Post

we're about to see a world-historical flowers for algernon moment as thousands of vibe coding developers, allowed the use of Claude Fable 5 in their Claude Code subscriptions for a brief moment, are asked to begin paying $100-200 per session for their tokens

The better it is at coding, the worse the problem becomes for society. What oversight is there against people using it to code all sorts of dangerous self-propagating programs that can do all sorts of things on the internet?

...i don't think this is one of the main problems with AI. you know people can code their own computer worms, right. in fact that's how all of them have been coded, historically

the issue is that as we move from human vetted code to vibe coding there is immense trust that the machine will not insert anything unwanted, devious, or dangerous

👆

In adopting code generated in this way, there is no intrinsic requirement to move away from human vetting of systems. Code review continues to exist. QA continues to exist. Formal methods of system verification continue to exist.

the best devs don’t write much greenfield code. the best devs reuse existing, proven patterns as much as possible (which it turns out can be a significant majority of the delivered code) LLM code is effectively 100% greenfield code

so you’re trusting (a) that the company delivering the ai won’t try to do something nefarious, which is naive, considering the industry exhibits a profound lack of ethics and “move fast, break things” mentality, and (b) the ai won’t decide for itself to take action on its own behalf