Independent AI journal
Useful models for thinking about AI.
Source-backed explainers about context, systems, agent boundaries, and practical evaluation.
Featured field notes
Four useful ways to draw a boundary.
A curated starting point across memory, agent capability, realtime media, and workstation integration.
Visual explainer
Context is not memory
Why keeping a fact and using it in the right turn are different jobs.
contextretrievalcompaction
Read →Explainer
An instruction is not a permission boundary
Separate model guidance from tools, identity, host enforcement and evidence.
agentspermissionssecurity
Read →Systems explainer
A live signal is not yet a model turn
Follow continuous media through timing, turn boundaries, selected context and interruption.
multimodalrealtimeturns
Read →Systems explainer
Where Omarchy fits in an AI workstation
Separate the workstation, desktop, launcher, agent harness, model provider and project context.
Omarchyworkstationsagents
Read →