AI Is Changing Who Gets to Do Math and Software

Two threads circled the same question from different angles. The 'After Math' thread asked what AI means for mathematics as a field, with one commenter arguing that merely providing a certified answer is not the same as solving a problem, because the insight is missing. Another pushed back: AI might make mathematics more ambitious by lowering the cost of exploring risky ideas, the same way it does in software.

The 'Don't be the out of touch Kung Fu master' thread made the software version explicit. The argument: we are moving from line-by-line mental model transcription to a world where engineers focus on data structures and architecture. John Carmack's skepticism was cited as a counterpoint, and the community was split. Some people find AI-assisted coding deeply fulfilling. A thread about a 60-year-old who said Claude Code re-ignited their passion for building got warm responses. Others said it felt like getting an A on a test you didn't take.

The local model thread added a practical dimension: people are actively trying to replace Claude and GPT with local models for daily coding, with mixed results. Speed is the main bottleneck for local setups, but Qwen and DeepSeek are closing the gap.