Skip to content
Vinxi/Kernel THE BOOK OF RECORD

The Reality OS experience gallery

Charter: Spikes that reached a verdict, promoted from notes/ — what we tried, what we learned.

The Reality OS experience gallery

What it is. A set of standalone HTML explainers that answer the question every non-engineer asks — “where is NetworkAccess in all this kernel work?” — by showing the kernel dressed as the product: working, telecom-native screens for every seat in an operator, from the COO’s morning brief to a field engineer up a pole in the rain. The engine (ports, verbs, ontology, Acts, the statement log) is invisible on the surface and made explicit only in a labelled “under the hood” strip beneath each screen.

Why it exists. The kernel is abstract; sales, product and executives need to see the car, not the engine. These galleries are the presales / vision artefact — illustrative UI on sample data, faithful to the SmartInventory · SmartProject · SmartOps user stories (v3.1) and the modernization streams (configurable hierarchy, pluggable maps, AI-RCA, predictive analytics, ring/SPOF).

The artefacts

  • The full experience gallery — the centerpiece. ~37 screens across seven parts: SmartMarket & SmartFeasibility (sense & qualify) → SmartInventory (survey, auto-planning/DTP, BOQ, splice/OTDR, IE acceptance) → SmartProject (execution monitoring, validation queue) → SmartOps (dispatch/workforce, patroller, SmartSignal, assurance & AI) → Perception & attention (vision AI, spatiotemporal attention) → Configurability (N-level hierarchy, depth-aware RBAC, pluggable maps) → Make it yours (dashboard builder, node-based workflow engine, automation + in-app scripting, an AI build-assistant, object/trait/link/form/action/flow builders, and an agent layer — Studio, Manager, Library, multi-agent Situation Room, Approvals inbox). Every screen carries an ⚙ under-the-hood map to the kernel concepts that make it work.
  • Operations day-in-the-life — the first, tighter gallery: executive cockpit, NOC war-room, customer care, acquisition, troubleshooting, planning, projects, materials, inside-plant, field app, and the AI copilot.
  • Where NetworkAccess emerges, week by week — the milestone story: how the kernel becomes the product in a demo-able weekly arc.
  • Reality OS — competitive positioning — the category argument: incumbents bolt AI onto a system of record; we built the world model the AI was missing, and absorb each competitor’s best idea natively.
  • Kernel component mind-map — the engine view: clients → transport seam → verbs → core → the 14 ports → adapters, plus the roadmap.

What we learned

  • The product is the kernel, dressed. Every business feature maps cleanly to a small set of kernel primitives; the per-screen “under the hood” strip made that traceability legible and became the strongest argument for why each kernel concept earns its place.
  • Extensibility is the demo. The userland layer — building new dashboards, workflows, objects (with reusable traits), relationships, forms, actions and field-app flows without a code release — is what makes “keep growing” concrete for a business audience.
  • Agents want to feel hired, not scripted. Framing background agents as teammates you hire, give a beat, and hold on an autonomy dial (watch → suggest → act-with-approval → autonomous) — with approvals and an audit trail — reads as empowering rather than threatening, and reinforces the “AI advises, humans decide” invariant.
  • Ambient intelligence belongs on every screen, not a dedicated tab: a global attention pill and a working agent dock made the whole surface feel alive.

Honesty notes

Illustrative UI on sample data — not a running app. Several “native” kernel mappings point at capabilities that are early-stage rather than shipped (agent-host, /sys/describe, branch/simulation via the LTAP experiment, DuckDB spatial); treat those cells as emerging in front of a technical audience. Source galleries were authored in a workspace scratch directory and promoted here.