Has anyone here been using Buzz yet? The idea of having humans and AI agents working together in shared channels (instead of jumping between Slack, GitHub, and separate chat windows) seems really interesting. I’m curious about real-world experience though: - Is it actually useful? - What workflows are you using? - Any surprises, limitations, or things you wish you knew before trying it? Would love to hear from anyone who’s spent some time with it.

I organize in folders like everyone else here, but I'm stuck on something. What I'm really trying to get to is being able to have one agent working on one task/department while another agent works on a completely different task/department at the same time — without either one pulling in the other's context, and without burning through tokens re-explaining or re-loading things that aren't relevant to what it's actually doing. I'm about to add a landing page, Terms/Privacy, and marketing on top of an existing app, and I want to know whether to give each its own folder or fold them into the app's existing one. My app is set up with its own folder, its own running notes, one thing actively being worked on at a time. I now need to add work that isn't the app itself: a landing page, Terms/Privacy, and marketing copy. Different content, different audience, different pace of change than the app — and importantly, nobody's touching the same files for these at the same time. That's exactly the situation where I'd want to point a separate agent at each one in parallel — one working on the landing page, another drafting the Terms of Service, another on the app itself (keeping their own feedback loops) — without any of them needing to load or wade through the others' notes and files just to get started. My instinct is that this argues for giving each one its own folder, all sitting side by side, rather than cramming all of it into the app's existing folder and juggling several things "in progress" there at once, where one agent's context ends up mixed in with another's. Two things I'd like input on: 1. Shared facts. A few things are true everywhere — the product's name, a one-line description, what it does, who it's for, pricing, what data it collects. My plan is to write those down once, in a shared folder that each department can read, and have the isolated department folders point back to that shared context if there's something worth sharing across functions. When one folder finishes something the others need (like the legal folder finishing the Terms of Service, which the landing page then needs to link to), it gets handed over once as a finished document, not kept magically in sync. Does that "write once, hand off when done" approach hold up in practice, or does it fall apart over time?

If your system already does context management, retrieval, memory, and orchestration well — does upgrading from Model A to Model B actually move the needle? Or are we hitting a point where the system around the model matters more than the model itself? Personally, I’ve started caring less about which model is “best.” I often run two models at once when I get stuck on something — not because one is smarter, but because I end up acting as a mediator between their ideas, pulling the best parts from each to solve the problem. The model feels less like the bottleneck and more like one input into a process I’m steering. Curious where you land Are you still upgrading models the day they drop, or has that urgency faded? Has better context/retrieval ever made a “worse” model outperform a “better” one in your stack? Anyone else running multiple models side by side and playing mediator like this?

Do someone have tried claude Tag. I would like to know the difference between Claude Tag and the ICM methodology. Maybe antrhopic with Claude Tag has find a scalable solution for a clear context, and memory that evolve with time ? By the way, how did you do to make an memory evolve with new information? For exemple : T0 : your knowledge.md contain a specific information about something so this information is true 👍 T1 : after some research your knowledge.md evolve but the specific information that you find in not correct T2 : how does your system know which information is true and accurate ? Do you understand what i mean ?

Hey everyone, Before finding Clief's content, I had gradually stumbled into my own version of this way of working. I/ my agent, was capturing ideas, building systems, and connecting notes across different projects, but I didn't really have a name or framework for what I was doing. One reason the content here resonated with me is that I come from a completely non-technical background. I have no formal computer science or coding education. As I've worked on more technical projects, a lot of the discussions/videos in this community have helped fill in foundational gaps and given me better mental models for understanding the larger systems and concepts behind the technology. Finding Clief's content felt less like discovering something entirely new and more like finding language and structure for something I had already been doing. I'm curious—what was your path?