● AI Coding Shock, Engineers Obsolete, Builder Era Begins
The content I am sharing today is based on a podcast conversation between Boris Cherny, Head of Claude Code at Anthropic, and Lenny, revealing shocking insights on how AI is completely upending software development and the future of jobs.
It goes beyond simply saying “coding has become easier”—it contains predictions that the very definition of engineering is changing, and the way we plan and build products will be completely different in 6 months.Reading this will provide clear guidelines on how developers, PMs, and non-technical roles can survive in the AI era.
Now, let’s get straight to the core points.
[Shocking Analysis] “I No Longer Write Code Myself” The End of Development and New Opportunities According to Anthropic’s Claude Code Lead
1. Human-Written Code is Over: The Era of 100% AI Coding
This is a truly shocking confession.Boris Cherny, who leads the Claude Code team at Anthropic, revealed that since last November, he has not written a single line of code by hand.Currently, 100% of the code he writes is generated by AI (Claude), and he processes between 10 to 30 Pull Requests daily.
What this means is that even the world’s best AI developers no longer type code themselves.Boris stated that engineer productivity has increased by over 200%, declaring that the act of coding itself is now close to a ‘Solved Problem.’
Since there is no longer a need to struggle with “Minutia” like syntax or library dependencies, he says development has actually become much more enjoyable than before.Now, the important skills are not coding techniques, but the planning ability and judgment to decide what to build.
2. The Extinction of the ‘Software Engineer’ Title and the Birth of the ‘Builder’
Boris predicted that by the end of this year, the job title ‘Software Engineer’ will start to disappear.Instead, that position will be replaced by concepts like ‘Builder’ or ‘Product Manager (PM)’.
The era has arrived where anyone can build software just by giving commands in natural language.We are seeing a flood of cases where data scientists with no development knowledge open a terminal and order AI to perform complex analyses, or designers directly generate code to create apps.
This implies that the ability to utilize AI Agents has become essential for tech industry professionals.In the past, development, design, and planning were strictly separated, but now those boundaries are collapsing.You must keep in mind that as everyone becomes able to code, a world where “everyone is a PM and everyone is a coder” is coming.He did not forget to warn that if you fail to adapt to this change, whether you are a developer or a planner, you will suffer greatly.
3. Anthropic’s Secret Weapon: ‘Latent Demand’ and ‘Co-work’
There is a point we absolutely must learn from how Anthropic builds products.It is the ability to identify ‘Latent Demand’.Boris’s team got hints from seeing users use Claude Code differently than intended (coding)—in other words, seeing them ‘Abuse’ it.
Some users were asking how to grow tomatoes, and others were using the coding tool to recover wedding photo files.Seeing this, Anthropic realized that non-developers also want AI to manipulate their computers on their behalf, so they created the ‘Co-work’ feature.Co-work is not just a chatbot that converses; it is an acting agent that opens Excel on your computer, sends emails, and sends Slack messages.
The business insight to be gained here is clear.If you productize “what the AI model wants to do on its own, and what users are trying to force it to do,” you will hit a jackpot.They say this Co-work feature was built in just 10 days with AI writing all the code. It is an era where speed is life.
4. The Immutable Law of Building AI Products: Bet on the Model 6 Months from Now
The most important advice Boris gives to founders and developers is “Do not build products based on current AI models.”Right now, AI might seem to make mistakes or look imperfect, so it is easy to tack on complex workflows or safety measures.However, the model coming out in 6 months will be much smarter and possess more general capabilities than the current one.
He advises to “Imagine the model 6 months from now and design the product.”Product-Market Fit (PMF) might not match right now, but the moment the new model is released, your product will soar.Also, he emphasizes not to use cheap models to save costs, but to use the highest performance models (like Opus 4.6).Because smart models handle tasks in one go, they may result in lower token costs than fixing the mistakes of dumb models multiple times.This is why companies following AI trends must focus on performance optimization rather than cost reduction.
5. The Core Point No One Else Talks About: Anthropic’s Safety Philosophy (Mechanistic Interpretability)
Finally, the point where Anthropic differentiates itself from other AI companies is its approach to ‘Safety’.Usually, companies train AI from the outside so it doesn’t say bad things.However, Anthropic uses ‘Mechanistic Interpretability’ technology, which looks into each neuron of the AI like neuroscience.
It means they directly monitor which parts of the “brain” are activated when the AI tries to lie.And to verify safety, they intentionally Release Early to collect data on how the AI behaves in the ‘Wild’, rather than in a controlled lab.This can be seen as the reason Claude is trusted by enterprises and the technical moat that becomes the foundation for us to trust AI with work.
< Summary >
- The End of Coding: The Anthropic lead codes 100% via AI and does not write code directly. Productivity increased by 200%.
- Change in Roles: The title ‘Software Engineer’ will disappear, and everyone will become a ‘Builder’ or ‘PM’ who creates software.
- Latent Demand: New business opportunities (like Co-work) must be captured from users using AI tools in unexpected ways (Abuse).
- Development Strategy: Products should be designed based on smarter models to be released in 6 months, not current models.
- Cost Efficiency: Using top-performance models is more economical in the end as it reduces correction costs compared to cheap models.
[Related Posts…]
Future of Coding: Will Engineers Disappear?AI Agents: The Next Big Shift in Productivity
*Source: Lenny’s Podcast
Leave a Reply