@melvindvivas deepseek flash(and oss models in general) is amazing for security tasks, if you want to audit your codebase and try to pen test it. OpenAI/Anthropic models, just straight up block you from doing it to your own codebase. The usage limit is very generous for 10$
@Moleh1ll Psyop , he is 100% getting blackmailed by an intelligence agency to spread this message to the uninformed. Perfect candidate with the perfect backgrouns
@TaylorLorenz He spent years learning to code and takes pride in it. Now AI can do things he worked hard to learn, and his ego takes a hit. he’s trying to cope, that’s normal.
@cystalphantom@theo@danielmichaelni In my experience, the captain always monitors them and if he finds a problem he makes them use worktrees or finds a solution to coordinate them. This is what burns the most tokens, Its unpredictable, but at the end he always makes it work.
@cystalphantom@theo@danielmichaelni You let the main agent decide, it is smart enough to manage all the subagents by itself. You have one captain that coordinates all and you speak to him directly, he speaks to the others for you