I all of a sudden feel behind

A question from a bin techie.. is this local set up or cloud?

Its cloud, but encrypted, in theory no one can see your chat, in practice hardware manufacturer and owner have access to their own keys and "can potentially decrypt", but its harder and the best option we have right now

I asked because my laptop is old and I seem to be constricted in local ai set up and agentic ai I'd love to play with... But sounds like this is workable and clawdbot could work without require upgrade if my machine, am I getting that right?

Cloud.

Same here, I gotta catch up fast.

Using through signal?

If you're not running any LLMs locally, it should be able to run fine on most computers.

I collected these notes from some exploration earlier. Figured if you’re this far along you’ll enjoy this detail gathered from leading users focused on scale. The focus here is on “Scaling Productivity With Agents” If you’re trying to scale capability, forget about building a hyper-smart agent. That doesn’t scale. The architecture that scales keeps the workers dumb and the system smart. Looking ahead, the winning teams won’t be the ones building elaborate agents for long drawn out execution. They’ll be the ones building ruthless orchestration in systems that can spawn, direct, and merge the outputs of hundreds (or thousands) of tightly-scoped agents autonomously executing. Think narrow tools, smart prompts, clean handoffs. Not one badass agent trying to save the day, but 100 decent ones getting shit done, fast. Most teams will fail at this because coordination is hard. Compute is scaling faster than your engineering team. But orchestration? Orchestration is a design problem, not a compute problem. You need tiers. You need external coordination. You need to build systems, not agents. The complexity belongs in the detailed and well structured system layer, not the agent brain. The agent should be stupid and disposable, like a script. The prompt is its operating manual. That’s it. If you get this right, you get output that scales with input. Then add agents to get more done. Get it wrong, and you get a coordination collapse after spinning up multiple agents that fail under their own context weight. “The magic isn’t in the agents. It’s in how they’re managed.” — Nate Jones This is the transition we’re living through right now. Multi-agent systems are about intelligence at leadership, and throughput at the dumber tier. The ones who figure that out will significantly outproduce the ones who don’t.

I'd do the same except I think it would burn through API credits.

Ya Maple is chincy with their allotment even on the max sub.

Need details on your setup Marty!

Ya it’s incredible. I named mine Stanley

Need the details on how you set this up. Hesitant after hearing the potential security risks with clawdbot.

Brian Roemmele on his Zero Human Company.

The Zero Human Company Run By Just AI. – @ReadMultiplex

Lol.. also bin = non*

Do they include any API credits in max?

Please elaborate on rhr.

Let’s go!!! Clawdbot with @Maple AI privacy 💪 View quoted note →

I understood that if you are using quantized models you are increasing the risk of prompt injection.

Not yet, it's something we're looking to do. As for regular usage credits on Pro and Max, we evaulate costs regularly and have expanded credit amounts multiple times.

How does it compare to Claude being the LLM?

Do you mind sharing the instructions?

How is this set up treating you so far, Marty? What do you love most about? Any hick ups and frustrations that are gonna need ironing out?