AVRAI

privacy-first discovery and model infrastructure

Roadmap

The roadmap is split across product, runtime, and model lanes.

AVRAI is building three systems in parallel, so the roadmap should look like an actual roadmap. Product, runtime, and model work move on different clocks and only come together at release time.

An actual roadmap, organized by lane and phase.

Now

Architecture and core framing

Building

Active product, runtime, and model delivery

Next

Longer-horizon expansion

Product

Discovery, planning, attendance, operations

  • Public framing for places, lists, groups, and events
  • Clear separation between user and operator surfaces
  • Reservations, venue tooling, and deeper planning workflows
  • Better end-to-end flows from discovery to attendance
  • Broader service pathways and richer operator intelligence
  • Deeper multi-party planning across more contexts

Runtime

Identity, policy, transport, rollout

  • Privacy-first execution model and contract boundaries
  • Consent, policy, and fail-closed direction in the architecture
  • Stronger transport, rollout, rollback, and recovery mechanics
  • Operator controls that stay inside the privacy boundary
  • More robust AI2AI coordination and endpoint orchestration
  • Wider deployments under the same runtime standards

Model

State, scoring, simulation, planning

  • State schema, model contracts, and evaluation logic
  • A compact model profile for constrained devices
  • Replacement of heuristics with learned scoring and transition models
  • Planning over lists, groups, and operator actions
  • Broader world-model behavior and longer-horizon planning
  • More capable federated improvement and ecosystem intelligence

What the public record should eventually contain.

Deployments

Where the system is live

Public proof should show where AVRAI is actually deployed and what workflow it is serving.

Evals

How the model is judged

Model changes should be backed by evaluation summaries, shadow results, and release gates.

Releases

What changed and what improved

Release notes should explain what shipped, why it shipped, and what measurable lift was observed.

See the workloads inside the product track.

The platform page shows how discovery, planning, reservations, and operations fit together.

Platform

See what is driving the model track.

The reality-model page explains state, scoring, simulation, and evaluation in more detail.

Reality model