XTC

The problem

Employers, recruiters, lenders, schools, strangers, scammers, and AI models all build a model of you from your digital exhaust — your search results, your old posts, your metadata, your linked accounts. They act on that inferred profile before any human reviews the logic. You don't get to see the model. You don't get to contest it. The asymmetry is total: institutions see the profile, you see nothing.

What it is

Optic is a private algorithmic identity audit. It shows you, in plain terms, how the internet and its institutions infer you — what they think they know, why, how confident they are, and what you can do about it. It is a personal instrument for reading the model being built around you.

How it works

A single inference pipeline. Optic ingests your public surface, extracts signals, derives the claims an outside system would draw, attaches a confidence level and the strongest counterargument to each, and routes every claim to a concrete repair action.

Subprojects

• IrisThe assistant and interface layer. The voice inside Optic that walks you through the audit, the evidence, and the fixes.
• NeuontologyThe inference engine. An ontology for personal digital identity that maps sources to signals to inferred claims.
• Inference CardsThe output format. Each card states one inferred claim about you, the evidence behind it, a confidence level, and the best counter-explanation.
• Repair ActionsThe practical layer. For every inference: clean, clarify, contest, bury, explain, or redirect.
• Audience SimulatorsRecruiter, investor, lender, admissions, stranger, and scammer lenses, each showing what that specific audience infers fast.
• Signal RepairThe discipline of fixing your perceived identity signal, not your "self."

The problem

The web verifies humans with puzzles — pick the traffic lights, twist the slider. Modern AI agents now solve those puzzles better than people do. The checkbox model is broken, and the replacement cannot be more surveillance. Something has to verify a real human without watching everything they do.

What it is

Gurm is human-signal verification built on play. Instead of a puzzle, the visitor plays a tiny game-world challenge — trivial and pleasant for a person, expensive and clumsy for an agent. Verification becomes a few seconds of interaction texture, not an interrogation. Gurm is also a whole aesthetic universe: characters, an alien-knight defender, an early-2010s mobile-game look. The charm is deliberate — it is the moat.

How it works

Gurm reads interaction signatures — timing, rhythm, hesitation, embodied friction — rather than puzzle answers. Humans have temporal and motor textures that bots fake badly. The challenge is a game; the verification is the behavioral fingerprint left while playing it.

Subprojects

• GurmCaptchaThe developer-facing product: tiny game challenges (Fruit-Ninja-, Cut-the-Rope-style) replacing checkbox slop.
• Gurm VerifyThe cleaner B2B verification layer.
• Proof of HumanThe protocol: an anonymous, privacy-preserving human-proof token other products can consume.
• TempoLockThe rhythm and timing verification mechanic; humans carry temporal signatures bots can't reproduce.
• Agent GurmThe operational intelligence inside the system: mascot, security agent, and admin helper in one.
• The Gurm universeThe character system, the alien-knight defender, the game-world lore that makes verification something people don't mind doing.

The problem

Biotechnology can already edit organisms, but the field productizes slowly and dangerously because editing alone is not a product. Without governed certification, welfare, reproduction, and monitoring, a "programmable organism" is a liability, not a venture.

What it is

ABx — Applied Biosciences — is a programmable-biology company that treats biology like governed infrastructure. The edit is one step in an auditable pipeline. Companion animals are the consumer wedge; agriculture, welfare, and environmental biology are the platform. Welfare is a design constraint, not an afterthought.

How it works

A single operating chain. Every animal is genotype-certified, health-warrantied, veterinary-mediated at delivery, and tracked with longitudinal phenotype data. Containment, reproductive control, and audit-ready records are built in from the first organism — the company is designed to be regulator-readable.

Subprojects

• MinoThe consumer wedge: custom, trait-engineered companion animals, genotype-certified and health-warrantied.
• NewFruitThe food spinoff: engineered consumer produce with novel fruit forms and flavors.
• PathxThe immunology direction: custom anti-virus and biological intervention.
• AGEngx / AgraThe agricultural engineering wing: crop and livestock diagnostics, hidden-yield-failure detection, genome intervention.
• Lux ProteinA food and protein biotech concept under the bio foundry.

The problem

The information a person needs to understand the world — geopolitics, markets, AI progress, conflict, energy, supply chains — arrives as an infinite, undifferentiated feed. A feed is not a model. It has no structure, no weighting, no situational awareness. You cannot steer with it.

What it is

SignalAtlas is a global situation terminal: a single instrument panel that compresses the state of the world into something one human brain can navigate. Not a feed — a cockpit.

How it works

Domain-by-domain signal dashboards feed a weak-signal radar that surfaces what is moving before it is obvious. Sources are evidence-weighted, not engagement-weighted. The output is a personal situation room — the world as a navigable model rather than a scroll.

Subprojects

• World Model_01The world-model dashboard and simulation core.
• Metacrisis SimA civilization-risk simulation: feedback loops, collapse early-warning, scenario modeling.
• Domain dashboardsDedicated panels for geopolitics, markets, AI developments, energy, conflict, weather, and supply chains.
• Oracle LayerThe synthesis and interpretation layer that turns raw signal into judgment.

The problem

The first wave of AI in education optimized for the wrong thing — getting the answer. Cheap homework bots do the work for the student, which produces a grade and erases the learning. Real tutoring builds the student's mind; it does not bypass it.

What it is

ExamCove is premium AI tutoring built on reasoning models, generated tutor video, and adaptive learning. It is engineered to teach — to find what a student misunderstands and rebuild it — not to complete assignments.

How it works

ExamCove maintains a per-student knowledge graph and a weakness map, detects specific misconceptions, and compiles personalized lessons against a concept-prerequisite map. A dynamic problem generator and an adaptive review scheduler keep the student at the edge of their ability; an exam-risk dashboard shows where they actually stand.

Subprojects

• Reasoning-model pedagogy engineTutoring built on models that reason through a concept, not retrieve an answer.
• Generated tutor videoResearch-grade explanatory video produced on demand.
• Student Knowledge Graph & Weakness MapA live model of what the student knows and does not.
• Dynamic Problem GeneratorFresh problems calibrated to the student's current level, plus adaptive review scheduling.
• Exam Risk DashboardAn honest readout of exam readiness by topic.

The problem

Affordable food is treated as either a charity problem or a thin-margin restaurant problem. Neither framing produces dignified, reliable, low-cost nourishment at scale. The gap is a systems gap, not a goodwill gap.

What it is

Bord is a \$5 dignified meal engineered as civic infrastructure. It is explicitly not a restaurant — it is a food-access system: a controlled, throughput-optimized buffet built around cost control, real nutrition, and human dignity. First location targeted in New Britain, CT.

How it works

A controlled-buffet cost model holds the price point; throughput optimization and tight local meal logistics hold the unit economics. In-house growing is used only where it genuinely improves the math — not as decoration. The format is designed to repeat: a community food terminal, not a one-off venue.

Subprojects

• Controlled-buffet cost modelThe pricing and portion system that makes a \$5 dignified meal viable.
• Local meal logisticsThe sourcing and supply layer that holds margins.
• In-house growing where usefulSelective on-site production, used only when it improves economics.
• Community food terminalThe repeatable, civic-infrastructure format Bord is meant to scale into.