Architecture status date: 2026-07-18.
MD-OS (Artificial Prefrontal Cortex) is a prototype of a bounded quasi-autonomous cognitive agent and its persistent operating context. It is not demonstrated autonomous general intelligence.
MD-OS (Artificial Prefrontal Cortex) v5.0 is the Markdown-native Operating Filesystem compatibility release line that carries this prototype for persistent, inspectable, replayable agent and robotic operation.
It can be described as the Operational Context as Filesystem paradigm for persistent AGI-like operation: not the intelligence itself, but the architecture that turns a reasoning engine into a durable, bounded, correctable, memory-bearing operator. In this paradigm, intelligence is not only a model response. It is a loop of remembered goals, readable state, bounded action, connector-mediated perception, deterministic rebuilds, audit, replay, and human correction.
It is also a natural-language robotic-agentic programming substrate. The goal is to program the operating behavior of an ecosystem, not only one assistant or one robot command: humans, host runtimes, MCP resources, work tools, devices, sensors, robots, policies, telemetry, approval gates, and recovery paths.
The core claim is therefore:
MD-OS (Artificial Prefrontal Cortex) is a bounded quasi-autonomous cognitive-agent prototype.
It can support self-directed research and problem solving inside an authorized
and delimited task environment.
It is not demonstrated autonomous general intelligence.
MD-OS is the persistent Operating Filesystem that makes those bounded cycles
inspectable, verifiable, replayable, and correctable.
Self-directed means that the prototype can choose intermediate questions, hypotheses, tests, plans, and allowed actions after the objective and authority envelope are supplied. It does not mean that it may create its own authority, run indefinitely, bypass approval, or certify its own success.
If intelligence becomes persistent operation, it needs more than a model and a context window. It needs:
MD-OS provides that layer through Markdown, JSON, NDJSON, bounded connectors, and deterministic builders.
Stated plainly, this layer sits between MD-OS and real host substrates: the operating system, hardware, peripherals, applications, desktops, filesystems, terminals, browsers, APIs, services, queues, robots, controllers, sensors, and actuators. MD-OS coordinates natural-language intent over those substrates; it does not replace them.
The MD-OS paradigm treats increasingly general, persistent cognitive operation as an operating-system problem as well as a model problem.
In this view:
This is why MD-OS can grow and evolve: new repeated needs become readable knowledge, natural-language programs, connector contracts, deterministic scripts, tests, and runtime state. The system expands by making new capability explicit and bounded.
The cross-domain extension path is defined by the Cross-Domain Cognitive Unity Model: Cortex constructs competing frame-transformation laws, seals a unique candidate before target evidence, verifies preserved relations and consequences, and reuses only hash-bound verified transformations. This is an implemented bounded mechanism and promotion contract, not a demonstrated AGI result.
MD-OS does not replace:
MD-OS coordinates the persistent operating context above them.
The optional continuity service is a runtime convenience, not a consciousness
claim. It makes the operating layer easier to keep interactive by maintaining
heartbeat and rebuild cycles, while the durable source of truth remains the
readable filesystem state under md-os/.
A Raspberry Pi can host MD-OS as a physical edge node above Raspberry Pi OS or Linux.
Raspberry Pi OS / Linux = device runtime, drivers, networking, and low-level execution
MD-OS (Artificial Prefrontal Cortex) v5.0 = persistent agentic memory, state, connectors, audit, and replay
This is useful for sensor collection, local automation, lightweight robot
supervision, monitoring, MCP serving, and bounded device coordination. MD-OS
keeps continuity in readable files under md-os/ops/ so the node can resume after
reboot without relying on hidden chat history.
Model = reasoning engine
Tools = hands and sensors
Context window = short-term attention
MD-OS = persistent Operating Filesystem, memory, and action ledger
Expanded formula:
AGI-like operation = reasoning engine + bounded tools + durable memory +
connector-mediated perception/action + deterministic rebuild + audit/replay +
human correction
MD-OS = the Markdown-native Operating Filesystem that stabilizes those parts
into an inspectable persistent control plane