A developer used Claude Fable, an AI language model from Anthropic, to create the majority of a software library. The special thing: The AI costs were less than 150 US dollars. `sqlite-utils` is a popular open-source library that simplifies SQLite databases.
This development is a direct blow to the gut of traditional software development. If an AI delivers complex, functional code for a fraction of the cost, it massively shifts the value of human labor. Companies can implement projects in the shortest possible time, while developers must ask themselves which tasks will still require their expertise in the future.
Simon Willison, a well-known developer, used Anthropic's Claude Fable to finalize version 4.0rc2 of his `sqlite-utils` library. The AI generated most of the code, which was then only reviewed and optimized by Willison. The total cost for interacting with Claude Fable amounted to **just $149.25** for the AI's computing power.
For freelancers and young developers, this means: The pressure is increasing. Routine coding tasks that previously secured fixed incomes will increasingly be taken over by AI. You must continue your education and focus on complex problem-solving or architectural tasks that still require human judgment. Those who only generate standard code could soon become obsolete.
Companies face a huge opportunity, but also a challenge. Software development costs can drop drastically, enabling new projects and faster market launches. At the same time, however, companies must ensure that AI-generated code is secure and error-free. Internal review processes will become more important than ever to avoid unexpected errors or security vulnerabilities.
The biggest opportunity lies in **accelerating development**. Teams can turn ideas into prototypes much faster or extend existing software. **Cost control** also becomes easier, as AI usage is transparent and predictable. Smaller companies and startups can thus tackle complex software projects with less budget, which were previously reserved for large companies.
The main risk is **quality assurance**. Even if Claude Fable delivered good code, blind trust is dangerous. Errors in the generated code can have expensive consequences, from security vulnerabilities to system failures. Another risk is **dependence on AI models** and the question of who is ultimately liable for errors. In addition, the market for junior developers, who often take on simple coding tasks, could shrink significantly.
