It would have been worth it if the frontier models were open weight. Right now, if you invest time in mastering tools like Claude Code or Google’s Antigravity, there is no guarantee that you won’t be removed from their ecosystems for any reason, which would make your efforts and skills useless.
If there is a skill to using LLM coding agents, I think it is mostly just developing an intuitive sense for how to prompt and the “jagged frontier” of LLMs.
IME, the tools are largely interchangeable. They are all slightly different, but the basics of prompting and the jaggedness of the frontier is more or less the same across all of them.
Switching from codex to claude code is orders of magnitude simpler than switching from c# to java or emacs to vim.
AI is definitely a solid reason. Even a 10% increase in developer efficiency translates to roughly 9% fewer workers needed to do the same job. For AI to be cost effective, it must reduce headcount.
I have noticed this while working in a startup, how efficiency gains impacted work in the last 3 months with better coding models coming in. If you have someone senior that is effective at using these tools and can own the outputs and be capable of correcting things that are sloppy, you are immensely more valuable than 2-3 more developers in the same area of work. It's actually faster to empower fewer people than try to have 3 fast guys where there would be coordination overhead, which would become a bottleneck and bring down a lot of the efficiency gains. A good full stack engineer who can work with these tools at speed with caution is more valuable similarly as it requires less coordination. 3 junior devs shipping 90% good code and 10% slop would make the senior who is reviewing everything the bottleneck.
> I expect two of the listed to have very positive sentiment, and one generally negative in 2025.
You are quite correct! Crafting Interpreters actually has the highest average sentiment score across all books with more than 10 comments.
This is the average sentiment score of all three( range being -10 to 10) :
2012 : https://news.ycombinator.com/item?id=3987566
2014 : https://news.ycombinator.com/item?id=7367258
2021 : https://news.ycombinator.com/item?id=26837152