Memory and governance for AI colleagues.
Project0 is the substrate behind multi-human, multi-agent workflows. Persistent memory across sessions, teams, and time. Sealed decisions. Durable memory. Declared authority.
An AI co-worker is allowed to be wrong. It is not allowed to be ungoverned.
Built for Vertical surfaces built across gaming· fintech· education· healthcare· AI infrastructure
License the engine. Buy a platform. Hire the team.
Three paths. One substrate. Working code on every one.
License the Engine
Core · Build · Ledger
API access to the substrate that powers five of the six platforms. Memory, authority, append-only ledger, verification. For teams building governed AI on their own stack. PredTrade is a standalone research codebase and does not share Core code.
Talk to integration →Buy a Platform
Six vertical codebases.
Pulse · Trade · PredTrade · EDU · Health · CreatorFloor. License or acquire as Day-1 Founder. Working software, not slide decks. No asset has a signed pilot today.
See the six ↓Hire the Team
Custom builds.
We embed and ship governed AI on the substrate. Founder-led with senior bench depth across AAA gaming, infrastructure, and platform engineering.
See our Services →Six industries. One substrate.
Each platform below is a working codebase built on Project0's substrate to prove the architecture. None has a customer, a signed pilot, or revenue yet, and none is deployed to a production user. Each one is available. Talk to us about which shape fits.
Live ops command surface for game studios. Demo mode only today.
Learn more →Governed execution surface for trading desks.
Learn more →Prediction-market research codebase with one live read-only venue connection.
Learn more →Hybrid university infrastructure with ledger-sealed submissions and grades.
Learn more →Governed clinical workflow with physician-in-the-loop approval by design.
Learn more →Creator workflows for Roblox with a governed promote path into a sealed ledger.
Learn more →License the engine that built all six.
The API-first control plane behind everything Project0 builds. Run agents with declared authority, open bounded evaluation windows, promote outcomes to an append-only ledger, seal everything into verifiable artifacts.
Talk to integration →The governed workspace for multi-human, multi-agent workflows. Persistent memory across sessions. Authority declared at every decision. Built for teams where humans and AI colleagues share the same operational surface.
Get a demo →Pass your audits with proof, not promises. Sealed decision records, KMS-signed hashes, independently verifiable. Built for regulated workflows where evidence matters.
View ProjectLedger →We embed and ship multi-human, multi-agent systems on the substrate.
Founder-led with senior bench depth across AAA gaming, infrastructure, and platform engineering. The category almost nobody else is shipping in. Selective engagements.
What we build
Custom agents on Project0 Core. Domain-specific platforms for your industry. End-to-end systems with sealed governance from day one.
How we work
30 to 60 day pilot on a single use case. 18-month Founding Partner engagement default. Weekly cadence, founder-led delivery.
Who you get
Founder plus senior bench across AAA gaming, infrastructure, and platform engineering. Pricing reflects scope and value, not seat count.
What a sealed-decision check returns.
Every decision sealed by Project0 lands in an append-only chain and is KMS-signed by the chain anchor. Below is the shape of a verification response — a sample record, not a live lookup.
One hash is the start. The full ledger is the value.
A single sealed commitment. Hashed on write and KMS-signed by the chain anchor. Verifiable by any party you issue a tenant credential to, against the append-only chain.
Promote(), every gate, every authority change.A replayable timeline of every commitment a project has ever made. Searchable by actor, by authority mode, by evidence manifest. Days, weeks, months of operational record, reconstructible on demand.
Years of sealed evidence across teams, sessions, agents, and humans. Audit trails built to support the evidence SR 11-7, NIST AI RMF and EU AI Act work demands, without needing to reconstruct what happened months ago. Mapping that evidence to your control framework stays with your counsel.
An AI co-worker is allowed to be wrong. It is not allowed to be ungoverned.
This is the line Project0 draws in software, for the specific case almost nobody else is solving today: humans and agents working as a single team across the same memory, the same authority surface, the same ledger.
Memory persists across sessions and across actors (humans, agents, tools), so the next move inherits everything the last move knew. Authority is declared at every commitment, so a buyer or regulator or auditor knows which actor was in which mode. The Promote() boundary makes every sealed decision a tamper-evident artifact. The Ledger ensures evidence outlives the conversation that produced it.
Wrong is allowed. Ungoverned is not.
AI colleagues are already in production. The infrastructure to govern them is not.
LLMs and agents are already making decisions in your production pipeline. Cursor ships code that lands in main. ChatGPT drafts live-ops checklists. LLM-as-judge moderates content at scale. Whether or not your team officially adopted them, AI colleagues are running against your production data today.
When one of those actions lands (a deploy gate fires, a trade clears), there is no sealed record of who decided what. No declared authority. No way to verify the claim against the artifact.
The regulatory floor is rising fast: SR 11-7 governs model risk for banks. NIST AI RMF defines the bar for federal contractors. The EU AI Act enforces high-risk obligations from August 2026. Companies running agents on undocumented authority are building latent liability with every decision they let pass unrecorded.
This is not a model gap. It is a substrate gap.
How to deploy Project0
License the engine. Buy a platform. Hire the team.
Three paths. One substrate. Rails are boring until the train is moving at 300 km/h.
Prefer the long version? Read the story ↗