@corline0@goon_nguyen i hope they let people use their paid sub on any coding agent. seems like their model cannot do agentic well except in antigravity or gemini cli
@afifsohaili sure! btw you can get lower api cost for deepseek v4 pro, deepseek v4 flash, glm 5.3, glm 5.3-flash at @nahcrof. i think it's worth to check. and i use pi for high cache rate hit. happy coding!
@afifsohaili fable is off the chart ofcourse like any anthropic model. i think you can use multiple agent setup: ui use glm 5.3 and backend you can use deepseek or glm 5.3 flash.
@exponential_dan@ImLunaHey i dont think available distros are massively promoted like omarchy right?
and i see dhh and the team will adalt nixos as based distro. so it should be stable enough soon.
@alxfazio somehow when i use omp, it tends to drain my token a lot. seems like it does a lot of thing in background which make me migrate it to pi + pi subagents.
how do you find the performance and token efficiency so far on using omp?
@aibuilderclub_ i think using specialized agent on coder or using skill are better. Agents.md only contain repository context. then we can load the skills or told to dispatch the subagent to do the coding.