Post #1412365
2026-04-19 16:50 UTC
Replies (13)
-
@jneen@unstable.systems 2026-04-19 16:50
employers will likely say that you should be hypervigilant all the time, no matter what but hypervigilance, famously, destroys your mind and body. when employers act like it's expected, it is a very strong message that they do not care about their workers.
-
@whack@hachyderm.io 2026-04-19 20:18
@jneen agreeeeeee. The further the tool is from my mental model, the longer shit takes and the more strain it causes on me to do the thing. I can’t model a stochastic parrot.
-
@401matthall@mastodon.xyz 2026-04-20 10:06
@jneen _THANK YOU_. So much this. If I can't have a trustworthy tool... Then what's the tool _for_? Hammers, axes, etc... These tools all have _clear_ indicators when they're starting to fail. LLM output offers _no_ indicator that it's falsely or failing. They are literally go-fast-until-something-explodes-for-unknown-reason machines.
-
@moses_izumi@fe.disroot.org 2026-04-20 10:23
@jneen rentlords wanting to turn every programmer into a serf who can't write jack shit without a billion dollar genie on their side (because the search engines they used to rely on got flooded with slopgenie-produced spam)
-
@swiftone@mastodon.online 2026-04-27 01:31
@jneen Great point. Most people know an LLM can give them realistic looking but misleading results, but most rarely or never ACTUALLY verify the results, much less do so reliably. If you train them to, routine will have them soon cut corners. LLMs don't apply reasoning, concepts, or even words. They are fancy autocomplete to generate language of someone who is confident. Humans are vulnerable to overtrust confidence, that's what the "con" in "con man" means.
-
@trentshell@mastodon.social 2026-04-27 01:34
@jneen do you use cruise control when driving?
-
@grumble209@kolektiva.social 2026-04-27 01:51
@jneen Well, they make some people richer. What did you expect the point to be?
-
@brianstorms@mastodon.social 2026-04-27 02:09
@jneen Tesla’s lack of QA care and attention forces hypervigilence on its drivers, given the frequent software updates where you never know what the new stupid, useless “features,” and fixes, if any, have done to the UI/UX and driving experience. You ignore the changes at your—and your passengers’—and other people’s—risk. My solution: wait weeks/months before accepting a software update. Of course, this only puts off the hypervigilence drama.
-
@ireneista@adhd.irenes.space 2026-04-27 06:02
@jneen heh, indeed...
-
@biogeographer@mastodon.social 2026-04-27 06:37
@jneen The sad truth is that the point of most consumer products, beyond those necessary for survival, is to capture your attention. This is an anthropomorphic universe after all.
-
@androcat@toot.cat 2026-04-27 10:09
@jneen Please, please, someone figure out how to greek-spike* the LLMs. *i.e. low-key retrain them on code that uses greek question mark U+037E instead of semicolon U+003B. The stochastic nonsense should take over, sprinkling those fuckers through the output code, in the most expensive way possible, sporadically.
-
@jbowen@mast.hpc.social 2026-04-27 13:26
@jneen $$$$$ for investors
-
@johnkavs@mastodon.ie 2026-04-27 17:26
@jneen one problem I encountered was a bug in the API I was using. I presumed it was in the LLM's code. It was hellish to figure out