On gh-aw: it looks solid for the event-driven automation shape (triage, docs sync, CI fix). We're after a slightly different shape: interactive back-and-forth,

On gh-aw: it looks solid for the event-driven automation shape (triage, docs sync, CI fix). We're after a slightly different shape: interactive back-and-forth, steering from Slack or Linear, persistent sandboxes with a booted dev server for live previews. Thanks for the pointer, I'll dig into it more. On labs eating our lunch: it's definitely a risk. Our bet is that reusing lab-native CLIs is enough to position ourselves in the market On behind the firewall: it's something we're looking into. We open-sourced agentbox-sdk in that direction.