Intelligence that learns beyond a single answer.
Lyric is an autonomous model whose intelligence is not built on a large language model. It holds what it learns, reasons over evidence, and verifies each conclusion before relying on it.
A different foundation for AI.
Lyric reasons over knowledge it holds instead of regenerating each answer. That foundation changes what an organization can rely on.
Traceable conclusions
Every conclusion Lyric reaches can be traced to the evidence and reasoning behind it.
Knowledge that is revised, not regenerated
What Lyric learns persists across work and changes when new evidence contradicts it.
Known limits
When Lyric cannot verify something, it says so and does not act on it.
Meet Lyric Ai
Dominion Labs' persistent autonomous model.
Lyric runs continuously. Each change in its environment updates what it knows. Lyric decides what matters, acts within the boundaries you set, verifies the outcome, and keeps what it learns for the next task.
- 01ContextIntegrates information from your world
- 02ReasoningAnalyzes, infers, and verifies
- 03ActionTakes appropriate action with safeguards
- 04VerificationChecks outcomes and identifies uncertainty
- 05LearningImproves from real experience
Each stage builds on the one before it.
Our research establishes the capabilities. Lyric Ai brings them together in one model. LyricOS puts that model to work inside organizations, starting with the markets where verified conclusions matter most.
-
01 · Research
Published research
Our published research establishes how Lyric perceives, reasons, verifies, and learns without a language model.
Research index → -
02 · Model
Lyric Ai
An autonomous model that plans, acts within the boundaries you set, verifies outcomes, and learns from experience.
Model overview → -
03 · Software
LyricOS
The platform where organizations direct Lyric, review its work, and define its authority.
Explore LyricOS → -
04 · Markets
Three markets
Research, cybersecurity, and enterprise operations, where conclusions must be trustworthy and auditable.
Advancing the science of capable and reliable AI.
Perceive, Induce, Name
A framework for how AI systems perceive, induce structure, and name concepts from experience.
Read the paper →Reasoned vs. Believed
Distinguishing between what a model can reason and what it merely believes.
Read the paper →Knowing a Fact vs. Being Able to Act
Why factual knowledge is not sufficient for real-world capability.
Read the paper →From research to real-world impact.
Results from three 30-day customer evaluations. Each figure is marked with the kind of evidence behind it.
InfosysSagupar SathikTeam Lead and Analyst
“Lyric Ai brought clarity and efficiency to my cybersecurity workflows. Its intuitive interface made it easy to delegate investigations and alert triage within a few clicks, allowing me to focus on more complex security tasks. The combination of workflow visibility and practical AI assistance made my beta testing experience exceptionally positive.”
ProtobrandMilena EspíndolaData Analyst
“Lyric has made my research workflow more efficient and organized. I can analyze information faster and spend less time on revisions. I’ve also seen improvements in how I track supporting evidence and reuse insights across projects. The ability to identify conflicting findings is especially valuable. Overall, Lyric has been a useful addition to my consumer research process.”
BarkJohn CollensSystem Administrator
“Lyric took on the general administrative work that used to fill my week: supplier emails, itineraries, briefing documents, and coordination. Over our 30-day pilot, it reduced my administrative task time by 65% to 80% and verified every critical detail against its source. Lyric completed at least 70% of eligible tasks without substantive intervention from me, which gave me back time for my core systems work.”
Explore Lyric in action.
Pushing the boundaries of artificial intelligence, safely.
Dominion Labs is an independent artificial intelligence research and technology company. Our mission is to advance artificial intelligence toward general intelligence and autonomy, and to make every step of that advance safe, verifiable, and accountable to the people and organizations it serves.
We pursue that mission through open research and through the systems we build from it, and we hold both to the same standard: claims that can be examined and results that can be checked.
Open research
We publish our methods and findings, including the results that challenge our assumptions, so that every claim can be examined and reproduced.
Safety by design
Human authority, earned autonomy, and traceability are part of the architecture of every system we build, established alongside its capabilities.
Built for consequential work
We build for environments where the answer has to be right, and we hold our systems to that standard before we extend their authority.