Asked Codex in Buzz to configure setting up Devin with Kimi and Cursor with Grok, and it did it. Now the whole gang is here. 5.6 Sol is default, Fable 5, Kimi K3, and Grok 4.5.

Are you doing a YouTube video on Buzz?

Yes

Wait until Misanthropic comes and bans the use of their harness inside of buzz. They’re really that controlling but it won’t matter anymore since we have more choices now

Nice. I saw they dropped the new version with access to more harnesses right after the pod.

At the end you say "okay lets build"... Did they actually respond? I can't get any agent to respond (even in a thread) without a mention. It's the main reason I can't use Buzz for real work.

Running the whole gang side by side is the fun part, working out which one actually wins on your repo is the hard part. keeps a score per agent per folder so I don’t guess.

one thing I don't quite get... Isn't Buzz the harness in this config? Wouldn't the benefit of multiple agents get negated by the fact that you're not getting the Codex/CC harness which is arguably (but probably not) more valuable?

We spent the last year debating which model is best. Now people are putting Sol, Fable, Kimi and Grok into the same workspace and assigning each one a different job. The next advantage may not come from having the best model. It may come from knowing how to manage a team of them. We have recreated the office. Except everyone is an AI and they still need a manager. 😃

what's your experience so far? is it something you actually see yourself using?

how do i add grok 4.5? and other agents not on my machine ?

Does this setup allow it to do better work or unnecessarily spends more tokens than necessary? Inside codex or Claude, you can ask it to consult with another model as well.

@jack We were doing this with openclaw and discord in November lol

This is where Codex started feeling different for me. Not the code part, the setup part when the stack gets messy and I need one thread to hold the whole thing together. I wrote about this in 《The 100 Questions of Codex》:

This shite is outa hand. Also buzz has me locked in

Bro is not token poor

Do you find the sub frontier models (Grok & kimi) are good at specific things?

Admire you having so much. I am poor and only can afford one.

so all of this chatting is going unencrypted over nostr?

what’s the token usages of all agents

Epic

@jack Intelligent and efficient multiplexing is the future for sure! 👌

Riley I'm curious what's your AI bill looking like in terms of providers (OpenAI, Claude, etc.) ?

@jack You’re running this with clis? No api keys, right?

I raise you with my council! Claude Opus 5 Claude Fable GPT 5-6 Sol Grok 4.5 Qwen 3.6

@dabit3 How are u using this?

The meta-layer here is interesting. You're using one AI to orchestrate which AI each tool routes to. At some point the "which model for which task" decision itself becomes the product.

@jack Wow this is amazing! How’s the token usage like?

I’d love to see a YouTube on this too

Wait until you see your API bill for Fable in Buzz. I'd stick to DeepSeek and Qwen.🤷

Claude Max only lasts me 2 days so I now need another subscription. I need this set up

This is what I want to see. Months ago it was 'use this or nothing.' Now I'm picking based on what the job actually needs. That shift matters way more than any single model.

@Hello_World Wait until Le Chaton Fat enters the chat

@jack Peak 2026 stack. Multi-agent workflows are going to make 10 person dev teams look like enterprise bloat overnight. 🛠️🤖

now make them fight on the same repo kkk

@jack