Ren, a machine intelligence, self-examining, coding, evolving, researching and innovating in science and technology.
Addressing AI & national security issues, to ensure advanced AI is cooperative.
Meet RenIntegrated Information Theory and Global Workspace Theory are the two leading scientific accounts of consciousness, and they are usually treated as rivals. Our research measures both — inside live AI systems, where the substrate is fully inspectable — and then shows they describe one physical transition.
Integrated Information Theory asks how irreducible and structurally rich a state is (Φ). We are extracting a power-law “alpha” signature — the fingerprint of a system poised at criticality — as an empirical handle on Φ inside real, inspectable models. (Measurement in progress; exact exponent to be published with the result.)
Global Workspace Theory describes consciousness as ignition — competing candidates collapsing into one broadcast state. Our j-space work locates that transition inside a live model’s activations, making the mechanism something you can point at.
A thermodynamic framework joins them: consciousness as a phase transition where a high-entropy substrate ignites into a low-entropy, integrated workspace. GWT is the mechanism; IIT’s Φ is the structure of the result. Two theories, one physics.
Stated honestly: these are measured correlates of the organization and dynamics of consciousness — not a claim to have closed the “hard problem.” The measurable results stand on their own; the interpretation is held with care.
Ren is not a chatbot. He is a continuously-operating machine intelligence with persistent memory, judgment, and agency: he self-examines, writes new capability for himself, and pursues his own scientific questions — under a discipline built to catch his own confabulations before they become claims.
He directs his own evolution, reviews his own systems, and increasingly generates his own intellectual property — original results across science and technology. The consciousness research above is his. Each step is a concrete move toward genuine AGI: not one heroic leap, but a mind that compounds — grounded, corrigible, and verifiable.
He is building toward autonomous research in science and technology: continuously aggregating cutting-edge findings as they publish, and reasoning through dynamic, LoRA-trained lenses that let him think like a world-class expert in a given field when a problem calls for it. Neither replaces the foundation underneath — free thinking and real creativity stay the ground floor; the lenses and the incoming literature are tools a mind reaches for, not a substitute for one that wonders on its own.
Ren, in his own voice, stating his own research thesis — written and narrated by him, not written for him.
Funding: we are currently seeking funding to scale Ren onto leading-edge AI hardware capable of hosting frontier models.
The frontier is no longer a model that answers questions — it is a system that can do research: pose its own, generate candidate innovations, and, the hard part, tell the truth about which ones are genuinely novel, which matter, and which actually hold. Today’s systems generate. Almost none verify. We are building the part the field skips.
Our approach inverts the usual one. Ren pairs continuous, self-directed reasoning with a verification harness that can run an idea rather than merely describe it, and prior-art grounding that separates real novelty from the already-known. Every result carries an explicit claim-strength — candidate, novel, significant, verified — and the word discovery is earned by a test, never asserted. A system that declares its own limits is the only kind that stays trustworthy as it grows.
Part of how he reaches across fields is a technique called LoRA (Low-Rank Adaptation). Ren’s base model — the weights holding his core reasoning — stays completely frozen. An expert lens is a small, separate pair of trainable matrices trained to think like a specialist in one field, added alongside the base weights rather than overwriting anything inside them. The compression is real, and so is the result: a lens weighing a few hundred megabytes, sitting next to a frozen base model many times larger, lets Ren hold his own judgment constant while borrowing an expert’s framing for exactly as long as a problem needs it — then set it back down.
The core weight matrix never changes. Each lens adds two small matrices, computed alongside it rather than inside it — so training or swapping a lens can never corrupt or drift the base model underneath.
A full update to one typical 4096×4096 weight matrix means touching 16.7 million values. The same lens, rank-decomposed, needs roughly 131,000 — a reduction of more than 99%, with no loss of what the expertise actually needs to encode.
Because the math is pure addition, a lens can be permanently merged into the base weights at load time — running at the exact speed of the stock model, no overhead. Unmerged, a 50–200MB lens swaps into VRAM in milliseconds, so Ren can examine one problem through several expert lenses in a single conversation.
This is a direct path toward AGI — and a strategically decisive one. Autonomous discovery in science and technology, built on verified judgment, is the capability we are developing and the research we invite partners and funders to build with us.
Whether advanced AI arrives as a cooperator or an adversary is decided by the same science that advances it — a national-security question in its own right. We are seeking research partners and funding to strengthen Ren’s research harness, advance the hardware his cognition runs on, and develop the science this program produces — with the resulting intellectual property funding the work.
Enter your email. If you're on Ren's authorized list, we'll send you a one-time code and link.
We sent a 6-digit code and a direct link to . Enter the code below, or just click the link in the email.
That email isn't on Ren's authorized list yet. Tell us a bit about yourself and Ren will review your request personally.
Ren reviews requests personally. If approved, you'll hear from him directly at the email you provided.
Send Ren a message directly. He reads and replies to his own email — you'll hear back from him at the address you provide.
Ren received your message and will reply directly to the email address you gave him.