> AI is the silver bullet - my output is genuinely 10X what it was before claude code existed.
This reminds me of ThePrimeagen noting that Dario was right about 90% of all software being written by AI. 90%+ of all useful software is still being written by humans, but now we've also got non-programmers like Gary Tan pushing the button to generate garbage piles of AI generated code at 10x the volume of what humans are generating.
Or look at Claude Code itself, the poster child for what AI generated code looks like, especially when you're operating in YOLO "never mind the quality, count the LOC" mode. Over a half million lines of bug ridden crap when 10-50K could have done the job.
It's always worth remembering that code is a liability while software is the asset. Wiring 10x more code is accumulating 10x the liability, with the often unjustified belief that the software is 10x as valuable. It might be, but that's what the market is for.
I highly doubt that, unless you are talking about something much more specific like test cases.
There is certainly no sign of these companies becoming more productive.
Most of the software that "billions" of humans are relying from any of these companies are things like Microsoft Windows, Microsoft Office, Google Search, FaceBook, that were not written, or rewritten, in last 1-2 years, so were certainly not written by AI.
Well, not exactly productivity, but yes there are some symptoms of a reduction in quality from some companies. I was about to say not the three mentioned, but GitHub (owned by Microsoft) is certainly one, and Amazon seems to be having an uptick in outages that are anecdotally linked to use of AI. Anthropic, who we know are using AI a lot internally also have reliability problems.
This reminds me of ThePrimeagen noting that Dario was right about 90% of all software being written by AI. 90%+ of all useful software is still being written by humans, but now we've also got non-programmers like Gary Tan pushing the button to generate garbage piles of AI generated code at 10x the volume of what humans are generating.
Or look at Claude Code itself, the poster child for what AI generated code looks like, especially when you're operating in YOLO "never mind the quality, count the LOC" mode. Over a half million lines of bug ridden crap when 10-50K could have done the job.
https://techtrenches.dev/p/the-snake-that-ate-itself-what-cl...