I’m building an AI/AGI architecture that is not based on LLMs or transformers.
I came to the conclusion that intelligence needs a different computational foundation: nonlinear interaction, persistent internal state, feedback and multidimensional connectivity — closer in principle to biological networks than a simple module → module → module pipeline.
So I built the computing core first.
R-Core Gen7 already exists, but I’m giving you Gen1.
Source, RTL, tests and documentation:
https://github.com/etoyruben-wq/R-CORE-A patent application has been filed for the architecture.
Run it. Break it. Change the topology. Scale it. Show me where it fails.
I don’t need “interesting.”
Show me tests.Those who really understand Gen1 and produce results will also get early access to another technology I developed for memory reduction and data transmission.
Let’s see what this community can do with it.
Thanks to the admin for not blocking me… I’ve been banned on some other sites. I’m from Ukraine, and I work sometimes without electricity or internet, with Shaheds and missiles flying and exploding overhead. And when you try to post something and get blocked, it’s really unpleasant…
As far as an AGI architecture goes, I can’t really comment because I am not highly interested in their scope or viability. I would say that I’m not sure you’re on the right track when I see the word “deterministic” being used, but that’s probably more of a philosophical discussion than the testing rigor you’re looking for so I’ll refrain. I don’t have the PhDs required for that anyway.
I wanted to offer a different feedback: I get the sense that you’re using an LLM to generate all the copy in this initial post and in the github. It’s generic feeling and off-putting to me. When I need to generate copy, rather than getting LLM generated first drafts I find the quality significantly improves when I do the first draft myself (a lazy one, at that, because I’m a lazy guy) and then have the LLM fill it out and clean it up. Then it properly retains my style and all of my intent, which goes toward a better presentation.
Even if you do that and then get it to translate, at least it means the first draft is entirely based on your word choice. And I think that comes through strongly to others in a day of dead internet. Thanks for joining and sharing, btw.
Thanks, that is fair feedback.
The main reason is actually very simple: I do not understand English well. When I use ordinary translators, the result is often pretty bad, especially with technical text.
So I usually write what I want to say in my own language and use ChatGPT to translate it. It also tends to clean up the structure and pull the wording a little more toward a scientific/technical style. Sometimes, apparently, a little too much. 
I understand your point about this making the text feel generic, and I will keep that in mind. Maybe I should ask it for a more literal translation and leave more of my own wording in.
But that part is only presentation. The RTL, tests, architecture and characteristics of the core do not change depending on who translated the README.
So I am much more interested in criticism of the actual core. If you ever notice something technically wrong, strange, inefficient, or worth testing, I would genuinely like to hear it.
And thanks for the welcome.
I’m not the guy you’re looking for on that, I’m afraid.
Not sure where and how you’re using translation, but the other two posts you’ve made are much more “you” to me. For whatever that’s worth to ya.
I’m not really looking for one specific person.
In fact, over the last six months I haven’t found anyone who immediately looks at this architecture the same way I do. Most people are used to a computing core looking more conventional — CPU, GPU, separate modules, clearly separated functions.
What I am trying to build combines ideas that are normally separated, so it naturally looks strange at first.
The work is still continuing, and I am still learning too.
I am not a professional programmer. I started learning because I needed to build the architecture I had in mind, and a lot of what I publish is basically me showing that learning process in public.
So I’m not expecting someone to “join the project” or prove anything to me.
If someone looks at the core, tests something, points out a weakness, or simply gives me an honest technical opinion, that is already useful.
And if the other posts sound more like me, that is actually useful feedback too. I probably need to keep more of my own voice when I translate.