Bun’s creator, understood this reality. When asked why Bun needed to be acquired rather than replaced, he was blunt: “If Bun breaks, Claude Code breaks.” The AI that supposedly makes developers obsolete depends entirely on infrastructure built and maintained by developers.
The Real Transformation: Not Replacement, Evolution¶
Here’s what the “software engineering is dead” prophets miss: AI doesn’t eliminate developers. It transforms what developers do.
Consider what actually happened at Bun post-acquisition:
“Over the last several months, the GitHub username with the most merged PRs in Bun’s repo is now a Claude Code bot.”
— Jarred Sumner
A Claude bot is the most active contributor. And yet Anthropic acquired the entire human team and is expanding it. Why?
Because someone needs to:
- Review what the bot produces
- Architect systems the bot can’t conceptualize
- Make judgment calls the bot can’t make
- Debug failures the bot creates
- Decide what should be built in the first place
The bot is a tool. The humans wield it.
This is the pattern we’ve seen with every technological revolution. The printing press didn’t eliminate writers. It created demand for more content. Excel didn’t eliminate accountants. It made them handle larger, more complex portfolios. CAD software didn’t eliminate architects. It let them design buildings that were previously impossible.
AI coding tools follow the same pattern. They don’t replace developers. They amplify what developers can accomplish.
The New Developer: Not Obsolete, Upgraded¶
The developers of tomorrow won’t write every line of code by hand. But they’ll need skills that matter more than ever:
1. System Thinking¶
AI can generate code. It cannot understand why a system should exist, how it fits into a business, or what tradeoffs matter. That requires human judgment, domain expertise, and strategic thinking.
2. AI Orchestration¶
The most productive developers will be those who know how to work with AI. When to delegate, when to intervene, how to prompt effectively, how to validate output. This is a skill. It requires practice. It requires expertise.
3. Quality Judgment¶
Anthropic’s own research found engineers worried about “reduced collaboration” and “becoming irrelevant.” But the solution isn’t to abandon AI. It’s to become the person who ensures AI output meets human standards. Someone has to be the quality gate.
4. Infrastructure Mastery¶
The Bun acquisition proves infrastructure matters more in an AI world, not less. As Sumner wrote: “If most new code is going to be written, tested, and deployed by AI agents: The runtime and tooling around that code become way more important.”
Who builds that infrastructure? Developers.
Why the “10x Engineer” Becomes the “100x Engineer”¶
Y Combinator CEO Gary Tan claimed “vibe coding” lets “10 engineers do the work of 50 to 100.” He meant this as a warning to developers. But read it again.
Those 10 engineers still exist. They’re still employed. They’re still necessary.
What changed is their output. Each developer is now more valuable, not less. The 10 engineers doing the work of 100 don’t get fired—they get paid like they’re worth 10x their previous salary, or they tackle problems that were previously impossible.
When Zoho founder Sridhar Vembu responded to Tan’s claims that his company would be “competed away by vibe coding,” he pointed to reality: Zoho is experiencing “rapid customer growth exceeding 50%.” Why? Because “vibe coding just piles up tech debt faster and faster until the whole thing collapses” without real engineering discipline.
The developers who master AI tools become force multipliers. The developers who ignore AI tools become less competitive. But the profession itself? It’s not going anywhere.
The Historical Pattern No One Mentions¶
Every generation produces prophets who predict the end of programming:
- 1950s: “Machine code will eliminate the need for programmers”
- 1960s: “COBOL is so simple, businesspeople will write their own software”
- 1980s: “4GLs will replace traditional programming”
- 1990s: “Visual Basic means anyone can be a developer”
- 2000s: “Offshore outsourcing will eliminate domestic programming jobs”
- 2010s: “No-code tools will make developers obsolete”
- 2020s: “AI will write all the code”
The U.S. Bureau of Labor Statistics tells a different story: software developer employment grew from 1.4 million in 2019 to 1.9 million in 2024, with projected growth of 17% through 2033, far exceeding average occupational growth.
Each wave of automation created more demand for developers, not less. The tools changed. The skills evolved. But the need for humans who understand systems, solve problems, and build things? That need only grew.
What Anthropic’s Actions Actually Tell Us¶
Strip away the hype and examine the Bun acquisition strategically:
| What They Said | What They Did |
|---|---|
| “Software engineering is done” | Acquired 14 engineers for hundreds of millions |
| “AI will write 90% of code in 6 months” | 6 months later, their own research shows 20% delegation |
| “You won’t need to check generated code” | Require human engineers to maintain the runtime Claude depends on |
| “Vibe coding replaces traditional development” | List 200+ engineering positions on their job board |
The gap between rhetoric and reality isn’t subtle. It’s a canyon.
Anthropic isn’t investing in Bun because they believe engineering is over. They’re investing because they understand that AI infrastructure requires world-class engineering and that’s a bet on human talent, not against it.
The Path Forward for Developers¶
If you’re a developer anxious about AI, here’s the honest truth:
Yes, your job will change. The tasks you do daily in 2025 won’t be the same tasks in 2030. Some of what you do now will be automated. That’s reality.
No, you won’t be replaced. You’ll be upgraded. The developers who thrive will be those who:
- Learn to work with AI as a collaborator, not a competitor
- Develop judgment that AI lacks understanding context, making tradeoffs, knowing when the AI is wrong
- Build expertise in areas AI amplifies rather than replaces: architecture, systems design, problem framing
- Stay curious about new tools rather than fearing them
The most active contributor to Bun’s codebase is now an AI bot. But that bot runs on infrastructure maintained by humans. It’s reviewed by humans. It’s directed by humans. The roadmap is set by humans.
AI is a tool. The most powerful tool developers have ever had. But a tool nonetheless.
The Check That Tells the Truth¶
Dario Amodei can tweet whatever he wants about software engineering being “done.” His employees can post provocative threads about never checking AI-generated code.
But when it came time to secure Claude Code’s future, Anthropic didn’t trust their AI to solve the problem. They wrote a check for hundreds of millions of dollars to acquire human engineers.
That check is the truth.
That check says: Human expertise matters. Engineering talent is valuable. Developers aren’t going anywhere.
The only lie is the one suggesting otherwise.
The developers who learn to wield AI as a tool will build the future. The developers who believe the hype and abandon the craft will miss out on the most exciting era in software history. Don’t let fear-mongering tweets cost you a career that’s only becoming more valuable.
Software engineering isn’t dead. It’s evolving. And evolution rewards those who adapt.