Actor Infrastructure Platform

The foundation layer of the Actor ecosystem

AIP provides all the Actors, Services, runtime, communication, and security required to deliver distributed, concurrent, and autonomous computing — from edge devices to high-performance computing.

It underlies every PointOfData.ai product, so understanding it is important. The model is simple and clear.

Footprint on a 64-bit machine

Actor binary
786 KB
Actor memory at startup
5 MB
Neural Memory binary
894 KB
Neural Memory at startup
10 MB

Native by design

It runs on a bare operating system

Written in C

Actor and Evolutionary Neural Memory services need no running software beyond the base OS.

No layer cake

No interpreter runtimes, no phalanx of packages, no version dependencies to keep in sync.

POSIX everywhere

AIP supports all modern POSIX-compliant operating systems, from edge devices to HPC.

Actor architecture: operating system with Services connecting to an Actor via socket messages

Actors and Gateways

Everything talks over sockets

An Actor Gateway accepts messages, routes them to the correct Actor as necessary, and handles the message returned from the Actor to the requestor. It manages the Actors assigned to it, turning them on and off as required.

Actor Gateways communicate with everyone — other Actors and any other software — using TCP sockets, a high-performance, flexible, well-proven communication protocol.

Messages

Every message is an Event

An Event is something that happens at a specific time and location, sent from a specific entity to a specific entity. For example, edge software captures telemetry in surgery and sends it to Evolutionary Neural Memory.

Messages are like e-mails, but they carry instructions and data for what to do: store an Event, search Evolutionary Neural Memory, request a service, and infinitely more based on design requirements.

Event Object carrying its embedding, typed fields, validity interval, source, confidence, Tags, and Links

Services

Host any runtime

Services run the software an Actor uses, hosting it in the appropriate runtime. Running Llama-3, for example, uses the llama.cpp kit for a performant solution. This lets an engineer run and support any software in an Actor architecture.

Security

Layers you control

AIP manages security for the creator, giving full control over the layers of security, including SSL.

Evolutionary Neural Memory service architecture: Event Objects with unlimited Tags and Links

Evolutionary Neural Memory

The database runs as its own Actor

Our native Evolutionary Neural Memory DB service runs as its own Actor type, and its service architecture has unique characteristics that are important to understand.

Searching high-fidelity context requires techniques far beyond simple keyword or vector searches. Evolutionary Neural Memory search is programmable with conditions and branches, pattern matching, and mathematical searches across Events, Tags, and Links.

Open source SDKs

Build on Pod-OS

Public clients for Python, Rust, Go, and Java — plus a CLI for live gateway testing. Assert claims, link them, and query as-of state from your language of choice.

View all developer tools →

What this adds up to

A simple architecture that expands infinitely

By design, both AIP's platform software and storage are performant and efficient. That same small footprint scales without change from an embedded sensor to a hypercomputer, providing distributed, concurrent, and autonomous computing.

AIP is the foundation of the Actor ecosystem.

Product documentation

Pod-OS Documentation

Gateways, Actors, Evolutionary Neural Memory, deployment, and how to add and query data.

Open pod-os.com docs →

Build on the platform

Free access for builders, and direct conversations for teams evaluating Actor-based infrastructure.