Remember how Meta monitored employee activity closely for a few months, and then had a bunch of layoffs related to AI efficiency? (oh right that was like 3 days ago).
What if you could do the same thing to your workforce of Claude Code agents? Aka use Claude Code to observe and distill how to make a different AI model/harness work nearly as well, so that then you don't need Claude anymore? If AI is doing it to humans in the workforce, why can't we do the same thing to AI... I love Claude but it's always good to have a backup, in case, say, the subscription plan goes away and you're expected to pay for the API usage rates.
If you'd like to follow along, discuss, or contribute, this is the repo where the work is going. I'm working on this using the DeepSeek V4 Flash repo but it would be interesting to see how the process works if trying to teach a smaller LLM.
Comments URL: [https://news.ycombinator.com/item?id=48326539](https://news.ycombinator.com/item?id=48326539)
Points: 1
source & further reading
github.com — original article