Richard Levitte
I write software. I play with software.
I take pictures, photographic pictures.
I dance; tango, Swedish and continental folk, medieval inspired.
Sometimes, I also participate in #queer, #gsrm, and primarily #polyamory / #relationshipanarchy spaces.
Sometimes, people refer to me by nickname. #Wolfie #GrayFox
What I express here are personal opinions, unless I clearly mark them otherwise.
@rkaj@mastodon.nu https://codeberg.org/ethical-foss/open-slopware uses a self-fulfilling definition:
slop | Anything generated by AI, typically code, books, articles, or images.
It's pretty dumb and nonsensical if you ask me.
Ok, rant mode on...
Why. The F*CK. Does. Everyone. *Assume*. Dark mode?
Yeah, I know, "everyone" loves dark mode.
I. Don't!
But apparently, I'm supposed to.
So here's what I get for just starting a new tool I wanted to try out. Completely f*ing unreadable.
@Infoseepage@mastodon.social @mrundkvist@archaeo.social
If the LLM didn't give you links to the stuff it researched for you, you're using the wrong tool. Use a better one.
If you didn't bother to have a look a the linked material, that's on you.
(General "you" used)
Also, if you're getting much slop, you may want to choose a more suitable model. Some are just not made for this sort of thing.
@ahltorp@mastodon.nu @rkaj@mastodon.nu Not sure what you're looking for in terms of scenario, but perhaps a summarised session could do?
- Human chooses a task to be done, and describes it to the agent, with criteria and guardrails.
- Agent produces something.
- Human reads what agent has produced.
- If human understands and likes, the product is shipped (committed, packaged, whatever), and the task is done
- Otherwise, human asks for explanations, or opines, and gives the agents refined instructions.
- goto 1
