Learning Programming in an Age of LLMs
3 hours ago
- A reader built a large AI-assisted system but realized it exceeded their understanding, leading to struggles with production development.
- Mark Seemann is skeptical about AI's impact, fearing mass unemployment among knowledge workers, but hopes to be wrong.
- He is glad he learned programming before LLMs, as foundational skills remain valuable; he suggests considering trades if starting today.
- Seemann advises that people have long worked on abstractions they don't fully understand, but should grasp the layers below and above their work.
- He recommends systematically learning fundamentals for those with weak foundations, though acknowledges this might be too slow in the current era.
- Seemann learned programming slowly through trial and error, books, and examples, and now picks up new languages quickly from existing code.
- He uses LLMs mainly for falsifiable questions (e.g., code optimization) but distrusts them for open-ended advice due to bullshitting.