A company built by engineers who saw the limits of AI and are now obsessed with building the future
Our Theses
- The current AI paradigm is a dead end.
- The focus on model scaling is a race toward a local maximum—building faster ways to arrive at incoherence.
- Today's models are performers, not learners.
- A system that cannot write new knowledge to its own memory is not learning; it is performing.
- The vast majority of compute is wasted on disposable outputs.
- RAG is an architectural crutch.
- Vector proximity is not a substitute for logical integrity.
- Intelligence is an architectural property.
- The industry's focus is inverted—LLMs are a commodity—the durable, compounding asset is knowledge.
- The next breakthrough isn't a bigger model. It's a self-forming, organic architecture that learns like a human.
- We are building it.
Joe Cohen
The Architect
Visionary and technical lead with a background in mathematics. A systems thinker obsessed with cognitive architectures compounding intelligence.
Pouya Javadi
The Builder
Elite engineer, formerly at Meta. A master of building and scaling resilient, mission-critical systems from zero to petabyte scale.
A New Beginning
We stand at a critical juncture. The prevailing AI architectures, for all their descriptive power, are a form of intellectual closure. They are vast, static encyclopedias of a world that has already moved on. Once trained, their capacity for new explanations is extinguished.
They can repeat explanations. They cannot create them.
This is not an infinite ascent. It is the perfection of a cul-de-sac.
True progress is not building a larger encyclopedia.
It is discovering a better process for correcting errors.
A process the merges knowledge, hypothesis, conjecture, and evidence into a coherent, verifiable, ever-evolving explanation of the world.
Our work is the construction of an architecture for this process.
We are not building another model to be trained.
We are building a system that can learn, reason, and correct itself.
Our architecture treats reasoning not as an ephemeral computation, but as a durable, storable asset—generating a web of causal, hard-to-vary explanations—forming composable context that scales awareness superlinearly.
We are building a living system designed for the perpetual refinement of its own understanding. A fallible, problem-solving entity capable of participating in the infinite journey of discovery.
This is the search for a new beginning.