have a subscription to Claude Code, as well as subscriptions to DeepSeek and Gemini through Vertex AI. I searched for a way to combine them and use them within a single environment. I came across a terminal tool called wmux. It allows me to run an agent in one terminal while enabling that agent to monitor, control, or collaborate with another agent running in a different terminal pane or split terminal. I defined rules for the terminals, the agents, and how they should operate. I also made Opos 4.8 the lead project agent. After trying this approach, I found that it works. However, I am not sure whether it is actually beneficial and saves tokens, or if it simply consumes more tokens overall. One reason I rely on wmux is its agent-awareness capabilities. Because of that, I thought it might be a useful setup. What do you think? Can this approach improve and speed up the work, or is it just adding unnecessary complexity? https://github.com/asadeisa/SpeakType

I watched the video. The reflection is quite impressive. I'm actually not expecting that level of accuracy from this model, but honestly, Mythos is consuming credits way too match, and we have to start thinking about the best strategy for using it.

Thank you for this content—it was very helpful and gave me some interesting ideas. One thought I had is: if we have multiple sub-agents working together, why not assign a dedicated supervisor agent whose main responsibility is to monitor the other agents? It could identify bottlenecks, conflicts, duplicated work, or situations where an agent gets distracted from its objective. The supervisor agent could then generate reports and insights that help us improve the workflow, reorganize the structure, and optimize how files and tasks are handled. I'm curious to hear your thoughts on this approach.

"Hi everyone, I'm Asad Eisa , a frontend developer working with Next.js and Vue.js. I'm currently interested in AI and its tools — specifically how to manage AI workflows and come up with clean, polished solutions. I've recently been using tools like Codex, Claude, and open-source alternatives, and I'm also interested in MCP services. Right now I'm exploring how we can integrate MCP into projects that are currently under development, and I'm curious whether it can speed up the workflow and make testing easier.