invisibleink

The point is, every new technology attracts its share of romantic skeptics, and every time they fail then they retreat to the same tired line:

"Just because all those other inventions didn't wreck humanity doesn't mean this one won't"

But that’s not an argument, it’s an evasion.

Given past inventions didn’t destroy us despite similar concerns, then the burden is on you to show why this one is fundamentally different and uniquely catastrophic.

YES and, also let's not use printing press, photography, word processors, spell checkers, internet and search engines because they lack human touch, make us lazy, prevent deep thinking blah blah...

interesting. is the answer not context specific most of the time? even if I ask LLM the same question again and again the answer depends on the context.

what are some use cases where you need deterministic caching?

The question is a good one. The responses? Deeply unoriginal, and technocratic.

What if the notion of work and the whole society built around working changes? What if?

All life is politics, and workspaces are not politics exempt. The world we live in understandably makes many cynics. Yes, still we want no kings, and more politics in and out of our workspaces.

we redesigned the whole system from ground up for robustness, speed, scale and multi agent capabilities.

thanks. would love your feedback once we launch.

> I’ve been using tools like Cline and Roo-Code quite a bit, and the idea of rethinking the core stack/UX from scratch resonates a lot.

Would love to learn more about your experience with cline. We spent quite some trying to add multi agent capabilities and improve the overall ux/devex of Cline (and its clone Roo) to make it more intutive for developers. we found that its stack is not built for these capabilities, and the codebase was not as stable as we would like it to be. it required a major rewrite, at that point being another cline clone made no sense to us.

> The agent workflow and devex feel like the hardest problems to get right

yes, AI coding agents are probably the most complex agents out there. our approach comes from actor model, where each agent manages its own event-loop. we found this is incredible robust way to build specialized agents that interact with one another within multi. Too early to say but

> Excited to see it open-sourced and will definitely give it a spin.

thanks. still brewing but will let you know. appreciate your feedback.

Building https://multi.dev, an AI coding agent with bunch of FOSS contributors

We took a great amount of learning from tools like Cline, Roo.. After spending some time on their tech as active users/devs, we decided to build multi from scratch with drastically different take on core features, tech stack, ux/devex..

If you are an active user of similar tools, and/or want to try multi.. We want to hear from you.

-- edit: I am one of the core contributors to multi. And we are in the process of open sourcing it.