But Codex doesn't survive a reboot by default or a laptop going to sleep. Also, herdr is abstracted up a level from the agent, so you actually get more benefit by using Codex with herdr because herdr knows how to operate Codex, and other harnesses. So if you're using multiple Codex instances you can orchestrate them because each harness can talk to the others. You can still interact with Codex running in herdr via remote control (ideally you'd target your "orchestration" Codex instance). It just gives you way more power.
Hey thanks for trying it out! The 2 participants you see are just you and the agent.
AI Prompt means that the AI will always reply, Team message just means it's meant for a human. We can see how this is confusing though and we want to simplify this by just making it smart enough to know when you're replying and when you're not.
As for the burning of the tokens it's hard for me to know without troubleshooting further. Can you email me your org name k at type dot com ?
Underneath the hood it's the codex cli and claude code cli as the harness. So product wise it's similar to ChatGPT Work and Copilot Work, but technically our harness uses the CLIs.
For translations, the score is basically 1 or 0. For some tasks, the least amount of LOC gives the highest score, and so on. Basically, you need to figure out how to score it, so you can compare scores across agents/models.
I have this cheap B movie in my head with a primitive people living on an island. They compete in hunting, fishing, building boats, houses, cutting trees, growing crops etc they use sea shells as currency. Someone finds a spot with countless sea shells, 95% of the population spends their days digging up more and more. Almost everyone is insanely rich, everyone except from the dumb people still hunting, fishing, building boats, houses, cutting trees, growing crops etc
reply