Our Mission
A continuous stream of consciousness, dynamic memory, intelligence, automatic context-relevant memory retrieval, structured thought, and fractal goals. Those are just some of the requirements for a productive adult human being, those creatures that studied math, built computers, and walked on the moon.
Today, LLMs provide only one thing, intelligence. There is no extreme-scale memory with dynamic creation and retrieval like what we humans enjoy, there is limited structural thought of the kind a trained human, like a mathematician or a programmer, might do, and critically there is no very-long-term planning and nested goal execution, where as humans can coordinate and work towards goals over years, even lifetimes.
The combination of all these things, and more, contributes to our experience of consciousness, that is to have a consistent self-identity that exists and operates in the world and remembers what was and imagines what's to come.
The point about nested goals is very important, so lets clarify by taking a simple example. Say you want to make a latte, so you need some milk, you go look in the kitchen and fridge but can't find any, so you decide to order some, but you need to remember where you put your phone, now you go grab it, search for your delivery app and open it, search for a close grocery that has milk, add milk to your cart and go to checkout, oh your card is expired, so you go search for your wallet, find it in the living room, grab your new credit card, grab the phone again, fill in the new card info and make it default, verify the OTP, confirm the purchase, and success! wait, what was all this for again?
This example hopefully illustrates the hidden complexity of even menial tasks, but the one thing we want to focus on is how one goal, making latte, has an unknown number of nested goals, like ordering milk because we ran out, and those nested goals themselves have other dependencies, like adding your new credit card, and so on. With this in perspective, the amount of thought and nesting behind even a small feature, like signup for a website, would surprise most people.
Since it is simply not possible to know upfront what actions would be required, serious AI applications need to maintain long term memory, be able to handle fractal goals, global context (e.g. we want a latte), specialized knowledge (e.g. the password of the phone), local context (e.g. got milk, now I can go back to get coffee), along with many other things.
Current AI systems, regardless of context size, are fundamentally incapable of handling such problems because they lack critical architectural pieces. Base model capability quickly becomes insufficient when facing the superlinearly growing complexity of fractal problems. No context window expansion or model tweaks can bridge such fundamental deficiencies.
As an example, the current situation of AI in programming, where its good for fresh projects or isolated small problems but immediately fails when you hit a few thousand lines of code, is a manifestation of the necessity but insufficiency of intelligence.
Overlord Systems has been founded to understand and solve the hard problems standing in the way of using AI in complex real-world scenarios, founded to complete the digital mind. Our belief is that building true AI agents is a hard engineering problem, but does not require new LLM models or capabilities nor is it even solvable through such means!
The economic value of crossing the threshold into persistent, learning, temporal agents capable of long-term complex work is nearly unbounded.
To share with you a small glimpse: we envision AI going from barely coding tiny apps to working as part of an engineering team on complex features requiring weeks of work, we envision the emergence of real AI-based software development agencies, we envision AI working in research labs and developing domain expertise, we envision people creating software without having heard the word 'code'.
In God's name and by His grace, we embark to build the foundation that enables AI to become a cornerstone of life, just as computers have become. Overlord Systems is an engineering-first company operating with a long-term vision, unafraid to tackle the hardest technical challenges. We have just started, but hope to share more about our work in the coming months.
If you are interested in what we do, would like to work with us, have thoughts to share, or are looking to invest, please feel free to drop us a line at contact[@]overlordsystems.com, we look forward to hearing from you!