In early 2026, Boris Cherny, the primary creator of Claude Code, made a statement during a public conversation that left the entire developer community in silence: "For me, programming has been solved."
He mentioned that he uses his phone daily to orchestrate hundreds of AI agents, allowing them to handle coding, reviewing, and even communicating with each other autonomously, while he no longer writes a single line of code himself. This is not just a personal experiment by a tech geek—it is an engineering paradigm being rolled out on a massive scale by Anthropic, a company with a valuation approaching a trillion dollars. Co-founder Dario Amodei subsequently confirmed that engineers within Anthropic have almost entirely stopped writing code by hand. Instead, they accomplish their work by managing vast AI agent systems, with each individual's output being two to three times higher than before.
The true weight of this event lies not in the fact that an AI company is using AI to write code, but in the fact that Anthropic is using its own operations to pioneer a completely new path for value creation across the entire business world.
01. Efficiency Reborn: From Linear Growth to Exponential Leap
To understand the commercial value of AI coding, the most intuitive entry point is naturally efficiency. Traditional improvements in development efficiency are often linear: faster compilation tools shave off dozens of seconds, better frameworks save some repetitive coding, and smoother collaboration processes reduce a few rounds of rework. These are all percentage-level optimizations, climbing upward along the same productivity curve.
What AI coding brings, however, is an entirely new curve. When a developer no longer types out business logic line by line, but instead describes their intent in natural language to have the AI generate complete modules; when unit tests, API documentation, and boilerplate code—tasks that once required immense effort—are completed in an instant; when technical research no longer means shuttling back and forth between documentation and forums, but rather engaging in multi-turn dialogues directly with the AI—the overall output of individual developers and teams enters a state of near-exponential leap. This efficiency boost is no longer just about being "a bit faster"; it transforms what was previously impossible to do at high frequency into an everyday occurrence.
The more profound significance lies in the fact that the quantitative change in efficiency is triggering a qualitative change in business decision-making. When the cost of trial and error becomes extremely low, enterprises dare to explore more possibilities. What used to be limited to validating three solutions in a quarter can now see thirty solutions validated in parallel. For the first time, the iteration speed of business strategies may be able to keep pace with the speed of market changes. This is the true commercial dividend behind efficiency.
02. Cost Restructuring: From Fixed Heavy Assets to Flexible Light Assets
The next chapter in the efficiency story is often cost. However, the impact of AI coding on cost structures is far more profound than simply laying off employees or reducing developer salaries. It is redefining whether the very activity of software production is a heavy-asset or light-asset endeavor.
In the past, founding a competitive software company required a technical team of a certain scale, which was almost an entry ticket. You needed front-end, back-end, data, operations, and testing personnel, with enough staff in each role to cover the workload. Even for a Minimum Viable Product (MVP), a fully-fledged small team was often necessary, which determined the threshold for startup capital.
The emergence of AI coding has changed the connotation of the "developer" role. An engineer equipped with architectural thinking and product sense can, with the aid of AI, cover the production capacity that previously required three to five people. These engineers; they are transforming into versatile generalists.
This means the size of the minimum viable team is shrinking drastically. We are beginning to see more and more "one-person startups," or "solo founders," building products that previously required angel round funding to launch. They are not superhumans; they have simply offloaded a massive amount of execution-level work to AI, focusing themselves on defining problems, designing experiences, and steering direction. For large enterprises, this equally implies a more flexible cost structure. There is no longer a need to rapidly expand a team for a short-term project only to face the subsequent challenge of personnel placement, nor is there a need to be constrained by the scarcity of talent in niche tech stacks. AI programming makes technical capabilities available like cloud services—callable on demand and elastically scalable.
Now, as AI significantly raises the baseline for coding efficiency, the advantage of low-cost labor is being eroded. Value begins to migrate to deeper levels: the ability to understand customer businesses, design solutions, and navigate complex integrations becomes more valuable than simply amassing code executors. The value distribution across the entire industry chain is being readjusted by an invisible hand.
03. Knowledge Democratization: Truly Integrating Business Language with Technical Language
Furthermore, in the past business world, there was an ancient chasm spanning between "what can be conceived" and "what can be built." Business personnel are filled with insights into the market, users, and processes, yet they must translate these insights into requirement documents, which then pass through product managers and designers, and are finally handed over to engineers to be converted into code.
Low-code and no-code platforms have been promising "empowering business personnel" for many years, but they have always faced a ceiling of flexibility—where templates run out, innovation stops.
AI coding, however, provides a completely different path: it allows people to directly manipulate logic using natural language. A marketing director can generate a script for automated user data processing by describing their requirements without needing to learn Python; a supply chain manager can articulate their inventory optimization logic to the AI and instantly see a runnable prototype.
When the translation cost between business language and technical language approaches zero, the speed of innovation within enterprises will undergo a fundamental change. Previously, an idea for business optimization might have to wait in the technical department's queue for weeks; now, the time gap between idea and validation is compressed to the time it takes to drink a cup of coffee. This change will breed a completely new organizational culture: people no longer default to "we need to get tech to implement that," but rather "let's just run it and see." Once this "just-do-it" spirit enters the organizational DNA, business agility will no longer be just a slogan in consulting reports.
More importantly, it changes the mechanism for discovering business opportunities. In the past, many requirements hidden in gold mines remained undiscovered because business personnel simply could not imagine what a technical solution might look like, leaving them unable to propose it. Now, when AI can instantly transform vague requirements into visual interfaces or interactive prototypes, the boundaries of imagination are opened. A customer service supervisor might accidentally stumble upon a completely new customer operations tool during a conversation with the AI. This ability for "serendipitous discovery" is the most desired yet hardest to systematically cultivate source of innovation for enterprises.
04. Asset Evolution: Beyond Code, the Model is the Moat
And if a company's core software is assisted or even primarily generated by AI, where lies the company's business moat?
This is a question that all business leaders must answer in the AI coding era. One of the traditional moats is the code assets themselves—massive and sophisticated codebases that carry years of business logic, which are difficult for competitors to replicate. But as the cost of code generation plummets, this moat is drying up.
The true value begins to settle elsewhere: in the knowledge itself used to generate the code. This includes extremely profound descriptions of the business domain—the business insights embedded behind prompt engineering, those structured business rules, encoded decision logic, and the continuous feedback loop of using business data to fine-tune AI models.
Future enterprise software assets will no longer be merely a static code repository, but a living "business model + AI generation pipeline." Code is just the instantaneous output of this process, just as a paper map is the instantaneous output of geographical information. What is truly valuable is the geographical data behind it, not the paper itself.
Consequently, a new dimension of competition emerges. Enterprises must not only manage code versions well but also manage "intent versions"—why business rules were defined in a certain way initially, what context prompted the AI to generate that specific logic, and how to have the AI regenerate code that better fits the current situation after the business environment changes.
05. Conclusion: The Return of Value
Looking back from this moment, the software industry has gone through two distinctly different stages. In the early days, software was a rare commodity, priceless; later, the internet brought prosperity, making software accessible at our fingertips, but the ability to write software remained a scarce resource, locked in the minds of a few.
The true commercial value of the AI coding era is not merely about improving efficiency or reducing costs once again, but about completing the final step of this long journey—to the hands of everyone with the will to create.
- End -