Skip to content

Engineering roadmap

Five stages, stated plainly

Stage 1 is live — it is this portal, the sandbox, and the certification engine. Stages 2 through 5 describe engineering work we intend to do. They are commitments of direction, not descriptions of shipped capability, and this page will say so until each one is true.

  1. 01Live now

    Now — this site

    Launch Portal & WebMCP Playground

    Establish the visual identity, publish the developer manuals, and ship an interactive WebMCP test surface.

    • Interactive portal with the dual-circle interoperability canvas
    • Developer integration guide for stdio (Group A) and HTTP stream (Group B) adapters
    • Client-side WebMCP sandbox with live JSON-RPC payload logging
    • 25-question technical certification engine
  2. 02Next

    Following Stage 1

    Stateless Host Engine & Pre-Execution Firewall

    Build a stateless host that ingests, parses, and verifies incoming tool-call payloads before anything executes.

    • Stateless JSON-RPC 2.0 handler over HTTP POST and local streams
    • Pre-Execution Firewall: schema assertion plus argument sanitisation
    • No sticky sessions — self-describing auth tokens carried in headers
    • Go reference implementation with an adversarial test corpus
  3. 03Planned

    After Stage 2

    Parallax Paradigm Execution Layer

    Physically isolate executing plugin runtimes from the host's reasoning environment.

    • Daemon spawning Group A sovereign plugins as isolated stdio subprocesses
    • Firecracker microVM manager for Group B integrations and high-bandwidth handshakes
    • Ephemeral volumes with cryptographic erasure at session end
    • A documented answer to the GDPR Article 17 'ghost vector' problem
  4. 04Planned

    After Stage 3

    Ledgerless Settlement Gateway

    Settle peer-to-peer usage without a shared ledger and without third-party gas fees.

    • HyperCycle wallet configuration inside the host runtime
    • TODA/IP asset files carrying local proof-of-work cryptosignatures
    • Stateless settlement receipts embedded in response headers
    • Performance harness asserting sub-5ms local state transitions
  5. 05Planned

    Long horizon

    Swarm Integration & Distributed Nodes

    Connect the runtime to A2A and AGNTCY routing, then push stateless coordination nodes to the edge — and eventually beyond it.

    • Automatic translation between flat-vector and structured-graph memory schemas
    • Unified multi-agent sessions across heterogeneous backends
    • Stateless microVM architecture ported to radiation-tolerant edge processors
    • Delay-tolerant networking for nodes outside terrestrial latency budgets

Verification

What has to be true before a release is tagged

These are the gates. Each one is a claim someone can check, which is the only kind worth writing down.

About the long horizon

The .space in ai4all.space is a statement of direction: stateless coordination nodes replicated far enough out that no single jurisdiction is a chokepoint. That is a genuinely hard, genuinely long programme — radiation tolerance, delay-tolerant networking, and power budgets are real constraints, not slideware. We name it because it shapes the architecture today: a host that holds no session is a host that can be replicated anywhere. We are not claiming it is close.