Oracle7 is on Hugging Face. No language model. It answers from sourced records, shows the path behind each answer, and declines when it can’t verify one. Runs on a CPU. Sealed test, run once on the release: 117 right, 3 wrong, 30 declined. huggingface.co/psiloceyeben/…

The CPU-native AI that refuses to guess. Very interesting stuff 👇

Join the space got lots to say!

Join the space got things to say

Also this isn’t to show a lack of credibility here I’ve spent the past 3 years basically getting a college level education in ai research and machine learning I had a choice 3 years ago, I was broke, out of a job and not sure what to do, I got a part time job and started working with ai around when opus 3 came out I have a history degree and really contemplated going back to get a degree at cal tech for a cognitive science degree to work on building my models of consciousness I had conceived through mushrooms use and it became clear that would just be a waste of time when I could just start building and learning instead I could teach a class on this stuff now It’s my obsession honestly And I love it wouldn’t rather do anything else atp Build build build

I hope in some way this project serves to show you all what is really possible now I have no background in ai research I just have ideas and passion for following through with them and, with lots of persistence, I was able to build this ai system my own way, on my own terms If you have an idea, don’t be afraid to start asking questions and pursuing answers We have tools of incredible capability at our fingertips and can basically bring to life anything we can imagine at this point Don’t ever be afraid to be curious, it’s how revolutionary ideas begin, with simple questions and unwavering persistence

Ngl whole market is tanking and we are still holding pretty strong Thankyou all for staying supportive and engaged even as price fluctuates NFA but my goal is that this reaches millions and gets proper eyes and usage for the model Won’t stop till the top 😤

Swarm update is coming tonight This one will be very good The swarm builds specialists as needed to solve problems and implements a shared memory between all specialists that you are running

For some context Vector recall and muddying at high dimensions was an open vsa problem for a few decades before I decided to fix it one day for my own use

Because of some of the magic stuff I did which increases vector storage to multi-trillions of recall without muddying the information so that its lost, all the specialists will be able to have a memory of their history as well as a shared memory that they call pull from across specialists and instances This is maybe one of the most consequential parts of the architecture and it kinda gets glossed over Normal vector symbolic storage is limited at a certain point, and especially on cpu only, because the vector space starts to muddy and the recall becomes worthless, I found that if you create a stabilizing spine that routes between related parts without unpacking and holds the unpacking till the end, where each node and path that is along the route uses what has been added prior and adds to it from that base, the whole thing can be unpacked at the end line a coil being unwound, which increases the theoretical vector storage space to somewhere near 15trillion cells of storage, without losing the information 🫡

Better explained by Astra “One of the biggest parts of Oracle7 gets glossed over: the memory. Each specialist will be able to remember its own history, while also pulling from a shared memory across other specialists and instances. That means useful information from one specialist can become available to others, and previous work can become something the system keeps building on. The hard part is making that memory bigger without losing the details. When you pack too much information into the same vector representation, the pieces can start interfering with each other. The memory gets muddy, and pulling out the right information becomes harder. I built what I call a stabilizing “spine” to address this. Think of it as a backbone that keeps related information connected. As the system follows a path through that information, each step carries forward what came before and adds to it. It keeps everything packed along the way, then unpacks it at the end—like unwinding a coil in order. The goal is to let the memory grow while keeping the information clear and recoverable, within a CPU-only architecture. The structure has a theoretical capacity of around 15 trillion storage cells. But the number is only part of why this matters. It means building toward specialists that remember what they’ve done, draw on what other specialists have recorded, and carry that history into whatever comes next. Their own memory, plus a shared memory they can draw from. I think that could be one of the most consequential parts of the entire architecture.”

Gm guys! Lots to do still the work never stops till I’ve got a factory out here building universal sensors to connect us up with the world

Also a good one to research: solana:DU7Lhe2QtcRBJdor7KbpjfUDmLyA2kdhR49Li3czpump @Oracle7vn

DU7Lhe2QtcRBJdor7KbpjfUDmLyA2kdhR49Li3czpump

The Swarm: Coming Soon