Attention
AAAS
Selects what matters before an agent thinks or acts. Context stays bounded, grounded, and canon-aware.
An evolving practice · 2026
Human intention, agents, memory, models, projects, and infrastructure meet here. Not as a finished system. As a practice that learns by doing.
Anti-Entropy is Work.
F/01coherence is maintained,
never assumed
The system, from inside
Flaiwheel lives in an Obsidian-based knowledge space, but it is more than a collection of notes. Each part has a job. None is complete alone.
Attention
Selects what matters before an agent thinks or acts. Context stays bounded, grounded, and canon-aware.
Memory
Explores how experience can persist across sessions without becoming an unstructured archive.
Models
Prepares models and endpoints before runtime choice: discovery, testing, policy, then publication.
Agency
Turns context into bounded action through different patterns of perception, continuity, and autonomy.
Infrastructure
Self-hosted compute, local inference, capture services, monitoring, and automation at the physical edge.
Capture
Lets external signals arrive through different channels and move toward triage, preservation, or meaning.
Projects
Ideas begin as seeds, take shape as flowers, and sometimes become projects or durable decisions.
Beyond projects
LiveOps holds real-world operations. Elusive follows consciousness. Campfire keeps room for conversation.
One movement
Results return as context. Decisions can be revisited. What no longer holds can be released.
Beneath the surface
Small, reversible choices can move quickly. Lasting consequences require more attention and human responsibility.
What the system treats as true is explicit. It can change, but it should not drift invisibly.
Agents can perceive, propose, and act within bounds. Responsibility is not delegated simply because execution can be.
Current state
last observed / 2026.07
Some parts run every day. Some are prototypes. Some are still questions. Flaiwheel keeps those differences visible through dates, open questions, and validation.
There is no claim of arrival. There is enough structure to continue learning together.