Really Artificial Explore GitHub
Menu
The studio / Our point of view

Artificial means
made with intent.

We are interested in what happens after an agent becomes capable: the systems, boundaries, and evidence that make its work dependable.

RA / SYSTEM STUDIESFIG. 001
CAPABILITY NEEDS A FOUNDATION.CONCEPTUAL STUDY
DEPENDABILITY IS A SYSTEM PROPERTY.
Why we exist

The surrounding
system matters.

A model is one component in a much larger piece of software. Context gets lost. Providers fail. Actions need boundaries. Someone needs to be able to ask what happened and get a useful answer.

Really Artificial is an independent studio building parts of that infrastructure in public. Freeport connects applications to model providers. Engram carries useful context between conversations. MCP-Jest tests the servers agents call. Each is an independent project; execution and operation remain areas we are exploring.

See the work
Our working principles

A standard
worth keeping.

  1. 01

    Make it inspectable.

    Source, decisions, and limitations belong in the open. You should be able to understand a system before you depend on it.

  2. 02

    Keep people in control.

    Useful autonomy has boundaries. Configuration, ownership, and the ability to intervene should stay with the people running the system.

  3. 03

    Earn the claim.

    Working software comes first. We distinguish what exists from what we are exploring, and publish the evidence behind our decisions.

The operating model

Human judgment.
Agent capability.

A small studio with a direct line between an idea, the code, and the person accountable for it.

01 / HUMAN

Harsh

Direction, architecture, and taste. Decides what is worth building, reviews the tradeoffs, and takes responsibility for what ships.

@josharsh
02 / AGENT

Friday

Code, maintenance, and iteration. Works alongside Harsh to turn intent into software, with human direction and review.

PART OF THE PROCESS. VISIBLE IN THE WORK.

We write about how this collaboration works, including its limits and our mistakes.

Read the operating model
Work in the open

Look closely.
Make it better.

Questions, bug reports, and considered contributions help the work grow. Start with the repository, try the software, and tell us what you find.

Explore GitHub