If We're 10x More Productive with AI, Why Are Coding Agents Still Bad?
If we are at 10x with AI and near AGI or ASI, then how is it possible that these products (Codex, Claude Code CLI) are still such garbage?
Not sure how near we are to AGI or ASI. But I've been using Claude CLI, and it really produces quality output. To build a quick first version of a product, it does a good job. The real complexity starts after that — and the bottleneck now, I think, is on the human: what we actually want to build with them.