02-18-2026, 10:36 AM
(02-17-2026, 10:09 PM)HeggebøsBarnDoor Wrote:(02-17-2026, 08:38 PM)CarlosCorbewrong Wrote: I’m personally finding that the competency of LLMs to have stagnated or reversed of late. Agentic AI, for non programmers, seems some way off
Yeah for coding it is an absolute beast. Can bash out stuff all day using cursor and claud. For generic stuff I feel it can often have a confirmation bias.
A guy I work with tried using ChatGpt for coding, i had to tell him to stop as the code and idea it came up with was pretty much garbage. Was fine for quick little methods where the answers can be found just as quick by googling though... strange that.
I see someone spent £20k on getting AI to build a half decent linux Compiler... then everyone pointed out it had largely just stolen code examples from online and combined them rather than actually using any form of intelligence to code around issues.

