In The Loop Episode 7 | Vibe Coding: Will Developers Be Out Of A Job In Six Months? Dario Amodei’s Take

This week, Dario Amodei, CEO of Anthropic, the creator of the Claude language model, made a bold prediction that within six months, 90% of all code will be written by AI.

Simultaneously, the concept of "vibe coding" has surged in popularity, becoming one of the most searched terms in AI and overtaking "prompt engineering" on Google Trends. Thousands of videos, blogs, and discussions are exploring this new approach to building software.

Additionally, AI startups like Cursor have experienced explosive growth, with Cursor reportedly achieving a 9,900% increase in 12 months, making it one of the fastest companies to reach $100 million in revenue.

Clearly, the way we build software is evolving, but the speed of this transformation and its implications for developers remain hotly debated topics.​ This is one of the most competitive and significant implementations of AI: technology that can build technology.​

In today's episode, I'll discuss Dario Amodei's claim, the rise of vibe coding, and what these developments could mean for business and technology. This is In The Loop with Jack Houghton.

‍

‍

Dario Amodei’s take on the future of developers

Let's start with Dario Amodei's statement. He shared his vision for the future of programming at a Council on Foreign Relations event. Have a look:

‍

Dario Amodei claimed that nearly all code will be written by AI within a year—a significant assertion.

It's important to approach this with caution, as Anthropic is currently finalizing a $3.5 billion funding round, which has tripled its valuation to over $60 billion. A key aspect of their strategy is positioning Claude, their language model, as the preferred choice for programming tasks. Many tools built on language models are selecting Claude for code generation, aligning with Anthropic's investment narrative. Additionally, they've introduced Claude Code, a coding agent, which benefits from the media attention surrounding such bold claims, potentially boosting their valuation and market share.​Financial Times

Financial metrics suggest that this level of AI adoption is ambitious but conceivable. Anthropic's valuation soared from $18.5 billion in February 2024 to over $60 billion by March 2025, with revenue increasing from $100 million to $1 billion in the same period. The company is backed by major players like Google and Amazon, with Amazon's Alexa integrating Claude as its language model. This has generated excitement among investors, as automating software development could significantly reduce the time and expense associated with engineering.

However, there are critical counterarguments to this narrative, which we'll explore throughout this episode.

‍

‍

What is vibe coding?

A key reason this interview has gained traction is the rise of "vibe coding," a concept that has gained global attention over the past three to four weeks. Vibe coding has become a more popular search term than "prompt engineering," indicating its rapid ascent.​

The term vibe coding was introduced by Andrej Karpathy, a notable figure in the AI space. He describes it as a new style of coding where one fully embraces the capabilities of AI, allowing the language model to handle the coding details. Karpathy mentions using tools like Cursor Composer, combined with voice-to-text applications such as SuperWhisper, to interact with the AI without manual coding. He admits that while this approach is effective for quick projects, it may result in code that is complex and requires careful review.

This concept quickly caught the attention of entrepreneurs, investors, and media outlets, particularly those affiliated with Y Combinator. Y Combinator’s CEO discussed whether vibe coding represents the next significant shift in programming. In a survey of Y Combinator entrepreneurs, some founders expressed feeling less attached to their code, given that AI can generate similar features in minutes. This marks a departure from the traditional pride developers take in crafting meticulous codebases.​

The discussion also touched on how vibe coding might alter hiring practices and technical interviews for developers. Instead of focusing on data structures, interviews could involve practical exercises where candidates use AI environments to build applications quickly, emphasizing their ability to produce functional outcomes efficiently. There's a narrative within the Y Combinator community that the ability to progress from an idea to a functional product is accelerating exponentially. This means that features or use cases can be developed and presented to users rapidly.​

As vibe coding gains popularity, developers may transition into roles akin to agent management. Some reports suggest that AI increasingly handles code reviews and generation.

While it's tempting to view this as the end of traditional programming, it's crucial to consider criticisms. The technology is still in its early stages. Some users have found tools like Cursor to be limited, working well for small experiments but struggling with large-scale projects or regulated industries like finance or healthcare. Concerns about security and debugging persist, as AI may not always provide reliable fixes for complex errors. Additionally, non-technical individuals might overestimate their ability to build functional software without a solid understanding of coding principles, potentially leading to unmanageable codebases.​

While these issues are likely to improve over time, they don't fully address the question of the future role of developers, which we'll delve into further in this episode.

‍

‍

The ever-changing role of a programmer—Tim O’Reilly’s take

Tim O'Reilly makes one of the most compelling arguments in this debate, and I want to take you through it because it’s a fascinating look at history—something I love. He explores the evolution of programming as a continuous process of abstraction and automation.

From the very beginning, computing has been about simplifying complexity. Early programmers had to connect circuits to make a computer execute a calculation physically. Then came machine code—binary instructions entered manually via switches. Assembly language simplified this by allowing programmers to write human-readable commands that the machine could understand. Later, high-level languages like C and Java abstracted away even more complexity, making it possible to write software without worrying about low-level hardware interactions.

O'Reilly points out that each of these advancements has sparked claims that programming as a profession was coming to an end. Yet, in every instance, the opposite happened—demand for programmers increased. Why? Because making programming easier expands what’s possible. More people could now create software, and businesses could automate processes that previously seemed too complex.

This pattern kept repeating. The rise of consumer operating systems like Windows and macOS further reduced low-level work. Developers no longer had to write specialized drivers for hardware; operating systems provided APIs to handle that. Far from eliminating developers, this shift created more of them—building more applications, specializing in new areas, and meeting the growing demand for software.

Then came the web, introducing HTML, CSS, and JavaScript—languages still in use today. Website builders like WordPress made it possible for non-coders to create entire websites. But rather than making programmers obsolete, this explosion of accessibility increased demand for specialized development—back-end engineering, security, UX, and performance optimization.

This connects to an economic principle I’ve mentioned before: elasticity of demand. When the cost of producing something falls, demand for it increases. In software, every automation, every abstraction, makes experimentation cheaper. That means more people attempt new ideas, which in turn creates demand for more specialized developers—those focused on security, AI, payment systems, or enterprise software.

So, if AI-based coding reduces the cost and time required to generate code, what’s the likely impact? More software projects, not fewer. The role of developers will shift, not disappear.

And this brings us to a historical analogy that O'Reilly brilliantly references: the Industrial Revolution.

Early textile mills displaced skilled artisans who wove fabric by hand. But over time, new roles emerged—machine operators, engineers who built better looms, and entrepreneurs who scaled production. Cloth became cheaper and more widely available, expanding markets and creating new business opportunities.

Similarly, AI might automate parts of coding, but it will also unlock new possibilities. Developers will focus more on high-level problem-solving, system design, and integrating AI into workflows. Just like in past technological revolutions, the nature of work will evolve, but the demand for skilled people will persist.

Satya Nadella talked about this too in a recent interview that I covered in a previous episode of In The Loop. Listen to it here.

It took about 50 years for the innovations of the Industrial Revolution to fully reshape the economy and drive wages up. Technological shifts don’t happen overnight, despite what headlines might suggest.

However, this time, things could move much faster. With the internet enabling rapid distribution and AI being embedded into every layer of software, the transformation will be accelerated—but it will still take time.

In software, automation has always followed the same pattern: removing barriers. Memory management, hardware drivers, and software interfaces were once major constraints, but as those obstacles disappeared, developers shifted to higher-level, more strategic work. The same will happen now. AI might free us from simple code generation, just as operating systems freed us from manually managing hardware.

‍

‍

Closing thoughts

So, rather than the “end of programming,” we should see this as the reinvention of programming. The real question is: What new layer of complexity will emerge? Massive computational challenges remain unsolved—climate modeling, real-time disease tracking, large-scale education solutions, and advanced robotics. Automating code doesn’t solve these problems, but it does make it cheaper and easier to attempt solutions.

I feel privileged to be part of this transformation and excited to see what unfolds over the next decade. Whether it’s in labs, startups, or ambitious teams, people will push the boundaries of what’s possible.

And while news headlines love to claim that “X changes everything,” it’s more accurate to say that “X changes the trajectory of something.” AI isn’t an endpoint—it’s a shift in how technology is built, and that shift will shape the world for the next 20 years.

To close, I want to leave you with something fun—a few clips from children in the 1960s predicting the future. It’s a great reminder of how our expectations evolve over time.

‍

Do me a favor: share this episode with a friend, a colleague, or even your pet cat—whoever you think might enjoy it. Word of mouth helps spread the best ideas, and I truly appreciate your support.

‍