@anah_sahh Wrote about this, after a guy told reddit they have to spend 10 prompts after each prompt to make things actually work. It all boils down to propositional logic, real experience, accumulated expertise.
https://t.co/FWhmXI0IyD
@ardent__dev 100% if you use AI to prototype and lead yourself. GLM comes 95%+ close to exactly what I need from the first prompt. Opus 5.5, Argon, maybe 30-40% at best. Usually require follow up prompts to fine tune things already explained in the forst prompt.
@gilles_ac We're a small team but a dozen people using it on local network. We pay 0, we don't share pur internal data, plans and problems with random third parties. It's gg. What's even better is that we can see what needs to be done asap, today, this week, this month etc on 50+ factors.
@sub0x_@ayoub_el77927 Our biggest mistake as self centered intel processing units, or humans, is that we assume no matter how grim or bright the future, we will undeniably be there. We can't even conceive, not even conceptually a future that doesn't carry us forward under the current format.
@gilles_ac The project management tool is now used internally in our organization as it eliminates the need for notion, jira, asana you name it. The latex rendering engine is private but boy is it on a next level?
@Ionkosm If it needs explanation you won't like the answer. Haven't used ChatGPT since 2023. Not even curious anymore. Sam Altman history, source of wealth also against my virtues.
@niqudrift That's a good question considering it eats up hours and stress for no reason. Guess our goal is to find people who understand AI the way we do, and want to buikd the future instead of inheriting it.