Est. 2024 — Roxbury, Massachusetts
Civic infrastructure for the age of AI agency.
A Massachusetts Public Benefit Corporation. We build open standards, schemas, protocols, and focused institutional software — public goods, released under open licenses, designed to outlive any single vendor, including us.
The thesis
The history of operator tooling has been a sequence of new sovereignties. Spreadsheets gave non-programmers computational agency. Databases gave them data agency. The web gave them distribution agency. Each was a new kind of control over a previously inaccessible domain.
The current moment requires a new one: cognitive agency in the age of AI. Not AI that thinks for you — that’s the path toward losing sovereignty. A substrate where the operator — the human whose judgment steers the system — is irreducible, AI handles the labor, the substrate itself accumulates knowledge, and the system surfaces what’s missing so the operator can decide what to do next.
None of this exists yet as public infrastructure. Most of what’s being built treats AI agency as a service to subscribe to or a product feature to embed. We think it should be substrate — owned by operators, federated by choice, signed at every layer, accountable through provenance.
That’s the work.
What we’re building
A public AI code qualifier.
Most code now passes through AI assistance somewhere in its lifecycle. Almost none of it carries verifiable provenance about what was generated, what was reviewed, what was tested, and what holds up under scrutiny. We’re building open infrastructure for qualifying AI-assisted code — not gatekeeping it, but making its qualities legible to the institutions that have to rely on it.
A law API.
Statutes, regulations, and case law are public information trapped in formats that make them practically inaccessible to the institutions and individuals they govern. We’re building open APIs and schemas that treat law as the queryable public substrate it should already be.
Civic OSS standards and protocols.
Identity for AI agents in regulated contexts. Auditable composition for multi-agent systems. Sedimentation models for institutional knowledge. Each piece submitted to the standards process where standards belong — NIST, NCCoE, W3C, IETF — and released as reference implementations under licenses that survive us.
Focused institutional software.
Tools built for specific public-interest use cases, sized small, owned by the institutions that use them, federated where federation helps and local where locality matters more.
How we work
The method is the product. Before we build anything, we run a structured process for identifying what’s missing in the current landscape, what already exists, where the prior art is, and where the genuine gap sits. The method is documented, repeatable, and auditable. It’s also the method we use to evaluate our own work — every artifact we ship is itself the output of a gap-hunting cycle.
Public submissions, public artifacts. Our work appears in the federal register before it appears in our marketing. Our NIST/CAISI submission on AI agent system security (Docket NIST-2025-0035) proposes the formal vocabulary of Intrinsic Access Control and the Enforcement Location Principle as candidate elements of federal AI guidance. We participate in the standards process in public.
Long horizons, small teams. We’re deliberately small. We work on civic infrastructure that takes years to mature and decades to matter. We’re a Public Benefit Corporation specifically because the work doesn’t fit cleanly into the venture timeline. The legal structure binds our charter to public benefit; the small team keeps the work coherent. We sustain through grants, implementation consulting, and scoped contracts — revenue models compatible with public-interest work.
Structure
The infrastructure layer of the agentic economy is being built right now. Most of it is being built by entities whose primary obligations are to private shareholders. Some essential pieces of that infrastructure — identity, qualification, the legibility of law — should be public goods. Not “open source as marketing” public goods. Actual public goods, with charters that survive their founders.
Massachusetts allows Public Benefit Corporations specifically to legally bind a company to its stated public mission. Cube Commons is incorporated under that statute. Our charter requires us to weigh public benefit alongside financial returns. The legal structure is the commitment.
What we’ve published
Technical Companion: Response to NIST/CAISI RFI on Security Considerations for AI Agent Systems
March 2026 — Docket NIST-2025-0035
Coordination as Rendering — applying computer graphics mathematics to multi-agent coordination visualization
CC-TR-2026-003
InAC: Intrinsic Access Control as a Sixth Access Control Model
Formal specification
Living list. Will grow as we publish.
Who we are
Cube Commons, Inc. was founded by Barton Nicholls in Roxbury, Massachusetts. We work in the lineage of civic infrastructure — library science, public standards, federal records, open protocols — applied to the conditions that AI agency now creates.
The corporation is a Massachusetts Public Benefit Corporation. The office is in Roxbury. The work is small, deliberate, and intended to last.