VENTURE · WORKING SYSTEMS
Different domains. The same demand for evidence, boundaries, and learning.
These systems span enterprise representation and RFP assurance, provenance, AI assurance, accounting, travel, investing, information governance, executive operations, creative production, literary work, and career decisions. The subject matter changes; the recurring architecture does not: authoritative state, explicit uncertainty, bounded decision rights, adversarial challenge, auditability, and outcomes that can correct the next decision.
The portfolio is ordered by evidence maturity, not by attachment to the idea. ERI / Auctrel leads.
CURRENT PORTFOLIO
Eleven systems, with an explicit evidence hierarchy.
The systems are not presented as eleven co-equal propositions. Auctrel / ERI leads because it is the portfolio’s clearest current expression of governed enterprise AI applied to a consequential commercial workflow. The remaining flagships follow on public evidence, technical specificity, and strategic relevance; active and specialist systems retain their own narrower jurisdictions.
01 · LEAD FLAGSHIP · ERI / RFP ASSURANCE
Auctrel / ERI
Most RFP systems help retrieve or draft answers. Auctrel asks the harder question first: what is the enterprise actually safe and authorized to represent? Built initially for RFPs, RFIs, DDQs, security questionnaires, and pre-contract diligence, ERI reconstructs institutional state before language becomes a customer-facing commitment.
WHAT IT GOVERNS
Buyer requirement → evidence → claim → scope → permissions → policy → disposition → human authority → released response → contractual position.
WHY IT EXISTS
Deal velocity, expert capacity, and representation control. The system is designed to resolve routine evidence questions automatically while escalating the smaller set of decisions that genuinely require specialist or executive judgment.
EVIDENCE BOUNDARY
Functioning governed execution architecture under controlled assurance. Enterprise identity, connectors, tenant isolation, lifecycle propagation, and live customer shadow validation remain the next authority stage.
ADDITIONAL FLAGSHIP SYSTEMS
Three additional public demonstrations of the portfolio’s core method: evidence before confidence, bounded authority, inspectable decisions, and learning from outcomes.
PROVENANCE + ACQUISITION INTELLIGENCE
The Provenance Engine
Objects in secondary markets rarely arrive with clean facts. A photograph, signature, dealer story, or auction description may be right, partly right, or misleading. The Provenance Engine determines what the object most likely is, what evidence could disprove that conclusion, what comparable objects actually sold for, and what action makes economic sense.
Under the hood. It separates observation from attribution, generates competing identifications, tests chronology and provenance, evaluates realized comparables, models condition and transaction friction, keeps distinct market, resale, personal-use, and offer ceilings, applies behavioral vetoes, and records outcomes so later decisions can be calibrated against earlier mistakes.
Evidence. Extensive operator testing · three public acquisition cases · one collection dossier · prospective calibration pending.
AI ASSURANCE + EVIDENCE
WarrantSpan
AI can write something that sounds expert long before it has earned the right to be trusted. WarrantSpan sits between generation and consequential human reliance, checking whether important conclusions are actually supported strongly enough to justify action.
Under the hood. It freezes the evidence corpus, raises the evidentiary burden as claims become stronger, forces serious competing explanations, repairs reasoning before polishing prose, audits citation-to-claim integrity and unsupported certainty, preserves an audit trail, and includes blinded comparative evaluation with reserved holdout material so the control system is not judged only on the examples that shaped it.
Evidence. Extensive operator testing · independent benchmark pending.
FINANCIAL CONTROL
TraceSum
Small organizations need accounting discipline but rarely have the administrative infrastructure of a large finance department. TraceSum reduces the reconciliation, evidence, and exception-management burden with AI while keeping the accounting system—not the language model—as the source of truth.
Under the hood. It keeps evidence, economic events, journal entries, reconciliation, and advice distinct; preserves canonical double-entry authority; controls duplicates and write paths; treats current-source research separately from bookkeeping truth; and tests the control surface through a versioned regression harness rather than assuming fluent output is correct.
Evidence. Controlled P0 run passed 84/84 checks · independent review pending.
ACTIVE WORKING SYSTEMS
Substantial operating systems under real use or testing, with meaningful external, commercial, performance, or prospective calibration still to earn.
EXECUTIVE + PERSONAL OPERATIONS
FORTH
Life does not arrive in separate apps. A moved meeting can change the flight, dinner, errands, groceries, and the rest of the day. FORTH carries the relevant context across email, calendar, travel, research, meals, planning, and everyday logistics so the user does not have to keep rebuilding the situation.
Under the hood. It separates stable context from facts that must be checked live, carries useful context across domains without treating all personal information as equally shareable, distinguishes harmless ambiguity from consequential uncertainty, and keeps a boundary between what the system can do and what it is actually authorized to do.
Evidence. Extensive operator use · invitation-only external beta.
INFORMATION GOVERNANCE
Determinant System Wizard
A retention policy says what should happen to information. It cannot prove what actually happened. Data gets copied into systems, exports, vendors, backups, and archives while legal holds, contracts, privacy requirements, and business exceptions accumulate around it. Determinant reconstructs that reality and turns it into defensible retention and deletion decisions.
Under the hood. It inventories records, systems, and owners; classifies data and dependencies; maps legal, contractual, privacy, security, hold, policy, and operational obligations supplied by authorized teams; records exceptions and accountable decision rights; and closes the loop with evidence showing what was retained, reduced, deleted, or left unresolved—and why.
Evidence. Extensive operator testing · commercial validation pending.
ADAPTIVE TRAVEL DECISIONS
ReturnVector
An itinerary assumes the plan stays true. Travel does not. Flights fill, prices move, weather changes, transit disappears, and the traveler gets tired. ReturnVector keeps asking not only “What should I do now?” but “What good options will still exist after I do it?”
Under the hood. It models constrained inventory, disruption risk, trip utility, confidence, and traveler state separately; builds time-dependent recovery paths home; discounts backup options that secretly share the same failure point; gives waiting an expiration time; incorporates fatigue and human capacity; and freezes forecasts before outcomes so later calibration is possible.
Evidence. Extensive operator testing · prospective calibration pending.
INVESTMENT DECISION INTELLIGENCE
ThesisDelta
When a stock falls, the difficult question is not what happened to the price. It is whether anything equally serious happened to the business. ThesisDelta investigates that gap and forces research to end in a bounded decision instead of an ever-growing pile of information.
Under the hood. It combines current-source research, forensic vetoes, opportunity-specific underwriting, explicit assumption sheets, hostile scenario analysis, portfolio concentration and factor overlap, adversarial review desks, falsifiers, monitoring triggers, and a performance-learning layer designed to compare forecasts with what later happened.
Evidence. Extensive operator testing · performance calibration pending.
SPECIALIST SYSTEMS
Narrower-domain systems that test whether the same governed architecture can survive contact with creative judgment, literary voice, and career decision-making.
VISUAL PRODUCTION SYSTEMS
KRITIFORM
Making an attractive AI image is easy. Producing the right visual artifact repeatedly is much harder. KRITIFORM acts less like a prompt and more like an autonomous creative studio: it interprets the assignment, directs the work, decides how to produce it, judges the result, and decides what should happen next.
Under the hood. It specifies composition and viewer response before generation, anticipates failure sites, routes between generation, editing, compositing, and regeneration, inspects the visible artifact rather than defending the prompt, preserves intentional difficulty when it carries meaning, converts criticism into binding revisions, and records artifact-level learning so production knowledge can compound across changing image models.
Evidence. Extensive operator testing · formal benchmark pending.
LITERARY INTELLIGENCE
Quireline
AI can write beautifully and still make every writer sound strangely alike. Quireline treats generation as only one stage of literary production. It decides what kind of piece the material requires, develops competing approaches, rejects weaker structures, revises the winner, and protects the writer from disappearing into the system’s own preferred voice.
Under the hood. It routes among literary modes, generates structurally different candidates, judges formal necessity and imitation risk, tracks syntax, diction, rhythm, and characteristic irregularities, distinguishes surgical editing from structural rewriting, scans for recurring AI “house style,” and uses comparison and failure classification to judge whether revisions actually improved the work.
Evidence. Extensive operator testing · external calibration pending.
CAREER DECISION INTELLIGENCE
Parallax
Most career software helps people apply for jobs. The harder question is whether a job—or any professional move—is worth pursuing in the first place. Parallax evaluates employment, internal moves, founder paths, and waiting as competing uses of a person’s professional capital.
Under the hood. It maintains a canonical career state, separates aspiration from provable evidence, weighs mandate, compensation, trajectory, manager risk, and reversibility, protects sensitive proof through disclosure controls, treats optionality as an asset, forces explicit action states instead of endless consideration, and records predictions and outcomes so future decisions can be calibrated against what actually happened.
Evidence. Extensive operator testing · outcome calibration pending.
Hierarchy is editorial, not permanent. A system moves upward only when stronger public evidence, independent review, real-world adoption, calibrated outcomes, or consequential external use justifies the change.
RECURRING METHOD
Understand. Build. Exercise. Revise.
Across the portfolio, the sequence is consistent: find the consequential decision beneath the surface request; build an inspectable artifact around the real constraints; put it into use; preserve failures, dissent, and outcomes; then change the system when reality disagrees with it.
RECURRING ARCHITECTURE
Authoritative state · source hierarchy · explicit uncertainty · decision rules · adversarial challenge · bounded authority · auditability · outcome recording.
Those controls recur because fluent output is cheap. A system becomes more interesting when someone can inspect what it believed, why it acted, what it was allowed to do, where it was uncertain, and whether later evidence changed it.
01 · UNDERSTAND
Find the governing problem.
Separate the visible request from the burden, failure, or decision that actually matters. State what would have to be true for the proposed intervention to deserve confidence.
02 · BUILD
Make the reasoning inspectable.
Turn the thesis into a workflow, control system, decision engine, schema, interface, or operating model with explicit state, uncertainty, decision rights, vetoes, and reconstruction.
03 · EXERCISE
Let reality change it.
Use the artifact, record what happens, preserve negative evidence, and revise from outcomes rather than attachment. A working system is evidence of existence; it is not yet evidence that the system is right.
WORK WITH ME
Bring me a consequential problem that has outgrown its current operating logic.
If the evidence is fragmented, decision rights are unclear, or a useful tool now needs to become a durable system, I am interested in the structure of the problem before the pitch. The strongest starting point is a real workflow, a real constraint, and enough evidence to discover whether the apparent problem is the actual one.
Please do not send confidential, privileged, regulated, customer-identifiable, or proprietary material in an initial inquiry.