Coding Is Not Solved – Alex Ewerlöf Notes
3 hours ago
- LLMs struggle with logic and non-functional requirements like security and scalability.
- AI cannot be held accountable for mistakes, unlike human engineers.
- Coding is not solved; LLMs are useful tools but have limitations, especially with large contexts.
- AI-generated code may be cheaper but doesn't guarantee quality or reliability.
- The economics of software favor SaaS for critical tasks, while AI is good for personal projects and POCs.
- AI output is 'Nordic Gold' – cheap but not always the real thing; engineers offer accountability and consistency.
- Be cautious with cloud AI due to data privacy concerns and potential misuse.
- New roles like AI managers and quality engineers will emerge, but understanding and ownership remain crucial.