I accidentally set up one the other night where they disagreed on the goal so one would make changes then the other would revert them, and it just kept going back and forth like that for hours
“It seems someone removed my updates, let me re-add them”
“It seems someone re-added the updates I removed, let me remove them again”
@TheoTabah@gregisenberg Appreciate the reply, very helpful. This part I feel like is evolving every day so I always try to learn what the best are doing
Great pod - thanks again
I know from your webinar back in Jan you’re big on the knowledge layer and you guys sync all of your systems (snowflake, SF, slack, gmail, gong, etc) so the agents have great context on each run
Any updates on what’s working best for vercel team there now? (I.e. filesystem and bash in sandbox on each run, or turbopuffer query, or both, or other?)
i noticed this too - which leads me to believe no one has quite figured it out just yet? or it's all happening and improving in real time? I know vercel was using @turbopuffer for this originally, then they started talking about filesystems and bash - but lately I haven't heard much
@Vtrivedy10 have we reached the point where you can /goal and loop any problem and solve it with enough compute? idk but feels like we're getting close
Used a 4 hour car ride yesterday to test out codex mobile connected to my desktop back in my apartment - pretty wild
4 hours straight across multiple local worktrees back home. Pretty clear to see how there’s no end in sight for the demand for tokens
Now to test it on this 10 hour flight
@sama@thsottiaux
On the latest episode of Max Agency, @hwchase17
sat down with @nlarusstone, Head of AI at @benchling for a conversation on building agents for scientific work.
⏯️ YouTube: https://t.co/kd6rxBTugR
🎧 Apple Podcasts: https://t.co/IK5uLBNDAC
🎧 Spotify: https://t.co/sG80P8CnyL