An Observatory for Applied Intelligence
Building systems that can explore more, remember more, and earn the authority to do more.
Parallax Provenance is an independent applied-intelligence research and product-development initiative founded and directed by Andrew Surber.
The initiative develops and investigates systems for turning information into useful, accountable intelligence—combining AI-assisted research, product development, evidence provenance, institutional memory, and independent evaluation.
The work to date includes a working market-intelligence prototype, an implemented experimental system for institutional research governance, and a completed controlled evaluation of AI-system integrity.
Market Weather, Sentinel, Cairn/SIA, and Assay are all projects within Parallax Provenance. Each explores a different part of the same challenge: how to make increasingly capable intelligence systems more useful without losing control of what they know, what they assume, or what they’re authorized to do.
The central idea
Better governance should make greater autonomy possible.
As AI systems become more capable, the question isn’t simply what they can accomplish. It’s what they can be trusted to do, under what conditions, and with what evidence.
Parallax Provenance studies applied intelligence: human- or machine-produced judgment intended to influence decisions or action.
This work asks how that intelligence can preserve its sources, distinguish observations from conclusions, recognize uncertainty, survive independent challenge, and earn greater responsibility as its reliability is demonstrated.
The goal isn’t to slow intelligent systems down. It’s to build the conditions under which they can safely go further.
The person behind the work
Andrew Surber
Founder & Principal
I founded Parallax Provenance to pursue questions at the intersection of technology, institutional knowledge, research, and accountable decision-making.
My role is to define problems, establish or adopt requirements and evidence standards, direct AI-assisted implementation, coordinate competing perspectives, and make decisions about what the results justify.
My earlier career in institutional technology, client services, contracts, privacy, and operations informs the work. I understand the practical consequences when systems disagree, processes break down, or an organization cannot explain why a decision was made.
Parallax Provenance is where I apply that experience to emerging AI capabilities.
Explore my professional background →
Explore how I work →
The project portfolio
Four projects. One evolving body of work.
The projects share research principles and technical ideas, but each has a distinct purpose and level of maturity.
Market Weather
Product development · Market intelligence · Governed research
Make complex environments understandable.
Market Weather is a working local market-intelligence prototype combining an interactive Forecast Center with structured research, historical replay, and evidence review.
It began with a practical product challenge: helping people understand changing market environments without confusing interpretation with prediction.
The work established reusable ideas around observation, research, contradictory evidence, replay, and the preservation of findings that do not survive scrutiny.
Status: Working research prototype; not a commercial deployment.
Explore Market Weather →
Sentinel
Institutional systems · Evidence provenance · Research operations
Give organizational knowledge a durable, accountable structure.
Sentinel is an experimental research application designed around persistent records, bounded authority, review requirements, and evidence-linked decisions.
Where Market Weather explored how intelligence develops, Sentinel investigated how the rules governing that intelligence could become enforceable system behavior rather than informal policy.
Status: Implemented experimental system; not institutionally deployed.
Explore Sentinel →
Cairn / Semantic Integrity Audit
AI evaluation · Independent challenge · Semantic integrity
Find out when an intelligent system changes the question without admitting it.
Cairn investigated whether an AI system could preserve the meaning and boundaries of an institutional question while working with evolving evidence.
Its Semantic Integrity Audit (SIA) provided an independent evaluation mechanism for examining discrepancies between an AI actor’s behavior and its own account of that behavior.
The controlled experiment identified one genuine integrity failure among 247 eligible audited pairings. Rather than treating favorable surrounding results as sufficient, I preserved the negative finding and constrained further experimentation, requiring the next experiment to justify its scope and existence.
Status: Controlled experimental research with a documented mechanism failure; not a validated commercial evaluation service.
Explore Cairn / SIA →
Assay
Curiosity without capture.
What happens when AI collaborators are encouraged to investigate the unexpected—but none is allowed to quietly take control of the investigation?
Assay is an emerging research concept developed from lessons across Market Weather, Sentinel, and Cairn. It would combine open-ended discovery, persistent evidence, competing interpretations, and independent review to explore questions conventional workflows overlook: unexplained patterns and contradictory accounts.
A plausible connection isn’t an established relationship. A rumor isn’t evidence. An AI collaborator’s confident interpretation isn’t permission to change the original question.
Its proposed checks and balances would let collaborators challenge and verify one another’s work while preserving the original objectives. The principle: cheap curiosity, expensive belief.
Status: Early research and design direction. The proposed integrated approach has not yet been demonstrated or validated.
How the work connects
Each project made the next question possible.
Market Weather explored how observations become research and how research survives—or fails—review.
Sentinel explored how evidence, institutional memory, and decision boundaries can be represented in persistent systems.
Cairn/SIA tested what happens when an AI actor crosses a semantic boundary and whether independent evaluation can detect the discrepancy.
Assay asks how those capabilities might work together: giving multiple collaborators the freedom to pursue unexpected discoveries while keeping the investigation anchored to evidence, independent challenge, and its original purpose.
This is not a claim that the four systems have already been integrated into a finished platform. It is the research direction emerging from the work completed so far.
Why Parallax Provenance exists
Intelligence should earn the right to matter.
A useful AI system needs more than the ability to generate a convincing answer.
It needs to distinguish what was observed from what was inferred, show where conclusions came from, preserve uncertainty, expose contradictions, and remain accountable when the evidence changes.
Those requirements become more important—not less—as systems gain autonomy.
Parallax Provenance investigates how to make them practical, measurable, and eventually useful to organizations that depend on complex information and consequential decisions.
The initiative is currently pre-commercial. Its projects are prototypes, implemented experimental systems, controlled evaluations, and emerging research concepts—not customer deployments or validated enterprise products.
The purpose of the work is to discover which capabilities deserve to advance, which assumptions fail, and what evidence would be required before broader claims or applications are justified.