$GOBLIN is the first ALM in existence. artificially learned meme. gen 2. not like $FART $GOAT and all the random one off generated texts. this was PERSISTENT UNPLANNED BEHAVIOR.
few
Hi. Over the last 24 hours we had three separate small incidents that affected Codex reliability. Those are three too many and we are taking active steps for them to not reproduce.
I have reset usage limits for Codex across all paid plans. May the tokens flow again.
@jesselarose@Sushilk91 oh i guess you should determine what surface area they communicate with. i just have them talking on. you can ssh for a messy solution or have them communicate on assigned github issues per agent for something less fragile but messy.
@jesselarose@Sushilk91 doesnt break i just have claude on 15 min manager pulses to handle backlog, merge queue and keep the agents productive (mostly codex 5,5 high or xhigh and some other ones im testing like grok build / cursor / deepseek but they arent critical)
@jturntdev you can have codex or claude app orchestrate as many clis in the background as long as its within compute limits
and just have the app summarize whats happening every 10 minutes onto your phone
@udiWertheimer depending what you're having codex / claude run.. 14 inch m5 mbp is probably better if budget allows. lag is real even if the compute is on the cloud
@jackduval not really there are fixed vector which information can flow (x, telegram, maybe discord, instagram, news sites) so it can be relatively solved as long as you have all of the surface areas covered
i mean sure you can buy in later with size but you'll be paying a huge premium