Tbf people aren't reliable either. Not that I'm defending the AI models but remember these long-standing bugs were introduced by people and not found by people for years.
Anytime anyone feels insecure that AI will genuinely do their job, I point them to creators like this [1]. Also, I find it so wild there are people, even here on HN, who don't admit that LLMs hallucinate, lie and forget.
I would have a hard time trusting any critical code produced by an LLM. This was a frequent complaint with OpenClaw (being Vibe coded) and the general advice was to not run it on any computer you're logged onto as you'll get hit by prompt injection attacks and the like.
Anyway, the models don't need to be reliable. They just need to be an excuse to fire people to suppress labor costs. That's the true use case.
I only seen one instance where developer came to work still under influence of drugs after the weekend. He was moonlighting as a DJ and as he said the afters was legendary. He was fired that day. Usually you get an odd Monday with younger devs that they call off sick or family/health leave with the older ones. But they never suddenly run in circles for weeks seemingly not knowing what they do after weeks of exemplary performance.
Anytime anyone feels insecure that AI will genuinely do their job, I point them to creators like this [1]. Also, I find it so wild there are people, even here on HN, who don't admit that LLMs hallucinate, lie and forget.
I would have a hard time trusting any critical code produced by an LLM. This was a frequent complaint with OpenClaw (being Vibe coded) and the general advice was to not run it on any computer you're logged onto as you'll get hit by prompt injection attacks and the like.
Anyway, the models don't need to be reliable. They just need to be an excuse to fire people to suppress labor costs. That's the true use case.
[1]: https://www.tiktok.com/@huskistaken