4 ms·
The people who hate AI seem to have plenty of time on their hands to respond to posts like this, so be ready. Expect, “If you’re 100 to 200 times faster and are
by paulcole 11mo ago
The people who hate AI seem to have plenty of time on their hands to respond to posts like this, so be ready. Expect, “If you’re 100 to 200 times faster and are managing a bunch of energetic ICs, what have you produced?”
You’ll never satisfy them with your answers because there is no world in which they are complimentary of AI tools.
IMO (as someone who strongly agrees with you and loves the AI revolution) keep this kind of thing to yourself. I just don’t see the upside of sharing given how overwhelmingly negative the prevailing sentiment is.
- keepamovin 11mo agoSo far seems good.
- al_borland 11mo agoI want to like it, but I can’t relate to people who seem to be able to direct the AI to do all this stuff for them. With what I do, the AI is barely functional. I’m always curious what people are working on when AI is able to effortlessly do it all for them.
- keepamovin 11mo agoThere could be some rose tinted glasses helping those folks, but I'm also curious what you do, and which models/tools you try?
- al_borland 11mo agoI’m only allowed to use Copilot at work. Within that, I’ve tried the various GPTs, Claude, and Gemini. So far, Gemini 2.5 Pro has worked best, but only on fairly limited scopes. I’m mostly using Ansible currently, which Copilot doesn’t seem great at. On top of that, there is a lot of OpenStack and other enterprise software, where there isn’t a lot of public code for it to train from. And then on top of that I’m having to add a significant amount of business logic and infrastructure specific stuff, which won’t exist elsewhere for the LLM to learn from. If someone has done it before, it is likely in another enterprise and not public code. Copilot has mostly been useful to write some json queries or regex, that’s about it. I know Ansible Lightspeed exists. I’m not sure how well that would do, but I don’t think I’m allowed to use that currently. We have a lot of governance around the use of AI.
- keepamovin 11mo agoThat makes sense right now
- jryan49 11mo agoThat's just silly. People are allowed to make grandiose claims and no one is allowed to question their statements? Anyone who is curious about these statements just "hates AI"?