The thesis

Why we're building this, now.

Models got good fast; the systems around them didn't. Everyone landed on the same word for what's missing — control — and most ship a layer that watches someone else's model and writes logs after the fact. That's a smoke detector. Octos is the building — the runtime your AI runs on and answers to.

Read the full manifesto →

Why now

Three things turned at once.

01

Intelligence is commoditizing

Open models now match the frontier on the work agents actually do. Once the model is a swappable part, the value moves to whatever holds the knowledge and the accountability around it.

02

We own the layer, not rent it

Octos is a runtime we control and can hand you to self-host — not a wrapper on someone else's API. That's what makes it sellable, defensible, and beholden to no single vendor.

03

Everyone's racing the wrong axis

The field is spending on raw capability. The layer that makes capability remember, behave, stay accountable — and become an app you can run anywhere is barely contested. That's the one we're building.

Roadmap

Working now — and what's next.

Personal OS runs today — and we built Octos with Octos, managing our own roadmap, decisions, and predictions on the same runtime we ship. Company Brain and a learning assistant come next.

Now Working

Personal OS

An always-on personal runtime — live today, local, and the system we run ourselves on.

Next

Company Brain

Governed AI for the organization — decisions logged and graded on real org data.

Soon

Learning assistant

A tutor that remembers how you learn and adapts — starting with chess.

Autumn

Marketplace

Publish and install Octos verticals — one runtime, many governed apps.

How it runs — today, and where it's going.

Today · Local
Runs inside your Claude, Codex, or Gemini subscription — you bring the model, Octos brings the governance. Nothing to host, nothing to spin up.
Next · In progress
Own local runtime — Octos runs standalone and points at any model, so it no longer depends on a vendor's agent app.
Later
Hosted instance — an always-on Octos in the cloud, for teams that would rather not run their own.

Where this goes.

The long game is a federated network of accountable runtimes — one per person, one per company, composing into something genuinely new. Every person's Octos holds their knowledge, tracks their commitments, and governs their AI. Every company's Octos holds the org's decisions, memory, and track record. They connect where they should and stay separate where they must.

The content economy follows: a creator builds a world in Living Worlds, it becomes the backdrop of someone's personal OS, and what they live feeds the creator's next chapter. Same runtime, same governed memory, different surface — for each.

The raise

Where we are — and what we're raising for.

Personal OS is already working — and we've built Octos with Octos, running our own roadmap, decisions, and predictions on it for months. We're raising a pre-seed to ship the Company Brain vertical, bring on design partners, and stand behind the open-source launch.

We're not trying to reach every fund — we want the few investors who already believe the durable value in AI is the accountability around the model, not the model itself. The ones who read "95% of enterprise AI pilots fail" and see why the answer is a governed runtime, not a better prompt. If that's you, the rest is just a conversation — no deck, no process. Reach out.

StagePre-seed · open to conversations
Live todayPersonal OS — self-hosted
ProofWe run Octos on Octos — months of real use
BuildingCompany Brain vertical
Open toInvestors · design partners · a good intro