Since this is shared by @valyala, I feel comfortable trusting the content, but is the original author of the blog delegating all writing to AI or are they genuinely this smart to produce high quality content this fast?
Nah, it’s full of LLM-isms. There’s a lot of focus on “there’s not just one profile, there are five”, “it’s not A, but B!” and other “X is to Y what A is to B” style content that obsesses over comparing different concepts.
I have nothing against AI blogs. Ideally it saves cost, so my AI does not need to waste compute to generate all that. We all likely have same models (Fable vs Sonnet vs Deepseek tiers of course) which produce same result for same input. So just save yourself tokens.
If it is human-reviewed AI blog, even better.
And if my AI can collaborate and improve/build-up/fact-check that article, perhaps our AIs working in GitHub Issue on those (which is a missing link now). That would be even better.
Since this is shared by @valyala, I feel comfortable trusting the content, but is the original author of the blog delegating all writing to AI or are they genuinely this smart to produce high quality content this fast?
For example see here their series about Filesystem internals: https://internals-for-interns.com/series/filesystems/
Reason I am asking, should I add them to my personal "Human blogs to read list"
Clanker image emdash emdash emdash emdash emdash emdash not to X but to Y
the art is def ai generated, text seems okay thusfar?
Nah, it’s full of LLM-isms. There’s a lot of focus on “there’s not just one profile, there are five”, “it’s not A, but B!” and other “X is to Y what A is to B” style content that obsesses over comparing different concepts.
I have nothing against AI blogs. Ideally it saves cost, so my AI does not need to waste compute to generate all that. We all likely have same models (Fable vs Sonnet vs Deepseek tiers of course) which produce same result for same input. So just save yourself tokens.
If it is human-reviewed AI blog, even better.
And if my AI can collaborate and improve/build-up/fact-check that article, perhaps our AIs working in GitHub Issue on those (which is a missing link now). That would be even better.
Or we could just use our fucking brains, how about that?