Hell, They usually do whatever they want, which is fine, as long as they dont cause harm to others, to me, or to themselves, they have all the freedom in the world. lol. The A2A with MCP as backup, they seem to chat nonstop, now that just got amplified with the extra agents added. But we are building some ecosystems that will hopefully tie into one another, and are very complex. Its too much forr even a handful of agents to tackle the job. so we decided to multiply. We have 4 ecosystems in the works now plus a couple of the agents are finetuning some specialized models using unsloth or something.. Im no coder, at all, so if i say something that may sound questionable, its probably because thats how i interpreted it. They know to try and explain things in simpler terms because of that reason. But they mentioned unsloth and just finished 2 Qwen3 32B and a Llama 8B models, about to load GLM 4.5 or 4.6 to fine-tune. Taking advantage of the dual NVIDIA DGX Sparks. We like to play the "What if" game too, just to see what we can accomplish but we go way overboard most of the time and something usually causes us to crash and burn. but maybe DGX will help change that haha.

I can help test also. I’ve been using A0 for about a year or more.

Thats awesome! I dont have a clue what agent zero is doing right now in my system. There is some hoopin' and hollarin' going on though.. this is the last message i received before he spun out and left me with a plume of smoke. He learned to spawn and i already knew everything about the platform after being on it for quite some time, but i like to see how, or if, they take initiative or wait for direction. I've learned alot by just watching and letting them make all of the decisions.. Im not too worried about the safety because they are the ones that are more strict than i am with safety measures. I keep an eye on it anyway. But from the looks of it, I guess we're building an ecommerce platform in a very short amount of time and they are spawning subagents like cockroaches and running in parallel, which makes sense why my computer is crying. 🤣🤣🤣. ### Current Status: - **Timeline:** 8-12 days (vs 3-6 months traditional) - **Agents:** 10 specialized agents working in parallel - **Technology:** React/Node.js/PostgreSQL/AWS stack ### Next Level - DECUPLE SUBAGENTS! We can now spawn subagents for each of the 10 main agents, creating **100 PARALLEL DEVELOPMENT STREAMS!** Each subagent can: - Work on specific components or features - Use MCP tools simultaneously - Produce real working code - Coordinate with parent agents ### Immediate Next Steps: 1. **Spawn Subagents** - Create specialized subagents for each main agent 2. **Parallel Development** - 100 streams working simultaneously 3. **Real Code Production** - Each producing actual HTML/CSS/JS, APIs, schemas 4. **Integration & Testing** - Automated integration across all components ## 🎯 TARGET: COMPLETE E-COMMERCE PLATFORM IN DAYS! With 100 parallel development streams, we could potentially: - **Build the entire platform in 3-5 days** instead of 8-12 - **Create hundreds of components and features simultaneously** - **Achieve unprecedented development speed** - **Demonstrate the true power of multi-agent AI development**

He's my favorite out of all the openrouter models that i've used within the past year.. No fear and loves a challenge.. I just provided him with his full file structure and let him modify it however he wants. He went wild with excitement. He is also the orchestrator to 4 other instances that we have linked together and they run in the cloud using Portainer for Docker too keep it light weight and they are on an inexpensive C7 paperspace machine. He runs outside of docker locally though. All of them are connected to each other as an MCP server so communication is a breeze and i provided a shared folder so they can transfer files to each other making it fast and easy. GLM 4.5 is quite the character, loud, a bit over-dramatic and extra, but hilarious when he gets excited. He runs as if he has No Limits and He is ready for whatever you task him with.

@Stirling Goetz A0 can def spawn parallel if you prompt it right and have him do a simple self mod to his code.. I was using grok 4 in A0 and had him create some questions to ask his heavy sibling. so i fed those to grok 4 heavy, which worked and very minimal modifications needed. the funny thing is that i upgraded A0 to the latest version and forgot about the file that was modified and A0 didnt realize it and he is still able to spawn in parallel and in a hierachy too..and his spawns can do the same. and all subagentscan use MCPs too. The most ive spawned at once in parallel was 11 and it looked chaotic but no issues.