LLMs are mostly a party truck that doesn’t work half the time. Like … Neat that it can kinda sorta do some stuff, but not yet at the level of consumer ready product with actual uses. The fact that it’s actively detrimental to the people it’s being forced on by the clueless ceos and management who are just protecting their other investments makes it just like all the idiotic return to office pushes because they want their real estate portfolios to go back up
the issue is at one point, like say a year or so ago, LLMs weren’t that bad. they were fairly accurate in their solutions. But lately within the past few months they’ve all collectively gotten noticeably worse. They ate up all the content available and then proceeded to start eating each others waste and vomiting out that as a solution. Claude for example at one point was a decent coding assistant. now? now 8 out of 10 solutions are hallucinations. GPT5 is a clear downgrade from previous versions and now the thing just rants and rants in hopes that somewhere in it’s rants and info dumps there’s potentially the correct solution. It also now fails to remember the context of a prompt most of the time. If you can’t get the correct solution within one answer from GPT5 you might as well just close the tab because it’s never going to get there.
Yep. My company has its own AI instance which we are encouraged to use in our work. I uploaded a document today, which I wrote myself, and asked it to summarise the high level requirements so I could copy/paste the bullet points. It got maybe 1/4 of the requirements, and what it output didn’t make sense without extensive knowledge of the topic. Went back to writing it myself, like I normally do.
LLMs are mostly a party truck that doesn’t work half the time. Like … Neat that it can kinda sorta do some stuff, but not yet at the level of consumer ready product with actual uses. The fact that it’s actively detrimental to the people it’s being forced on by the clueless ceos and management who are just protecting their other investments makes it just like all the idiotic return to office pushes because they want their real estate portfolios to go back up
the issue is at one point, like say a year or so ago, LLMs weren’t that bad. they were fairly accurate in their solutions. But lately within the past few months they’ve all collectively gotten noticeably worse. They ate up all the content available and then proceeded to start eating each others waste and vomiting out that as a solution. Claude for example at one point was a decent coding assistant. now? now 8 out of 10 solutions are hallucinations. GPT5 is a clear downgrade from previous versions and now the thing just rants and rants in hopes that somewhere in it’s rants and info dumps there’s potentially the correct solution. It also now fails to remember the context of a prompt most of the time. If you can’t get the correct solution within one answer from GPT5 you might as well just close the tab because it’s never going to get there.
Yep. My company has its own AI instance which we are encouraged to use in our work. I uploaded a document today, which I wrote myself, and asked it to summarise the high level requirements so I could copy/paste the bullet points. It got maybe 1/4 of the requirements, and what it output didn’t make sense without extensive knowledge of the topic. Went back to writing it myself, like I normally do.