ARCHITECTURE & EVALUATION OVERVIEW
AMOS™ is a proprietary reasoning-governance architecture built on The ANCHOR™ Method. It governs how AI uses its intelligence throughout a conversation, what it treats as evidence, what it examines, and how it develops and revises its contribution.
AMOS finds the knot and helps people see what they’re missing, so they can reason their way forward.
Implemented within GENE Alpha, the architecture produces contributions we can record and examine as conversations develop. Our next ambition is to bring AMOS into other AI products and applications.
The model supplies the intelligence. AMOS governs its participation. The human keeps the pen.
A person may know the facts and still lack a perspective, distinction, or connection that changes how they understand the situation.
AMOS works from the conversation to direct the model’s contribution toward what could make the reasoning more useful. Its interpretation remains open to correction as the person responds.
AI answers. AMOS governs how it answers.
Built on The ANCHOR™ Method, AMOS governs how the model uses evidence, develops its response, and adapts as new information arrives. This helps you see what you’re missing and reason your way forward.
Work from what the person has shared, distinguish it from assumptions, and consider what help they’re actually seeking. Develop useful reasoning without jumping to a plan, recommendation, or explanation the conversation hasn’t called for.
A person may know a great deal yet struggle to draw on it when stressed, overwhelmed, or caught in one interpretation. AMOS uses what they share to govern the kind and amount of help offered, so the response helps make perspective, understanding, or choice more accessible.
That assessment comes from the conversation and remains open to correction.
Connect relevant information, examine competing explanations, and follow implications far enough to make a meaningful contribution. Offer supported conclusions and practical help when warranted.
Build on what the person has shared so they don’t have to start from scratch with every turn. Revisit earlier interpretations when new information arrives, rather than treating previous conclusions or decisions as commitments the person must keep.
Across separate conversations, our aim is to preserve relevant context while allowing people to change their minds, revise their understanding, and move forward.
Our architecture governs how AI contributes throughout the conversation, without prescribing fixed reasoning steps. The model develops the reasoning the evidence supports. The person remains free to question it, change direction, and choose.
GENE™ is our first working prototype, now entering its third round of live-user Alpha testing. It turns AMOS from an architectural proposition into a system we can observe, challenge, and develop.
We examine recorded conversations turn by turn: did an assumption get questioned? Did new information change the response? Did the AI develop a useful connection or miss what mattered?
We build case studies from those records and participant feedback, examining where understanding shifted and what the system contributed. Alongside this work, blinded comparisons and repeatable test harnesses help us investigate what Method governance changes.
The question driving the work: can we govern AI’s intelligence so that people gain greater access to their own?
Early GENE Alpha participants describe recognizing patterns, seeing their situations from another perspective, and feeling relief as something previously tangled becomes clearer. Feedback also includes appreciation for practical help that supports their decisions without taking them over.
These accounts inform what we develop and examine next: which contributions people find useful, what becomes clearer, and what they feel better able to work through. They are encouraging early feedback, distinct from evidence of sustained improvements in reasoning or critical thinking.
We look for concrete moments in recorded conversations: an assumption reconsidered, a new perspective becoming available, or a distinction that changed how the person understood their options. We examine the system’s contribution alongside what the participant reported finding useful.
Alongside live-user research, we use blinded comparisons, repeatable test harnesses, and assessments from multiple AI model families as well as human review. These help us examine how faithfully AMOS governs responses, where it adds value, and where further development is needed.
Implementation integrity examines the runtime and governing inputs used for a contribution. Version information and structured records support traceability.
Reasoning fidelity examines evidence discipline, development of implications, usability, responsiveness to correction, and preservation of human authorship.
Participant feedback and human-outcome research examine whether the contribution helps people understand, reason, and choose. A faithful implementation and a beneficial human outcome are separate questions.
Our blinded comparison work includes responses from the same underlying model with and without Method governance, presented under neutral labels for review. Preserved responses and reviewer records support examination of those comparisons.
We are continuing to develop evaluation of what AMOS adds beyond strong instructions alone. Assessments by different model families help examine contributions; they do not establish that AMOS preserves fidelity when those models generate the conversation.
AMOS is being developed as a reasoning layer companies could integrate into the AI products they already use or build without replacing their existing technology or preferred AI models.
| Area | Status | What this means |
|---|---|---|
| GENE Alpha | Implemented | Working AMOS runtime within GENE’s existing infrastructure. |
| Model portability | Cross-model validation underway | Testing whether AMOS preserves its reasoning governance across different generating models. |
| External integrations | Development objective | Exploring licensed access to AMOS within third-party systems; delivery mechanisms and integration requirements remain under development. |
| Licensing | Commercial development objective | Potential models include key-based access to a hosted reasoning service or licensed implementations, subject to technical and fidelity validation. |
RESEARCH & DEVELOPMENT PARTNERSHIPS
We’re inviting technical teams, researchers, and AI product builders to collaborate on model portability, rigorous evaluation, and integration into other applications.
If your work connects with reasoning governance, human judgment, or the design of AI interactions, let’s discuss a concrete opportunity to work together.
Discuss a partnership ↗