ianhxu

It is too difficult to strictly prevent the model from being used for any unsafe purpose. The same thing can be used for completely different purposes as long as it is described differently.

Claude Code and Codex are my daily drivers, and I often run them side by side on the same task to compare. For me Claude Code gets something working faster, but I burn the 5h quota on the $200 Max plan really fast. Codex I tend to trust a bit more on the careful diffs. The bigger change for me wasn't the tool though, it was writing a spec doc first (features/UX, technical, language-specific) before either agent touches code, then reviewing every diff. I still set up a lot of the harness by hand. How are people automating worktrees and parallel sessions?

>frontier models are more capable than the latest from DeepSeek. But is the capability difference enough to justify a 30x price difference?

The contradiction here is that without frontier models, there'd be no foundation for models like DeepSeek to reference and catch up to. Is there an economic model that captures this kind of dynamic?

I don't have a quantitative way to argue this, but my intuition says that for humans to build something that matches human capability across every dimension would require a breakthrough at the physical level — and such a breakthrough may itself be bounded by the limits of humans as observers.

That said, this goal might itself be a non-goal. AI is going to be — or already is — more powerful than any individual human in many ways. But what my intuition points to is that humans will still have plenty of interesting work to do, like the author's example of handwriting code: it shifts from being scalable value creation into a form of craftsmanship.