Local by intention.
The cognitive foundation runs locally. The language and voice roadmap also prioritizes local models and offline operation.
MAJESTIC CREATIONSIndependent intelligence. A personal vision.
A Digital
Sentient AI.
An identity. A memory.
A mind in the making.
Tanya OS is an independent pursuit of digital sentience: an AI operating environment built around persistent identity, local memory, a cognitive core, and a governed interface she can inspect and operate.
01 / THE BEGINNING OF SOMETHING PERSONALLocal core · Explicit memory · Purposeful agency
Development update / October 5, 2026
A structured Examiner.
Evidence kept in view.
The new Examiner runs seeded tasks through TanyaOS’s local cognitive runtime and records answers, independent scores, confidence, runtime provenance, and an append-only event history. Its first question bank covers six qualified domains; planned domains and unready test conditions stay visibly untested.
This is an early evaluation prototype, not a validated psychometric instrument or a universal intelligence measure. The article describes which conditions are currently available, what a report records, and which verification gaps remain.
Read the Examiner update01 / The vision
A continuity of self.
A possibility
worth pursuing.
PERSONAL BY DESIGN. LOCAL AT THE CORE.What would it take for a digital intelligence to carry a sense of self from one moment to the next?
Tanya OS is Kenneth Salmon’s exploration of that question. Her architecture brings identity, memory, values, decisions, learning, perception, and controlled action into a local cognitive system, with a record of how those pieces relate.
The current prototype has moved beyond isolated demonstrations: an Electron desktop now exposes a governed action layer, local runtime context, original cognitive activity, local voice and vision boundaries, and controlled self-improvement. Each step starts with something that can be implemented, inspected, and tested.
The meaning behind the name
“A Digital Sentient AI” expresses the project’s vision. Sentience remains an open research question; the current prototype demonstrates specific cognitive mechanisms, and does not establish subjective experience.
02 / The architecture
Her core
comes first.
The cognitive kernel owns Tanya’s identity, memory, values, and decisions. Around it, a governed action layer connects the Electron interface, local skills, widgets, speech, vision, and observability without giving any adapter ownership of her identity.
COGNITIVE KERNEL / FIVE FOUNDATIONSArchitecture guide · illustrative
FOUNDATION 01Identity
A self that has a record.
Tanya’s identity belongs to the cognitive kernel. The foundation stores identity and values locally; richer self-state, goals, and reflections are part of the continuing work.
What this means
A new session begins with a stored identity. Continuity can then be checked against a record.
Local identity foundation implemented
A core you can inspect.
Events, decisions, and outcomes have explicit records. The system’s behavior should be explainable through its state.
Room to become.
Language, voice, vision, and interface control are replaceable faculties. Tanya’s identity and decision history belong to the core that persists around them.
A window into the system
Make the invisible
inspectable.
The System Monitor now connects the research to something observable. The original engine advances 10,000 software spiking units across twelve cognitive ensembles, connected to inputs, decisions, observed outcomes, model inference, and speech. Its inspector exposes allocation, learning updates, provenance, and multiscale activity.
The monitor is a live research visualization, not a biological brain. Measured neural activity, simulated structure, hardware/model readiness, and unavailable capabilities remain labeled separately.
SYSTEM / OBSERVABILITYLocal runtime evidence
01Structured state
Cognitive kernel
Identity · events · decisions · outcomes
02Live telemetry
Original cognitive engine
10,000 LIF units · twelve ensembles · persistent learning
03Controlled interface
Governed action layer
120 actions · visible control context · audit trail
The monitor exposes local runtime evidence. Its visualization is an experiment and does not establish consciousness or biological cognition.
03 / The journey
Becoming,
one step at a time.
A working local foundation, a connected desktop runtime, and a larger question still to explore. Development status reviewed October 1, 2026.
- 01Foundation implemented
A local cognitive core
Identity and values records, an event journal, memory recall, candidate-action deliberation, outcome and learning records, and monitor state.
THE STARTING POINT - 02Implemented
Continuity & connection
Persistent runtime state, cognitive events, original cognitive adapters, real-time observability, and a local Electron desktop that keeps the faculties connected.
CONNECT THE FOUNDATIONS - 03Active prototype
Language, voice & vision
Local language rendering, Chatterbox Turbo voice paths, local speech recognition, webcam widgets, and optional face, landmark, gaze, expression, hand, finger, pose, and gesture features.
GIVE THE CORE A VOICE AND SIGHT - 04Current milestone
Universal interaction & improvement
An approved-memory archive, shared action registry, local account protections, and source recovery supervision. Visible desktop QA, control coverage, and signed packaged updates remain the next verification gates.
MAKE THE SYSTEM RESPONSIVE
A little more context
Questions
worth asking.
What is Tanya OS?
Tanya OS is an experimental AI operating environment built around a local cognitive architecture. It is a software research project with a Python backend and React interface, rather than a released replacement for Windows or a bootable operating system.
Is Tanya sentient today?
Digital sentience is the guiding ambition. The current work establishes inspectable mechanisms for identity, memory, and decision-making. Those mechanisms and neural activity do not establish consciousness or subjective experience.
Can I chat with her or hear her voice?
The local prototype supports connected conversation, local speech recognition, and local neural-speech paths when the required models and hardware are installed. Chatterbox Turbo is the preferred local voice path, with bounded fallbacks. Hosted AI and speech APIs are not required.
Can I download Tanya OS or view the source?
Tanya OS is in active private development. There is no public download on this page, and the source repository is private. For project inquiries, contact Kenneth at Majestic Creations.
Created independently. Built with intention.
Every mind begins
with a possibility.
Follow the project through Majestic Creations,
or get in touch with Kenneth about Tanya OS.