Still working on this one. Honestly it's the piece I'm least finished with — and the one I'd most like to get some real security into, because the moment you connect things, the communication layer becomes the whole ballgame.
The concept
Quite simple, really: connect multiple Bobiverses. My Claude-Bobiverse talks to my Codex-Bobiverse. One team-of-teams reaching another team-of-teams.
And here's what's so amazing right now: almost everything is predictive — text in, text out. Which means it's mostly model-agnostic out of the box. You install a Claude Code plugin inside Codex and, boom, the Bobs are backed by ChatGPT now. Next week it's Cursor, and I'll happily wire those in too.
The part I'm really after is picking the model per Bob. Like Matthew Berman put it somewhere — I'd love to staff a team like this:
Bob (Fable) [Team-Lead]
Bill (Cursor) [backend-dev]
Riker (ChatGPT) [code-review]
...
Right model for the right role, right model for the right task. And every other day there's a new toy to slot in. Christmas never ends these days.
Why I'm not drawing you a fancy diagram
I could make some shiny graphics. But for what? This one is for the people who already get it. If that sounds root — it's actually the opposite. I'm just not good at writing, and especially not good at describing an abstract idea like this for "usual humans".
nerd: I genuinely need a translator for this one. 😩
