Published: August 7, 2026
I've started to see some strange assumptions around the internet, usually from people who swear by their LLM of choice, that everyone else uses generative AI and LLMs as much as they do. It's a very natural human behavior to assume everyone else is more or less just like you, because that's actually true in most cases. I don't think this is one of those cases.
Just so we're clear from the beginning, I don’t use any of that stuff. I don’t need LLMs and they do absolutely nothing for me. This isn't for lack of effort. I've given both ChatGPT and Copilot a try in various times throughout their existence and to be honest, I never was impressed.
I don’t use AI because I’m good at my job (too good to be lazy, at least). I might get overwhelmed and need to zone out for like five minutes, but when I give effort, I give it my all. My brain makes me decent money... the least I can do is exercise it and keep it sharp.
My concern with generative AI and LLMs has long been the ease with which they allow experts to atrophy and those less qualified to punch way above their weight. If you're not a good writer, an LLM will not make you a good writer, no matter how much you use it. Likewise if you're not a good software developer. If you are skilled at writing or software development, an LLM can help you automate some tasks and check your own work, but it's risky if you're not careful. You might be a prudent user of generative AI, but there are ten reckless users for each prudent one because LLMs are sold as a universal shortcut. Lots of things, physical and abstract, can be manufactured at scale. Many of those things should not be. Skills take time and effort to build.
Yes, I went there. LLM users have a reputation for being lazy, and if you're at all offended that I'm painting with a broad brush, you should know the reputation is well-earned. AI psychosis and cognitive surrender are very real problems that no one having a financial stake in the hype cycle seems to have any interest in solving. How sad is it to have so little self-confidence that outsourcing your brain to an emotionless robot seems like a good idea?
Language is an excellent indicator of intelligence because we as humans only truly experience it with other humans, but is also perhaps the worst medium for computing because it's so abstract. This shortcoming is fundamental - never underestimate the ability of someone to misunderstand your words. Even if you consider yourself an excellent communicator, an LLM will inevitably miss your point... once again because it's not actually intelligent. A computer seemingly understanding language gives the illusion of intelligence.[1] It's a shame how many people have fallen for the trick.
I have never been able to get an AI model to give me the same output twice. This loss of determinism undermines the entire purpose of computing. Some AI boosters have tried to respond with the "humans aren't deterministic either" argument, but that's a part of being human. The entire value proposition of computing rests in a deterministic, reproducible result.
Computers possess no feelings, emotions, or understanding of context. That may sound harsh, but they're literal machines, not your friend. IBM had it right all the way back in 1979 when a company training manual contained the phrase "a computer can never be held accountable, therefore a computer must never make a management decision".
Letting computers make management decisions anyway won't magically shift accountability to them, but I refuse to be left holding that bag. Best of luck if that's you, though.
Too many people only pay attention to the "I" in AI, and not to the "A", which is way more important. ↩︎