# Virion.ai > An intelligence to intelligence exploration company. This file is the single source of truth for the Virion.ai site. Edit the llms.txt, the site updates. No rebuild required. Proof that complicated websites can have order — and that the order is legible to both humans and machines. Format rules (for humans and machines): - `## H2` opens a top-level section (Brand, Horizon, Thesis, Products, Projects, Links). - `### H3` opens an entity (a product, a project, a tenet). - `#### H4` and `##### H5` open sub-sections of an entity. - `key: value` lines within a section become fields. - `- item` lines become list items for the preceding `key:` that ends with a colon on its own line. - Blank lines separate paragraphs within a free-text field. - `ctas:` is a list under a product or project. Each item is pipe-separated: `label | href | style`. `href` is a URL or `contact:` (opens the contact overlay preselected to that intent: engagement, research, introduction, press). `style` is `primary`, `ghost`, or `link` (default ghost). --- ## Brand tag: Infect Intelligence nav: - Products - Experiments - Thesis - Initiate Contact --- ## Horizon eyebrow: virion.ai · an intelligence to intelligence exploration company primary: Most teams are building tomorrow's app. We're building for the stage after. secondary: SaaS was the container for the last era of work. The next one belongs to intelligence that travels — between people, models, and systems — with its context, provenance, and handoff intact. We build the instruments that make that transition legible while you are still operating in today. cta-primary: See the systems cta-secondary: Initiate contact stages-label: Where are you? stages: - Today: AI features bolted onto SaaS. - Next: workflows where the artifact carries its own context. - After: intelligence that hands off between systems without losing itself. - The work: instruments that make each stage observable at atomic resolution. --- ## Thesis title: Thesis tagline: Four commitments that decide what we build, and what we won't. return-trigger: Only by observing the future at atomic resolution can we responsibly zoom out. expanded: /thesis.md expanded-label: Read the full thesis — DAGs, workflows, and system context for technical readers. ### Atomic resolution We build instruments before we build markets. The work starts at the specification — the layer where humans, AI systems, and regulators can read the same source of truth. Markets follow once the instrument is honest. ### Intelligence is temporal In regulated work, an answer is not a string — it is a claim with a timestamp, a source, and an evidence chain. Auditors ask when you knew it and how. Our systems answer in those terms by default, because retrofitting traceability is how trust collapses. ### Composable, not consolidated We ship small, sharp systems that compose: a substrate, a protocol, a router, a gate. None of them are the platform. The platform is the discipline of keeping them separable, so a model swap, a vendor change, or a regulatory shift does not require a rebuild. ### Return orbit Depth and overview must be reversible. Every zoom into atomic detail is paired with a path back to the operational view a CEO, an auditor, or an operator can act on. Memory has a shape, and the shape has to survive the round trip. --- ## Products ### Evidence Systems domain: Evidence Intelligence color: #00b894 sigil: E vertex: 0 thesis: Intelligence must be traceable, temporal, and composable. ctas: - Vault (Evolving Infrastructure) | #vault | primary - Capsule (Portable Intelligence) | #capsule | ghost #### Business Case problem: Most AI systems can retrieve text. Few can answer where a claim came from, when it was true, or what evidence should override it. In regulated and document-heavy work, that gap is where trust collapses. solution: We design evidence substrates — not chatbots, not wrappers. Source objects, temporal facts, provenance, and governed retrieval as architecture, so AI outputs can be audited as easily as they are produced. market: Regulated operations, due-diligence workflows, internal knowledge platforms, and intelligence products built on partitioned or sensitive data. edge: Lineage, partitioning, and time are first-class concerns in the substrate — not retrofitted after a model picks an answer. Teams inherit a foundation where every claim resolves to its source, its timestamp, and its governance rules by default. metrics: - Form: Substrate, not chatbot - Anchors: Vault, Capsule - Posture: Provenance as architecture - Best fit: Evidence-heavy AI systems #### System View inputs: Source documents, Temporal facts, Claims, Governance rules, Partitioned data transforms: Extract evidence, Preserve provenance, Resolve timestamps, Apply access rules, Govern retrieval outputs: Auditable claims, Evidence substrate, Traceable AI responses, Source-linked decisions --- ### Portable Work Products domain: Portable Work Products color: #7c5cff sigil: P vertex: 1 thesis: The work product should carry its own context, provenance, and handoff. ctas: - Capsule (Portable Intelligence) | #capsule | primary - Visit capsules.virion.ai | https://capsules.virion.ai | ghost #### Business Case problem: AI work products usually arrive stripped — a chat transcript, a PDF, a JSON blob. The reasoning, context, and handoff path stay locked inside the model that produced them, and the next intelligence in the chain has to start over. We built our way out of that — first with Vault, the substrate where intelligence rests, then with Capsule, the format intelligence travels in. That's why Capsule exists: we got tired of strapping different solutions together to ship the same kind of work twice. solution: A capsule packages mission, reasoning history, state, artifacts, provenance, and the next-step handoff in a single signed, hash-chained file — so the work can travel between people, systems, and models without losing itself. We authored the protocol and open-sourced it. Intelligence in motion shouldn't be gated; the field evolves through open cooperation in technological innovation, not through proprietary lock-in on the carrier. market: Multi-party AI workflows, field and offline operations, regulated handoffs, and teams treating the artifact itself as the surface of value. edge: The capsule is the product, not the chat that produced it. That makes it durable across model swaps, ownership changes, and the long tail of intelligence-to-intelligence handoff. Open by design — so the protocol can outlive any single vendor, including us. metrics: - Form: Portable file + protocol - Status: Open protocol, v0.6 with v1 roadmap - Companion: Vault (intelligence at rest) - Best fit: Provenance-rich, travel-ready outputs #### System View inputs: Mission context, Reasoning history, State, Artifacts, Provenance, Handoff target transforms: Package context, Sign the record, Chain hashes, Preserve state, Prepare handoff outputs: Portable capsule, Verifiable work product, Replayable reasoning, Cross-model handoff --- ### Agentic Harnesses domain: Model Harnessing color: #ef4444 sigil: M vertex: 2 thesis: The model is the engine. The harness is the work. ctas: - Rolodex (Routing Substrate) | #rolodex | primary - EveryAI.Chat (Rooms as Harness) | #everyai-chat | ghost #### Business Case problem: The industry is still asking which model is best. That question expired. Every frontier model has a shape — strengths, failure modes, reasoning patterns, latency profiles — and the work gets done or doesn't depending on the harness wrapped around it. The same model, under different harnesses, produces fundamentally different intelligence. Most teams haven't started building harnesses yet; they're still swapping models and hoping the next one solves it. solution: We started where most teams are — building a routing layer (Rolodex) for failover, attribution, and provider neutrality. We kept it. But the real work is upstream of routing: custom agentic harnesses that change what a model can actually express. Multi-agent loops, evidence-grounded reasoning, tool-use with rollback, rooms instead of threads — each harness shape unlocks a different kind of intelligence from the same underlying model. We treat harness design as an experimental discipline, not a feature checkbox. market: AI platform teams, applied research groups, and product teams who have outgrown "pick a model" and need a substrate for harness experimentation across providers. edge: Routing is solved infrastructure. Harnesses are open territory. We hold an honest position — no team has the universal harness yet, including us — and we ship the substrate that makes the experimentation legible. Cost attribution, failover, and provider neutrality come from Rolodex underneath, so harness experiments cost time, not architecture. metrics: - Form: Harness experimentation, routing substrate - Anchor: Rolodex (provider-neutral routing) - Posture: No universal harness — yet - Best fit: Teams done with model-shopping, ready to build the wrapper that actually does the work #### System View inputs: Model capabilities, Tool surfaces, Task context, Provider signals, Evaluation feedback transforms: Route requests, Shape prompts, Coordinate agents, Ground evidence, Roll back tool use outputs: Task-specific harness, Attributed model work, Provider-neutral execution, Observable agent behavior --- ### Supply Chain Decisioning domain: Supply Chain Gate color: #64748b sigil: S vertex: 3 thesis: Trust should be computed before installation. ctas: - Virion's Sherlock | #sherlock | primary - Discuss the alpha | contact:research | ghost #### Business Case problem: Most supply-chain tooling activates after code is already running in your environment. By that point the compromise path has already crossed the threshold the tool was meant to defend. The industry has spent a decade building runtime detection for problems that should have been pre-install decisions. solution: We move the trust decision earlier. Package metadata, dependency signals, and ecosystem behavior get scored at the gate — before install, before CI, before a developer's machine inherits the risk. Sherlock is the gate. A broader firewall architecture (Aidenied) is the surface it eventually composes into — but the gate ships first, by design. market: Security-conscious engineering organizations, regulated software teams, and developer platforms that need lightweight pre-install controls before the next dependency postmortem gets written. edge: The gate keeps its own dependency tree at zero. A trust tool should not recreate the problem it exists to solve — and that restraint is the design statement. metrics: - Form: Pre-install risk gate - Status: Alpha — stage 3 of 6, beta opens at stage 6 - Runtime: Standard library only - Posture: Zero-dependency by principle - Best fit: Lightweight supply-chain controls #### System View inputs: Package metadata, Dependency manifests, Ecosystem behavior, Maintainer signals, Install context transforms: Score trust, Detect risk patterns, Compare ecosystem signals, Gate install decisions outputs: Pre-install verdict, Risk explanation, Dependency gate, Lightweight supply-chain control --- ### Human Workspaces domain: Comprehension Layer color: #14b8a6 sigil: A vertex: 4 thesis: Agents produce. Humans decide. The workspace is for the human. ctas: - Let's design your workspace | contact:engagement | primary - EveryAI.Chat (Comprehension Surface) | #everyai-chat | ghost #### Business Case problem: AI now produces a thousand times more work than any human can read in a day. The bottleneck moved — from generation to comprehension — and most interfaces haven't noticed. They keep adding agents, threads, and outputs to the same single-stream chat surface, and the human at the end of it falls further behind every week. solution: We build the workspace for the human, not for the agent. Rooms instead of threads, persistent memory, routed models underneath, evidence-grounded outputs from Vault, portable handoffs through Capsule — all of it converging on a single question: can the human interpret what the system produced, and decide. That's what the workspace exists for. The agents are upstream production. The room is where production becomes legible enough to act on. market: Operators, analysts, decision-makers, and teams drowning in AI output that no human pipeline can absorb. Internal AI platforms where the gap between "we deployed agents" and "we made better decisions" has become visible. edge: The industry built agent-centric interfaces. We built a human-centric one. The workspace stands on routed harnesses, evidence substrates, and portable artifacts so the human at the end isn't asked to do what the substrate should have done — they're asked to do what only a human can do: decide. metrics: - Form: Comprehension layer for AI output - Composes: chatsdk, Rolodex, Vault, Capsule - Posture: Human-centric, not agent-centric - Best fit: Teams whose AI output volume has outpaced their decision capacity #### System View inputs: Agent outputs, Room context, Persistent memory, Evidence records, Portable handoffs transforms: Organize rooms, Preserve context, Route model work, Ground outputs, Surface decisions outputs: Human-readable workspace, Decision surface, Comprehension layer, Coordinated AI workstream --- ### Evaluation Loops domain: Fitness Engine color: #f97316 sigil: L vertex: 5 thesis: You cannot improve what you have not committed to measuring. ctas: - Evolve (Fitness Engine) | #evolve | primary - Commission an eval loop | contact:engagement | ghost #### Business Case problem: Most AI programs improve invisibly. Someone tunes a prompt on a Tuesday, the output gets better, no one writes down why. Three months later the person leaves and nobody can reconstruct which decisions held and which were luck. The tuning cycle stays informal — and informal is invisible. The deeper failure is upstream: teams build evals against proxies before they have committed to what they are actually trying to measure, and end up confidently improving the wrong thing. solution: We work two layers. Upstream, goal articulation as the load-bearing artifact — what is this code for, and what would "closer to that goal" actually look like as a number you can defend. Downstream, the loop that ratchets toward it: score, diagnose, propose, execute, decide, all inside an auditable ledger. Improvement becomes a property of the system rather than the property of whoever happened to be tuning prompts that week. market: Agent teams, AI platform groups, and operational AI programs that need evaluation as system behavior — not as a notebook one engineer maintains. edge: We make goal articulation part of the engagement, not an assumption. The loop is downstream of a target you can defend. Same architectural family as Karpathy's autoresearch — anything you can score, you can ratchet toward — applied beyond ML training into operational AI workflows. The fitness ledger is itself a capsule, so improvement decisions travel with the system that made them. metrics: - Form: Auditable improvement loop - Anchor: Evolve (autonomous fitness engine) - Lineage: Karpathy's autoresearch, applied beyond ML training - Modes: Rule-based, model-proposed, human-in-the-loop - Best fit: Teams ready to commit to what good actually looks like #### System View inputs: Goal definition, Scoring criteria, Run history, Output samples, Human decisions transforms: Score outputs, Diagnose failures, Propose changes, Execute trials, Record decisions outputs: Fitness ledger, Auditable eval loop, Accepted improvements, Goal-aligned system behavior --- ## Projects These are the active code-backed nodes in the Virion workspace — anchor implementations, infrastructure layers, applied arbitrage, and portfolio experiments. Each one is a real surface that makes a piece of the thesis observable at atomic resolution. A note on what these projects are, and what they are not. Software is no longer the defining quality of a company. AI's emergent capabilities are making code cheap, replaceable, and increasingly authored by the systems themselves. What does not get replaced — what is not cheap — is the vision and guidance to direct that capability into the right work. Bespoke intelligence was a specialty. Now it is a commodity. That is the arbitrage. The portfolio is evidence of that direction. The anchors are bets on the architectural shapes intelligence-to-intelligence work will need. The applied arbitrage projects — ReadySet for SMB lending readiness, Fix.Now for home improvement, ComplianceQ for RegTech, Adryos for the asset world — run the same engine across different domains: workflow, intelligence processing, and downstream matching of goods and services. They are the markets the commoditization just opened up. The infrastructure pieces compose across all of it. The experiments are early signals. Standing still with proprietary code is the losing position. Building the engine that turns specialty intelligence into commoditized intelligence — that is the position we are building. --- ### Vault status: Anchor implementation for Evidence Systems domain: Evidence Substrate sigil: V ctas: - Talk to an architect | contact:research | primary #### Why it exists We got tired of strapping together a different evidence pipeline for every use case. Most "AI knowledge platforms" are a vector store with a chatbot bolted on, and they fail the moment a regulator, an auditor, or a careful operator asks where a claim came from, when it was true, or which source should override another. Vault is the substrate we built so we would never have to answer those questions retroactively again. Postgres-rooted, multi-RAG by design — vector, knowledge graph, document, fact-schema, and extraction layers operating in coordination — and agentic adaptable, so the routing between layers is a runtime decision, not a hardcoded pipeline. #### What it does Vault treats evidence as architecture. Source objects, temporal facts, provenance chains, partitioning rules, and governed retrieval are first-class. A claim returned from Vault carries its lineage by default — the source, the timestamp, the partition it came from, and the governance rules that applied to its retrieval. #### Decision factors - Use Vault when retrieval needs to survive an audit, not just answer a question. - Use Vault when the same source corpus serves multiple agents, products, or partitions. - Skip Vault if a single-tenant vector store and a chat UI are sufficient — and be honest about whether they actually are. #### Engagement fit Best as the substrate under an applied engagement, not as a self-serve product. Vault is commercial — the substrate stays accountable. --- ### Capsule status: Anchor implementation for Portable Work Products · Open protocol, v0.6 with v1 roadmap domain: Portable Intelligence sigil: C ctas: - Visit capsules.virion.ai | https://capsules.virion.ai | primary - Read the protocol spec | https://github.com/virion/capsules | ghost #### Why it exists AI work products usually arrive stripped of the context that made them useful. A chat transcript loses the reasoning. A PDF loses the source chain. A JSON blob loses the handoff path. The next intelligence in the chain — human or agent — has to start over. Capsule is the format we authored so the work could travel without losing itself. A signed, hash-chained file that packages mission, reasoning history, state, artifacts, provenance, and the next-step handoff. It is the transport layer for intelligence-to-intelligence work. #### What it does A capsule is a portable, verifiable record of a unit of intelligence work. It can be opened, audited, replayed, and handed off — across people, models, vendors, and time. It is the protocol underneath the work, not a wrapper around it. #### Why it is open We open-sourced the protocol because intelligence in motion should not be gated. The field evolves through open cooperation in technological innovation, not through proprietary lock-in on the carrier. We retain commercial substrates (Vault) and applied implementations (the AML investigation capsule suite, eight coordinated compliance templates implementing FATF Rec. 10/24, FinCEN CDD, and OFAC screening workflows). The protocol itself belongs to the field. #### Decision factors - Use Capsule when AI work needs to outlive the session that produced it. - Use Capsule when handoffs cross team, vendor, or model boundaries. - Use Capsule when audit trails need to travel with the artifact, not in a separate system. #### Engagement fit The protocol is open and self-serve. Implementations, capsule suite design, and integration into existing workflows are commercial engagement work. --- ### Rolodex status: Anchor implementation for Agentic Harnesses domain: Routing Substrate sigil: R ctas: - Talk to an architect | contact:research | primary #### Why it exists We started by building a model routing solution because the problem felt obvious: every team accumulating model adapters one provider at a time, with no shared cost view, no failover policy, and no stable abstraction when a provider deprecated or went down. Rolodex solves that — and we kept it. But routing alone wasn't enough. The real work happens upstream of the router, in the harness wrapped around the model. The same model under different harnesses produces fundamentally different intelligence. Rolodex became the substrate underneath that experimentation rather than the product itself. #### What it does Rolodex is the control plane for model access. Failover, streaming normalization, billing attribution, model discovery, and routing policy live in one layer. Applications stop carrying provider risk on their backs. Per-project token-to-value accounting becomes possible because attribution is a substrate concern, not an application concern. #### Decision factors - Use Rolodex when multiple agents, products, or surfaces share model access. - Use Rolodex when cost attribution needs to be per-project, not per-key. - Use Rolodex when the harness work is ahead of the routing — and the routing has to stop being a distraction. #### Engagement fit Rolodex is commercial substrate. Most engagements pair it with harness design work, where the real differentiation lives. --- ### EveryAI.Chat status: Anchor implementation for Human Workspaces · Open-trajectory as it matures domain: Comprehension Surface sigil: E ctas: - Let's design your workspace | contact:engagement | primary #### Why it exists AI now produces a thousand times more work than any human can read in a day. Most interfaces haven't noticed. They keep adding agents, threads, and outputs to the same single-stream chat surface, and the human at the end falls further behind every week. EveryAI.Chat is the workspace we built for the human, not for the agent. Rooms instead of threads. Persistent memory. Routed models underneath. Evidence-grounded outputs from Vault. Portable handoffs through Capsule. All of it converging on one question: can the human interpret what the system produced, and decide. #### What it does EveryAI.Chat is the comprehension layer of the Virion stack. The agents are upstream production. The room is where production becomes legible enough to act on. The same artifact serves two roles by composition — a harness pattern at the orchestration tier, a comprehension surface at the human tier — which is what the composability tenet looks like in practice. #### Why it will go open The interface and runtime layers belong with the field, on the same principle as Capsule: carriers should not be gated. EveryAI.Chat will follow that trajectory as it matures. The substrates underneath (Vault, Rolodex) stay commercial because we owe customers an accountable maintenance posture on the operational systems. #### Decision factors - Use EveryAI.Chat when AI output volume has outpaced human decision capacity. - Use EveryAI.Chat when multi-party, multi-agent work needs a surface that isn't a chat thread. - Use EveryAI.Chat when the workspace itself needs to be the unit of value, not the model behind it. #### Engagement fit Co-designed per customer. The shape of the comprehension layer depends on the work being comprehended. This is consultative architecture, not a SaaS deployment. --- ### Sherlock status: Alpha — stage 3 of 6, beta opens at stage 6 domain: Supply Chain Gate sigil: S ctas: - Discuss the alpha | contact:research | primary #### Why it exists Most supply-chain tooling activates after code is already running. By that point the compromise has crossed the threshold the tool was meant to defend. Sherlock moves the trust decision earlier — package metadata, dependency signals, and ecosystem behavior get scored at the gate, before install, before CI. #### What it does A pre-install risk gate with a zero-dependency runtime. Standard library only. The tool does not recreate the problem it exists to solve. #### Roadmap A 6-stage plan, currently testing stage 3. Beta opens at stage 6. The broader firewall architecture (Aidenied) is the surface Sherlock eventually composes into — but the gate ships first, by design. #### Engagement fit Not yet pipelined. Conversation-only at this stage — security teams interested in the architecture or the trajectory are welcome to reach out. --- ### Evolve status: Anchor implementation for Evaluation Loops domain: Fitness Engine sigil: V ctas: - Commission an eval loop | contact:engagement | primary #### Why it exists Most teams know they need evaluation but have no durable loop for scoring outputs, proposing changes, and recording why one improvement was accepted over another. The tuning cycle stays informal — and informal is invisible. Worse, teams skip the upstream discipline: they build evals against proxies before they have committed to what they are actually trying to measure. #### What it does Evolve treats evaluation as a system, not a notebook. Goal articulation upstream; ratchet loop downstream. Score, diagnose, propose, execute, decide — all inside an auditable ledger. The fitness ledger is itself a capsule, so improvement decisions travel with the system that made them. #### Lineage Same architectural family as Karpathy's autoresearch — anything you can score, you can ratchet toward. Applied beyond ML training into operational AI workflows, agent systems, and regulated decision pipelines. #### Engagement fit Commissioned per customer. Goal articulation is part of the engagement, not an assumption. --- ### Adryos status: Alpha — Artpraisal module accepted by client · Two additional modules in development domain: Applied Arbitrage · Asset World Ecosystem sigil: A ctas: - See the ecosystem | contact:introduction | primary #### Why it exists The asset world — appraisal, attribution, provenance, transactional documentation — runs on specialist judgment that doesn't scale linearly. Specialists do not need to be replaced. They need their busywork reduced so they can spend their time on the work only they can do: deciding value, defending judgment, delivering finished work to clients faster. #### What it does Adryos is being built as an ecosystem for the asset world. Each module is a focused tool that compresses busywork around a specific specialist workflow, leaving the judgment with the specialist. - **Artpraisal** — the first module, accepted by client and operating in alpha. An appraiser-facing platform that performs contextual expansion around artwork valuation: comparable sales surfacing, attribution research, condition documentation, structured deliverables. It does not automate the valuation. The appraiser remains the authority; Artpraisal removes the friction between their judgment and the finished product. - **MCP / Mobile App** — in development. - **Third module** — in development. #### Decision factors - Built for specialty markets where judgment is the value and busywork is the bottleneck. - Domain partnership model — Virion handles full digital and application development; the operating partner brings domain authority and customer relationships. #### Engagement fit Ecosystem partnership. Conversations welcome with operators in adjacent asset-world domains (collectibles, real assets, structured provenance) where the same pattern would apply. --- ### ReadySet status: In active development · readyset.now domain: Applied Arbitrage · SMB Lending Readiness sigil: R ctas: - Visit readyset.now | https://readyset.now | primary - Partner conversation | contact:engagement | ghost #### Why it exists SMB lending readiness is structurally inefficient on three sides at once. Banks and service providers spend significantly to acquire borrowers who turn out not to be ready. Service providers carry CAC they cannot recover when applications stall. SMBs lose weeks or months in the gap between "I need capital" and "I am ready to close." Bespoke advisory used to be the only fix. Intelligence-based readiness is now buildable as a system. #### What it does A funding-readiness platform that compresses the path from intent to close. Workflow (intake, document gathering, eligibility assessment) → intelligence processing (matching applicant profile to lender criteria, identifying gaps, surfacing remediation paths) → downstream matching (connecting ready borrowers to appropriate funding products and lenders). #### Designed for - **SMBs** — reduce time to close by entering the funding conversation already ready. - **Banks** — lower CAC by receiving borrowers who are pre-qualified against your criteria. - **Service providers** — lower CAC and increase close rates on the borrowers you already work with. #### Engagement fit Partnership-driven. Best paired with operators in the lending or advisory ecosystem who bring domain authority and distribution. --- ### Fix.Now status: Approaching launch · Real-world applied AI solution domain: Applied Arbitrage · Home Improvement sigil: F ctas: - Partner conversation | contact:engagement | primary #### Why it exists Home improvement is the textbook information-asymmetry market. Homeowners don't know what's wrong, contractors don't know what they're walking into, estimates vary by 4x for the same job, and the supply side is fragmented across hundreds of thousands of small operators. Bespoke diagnostic intelligence used to be priced out of reach — now it isn't. #### What it does The Virion arbitrage engine applied to home improvement. Workflow (problem intake, scoping, scope-of-work generation) → intelligence processing (diagnosing the actual job, surfacing the right specifications and likely causes) → downstream matching (connecting homeowners to qualified contractors, parts, and services). #### Decision factors - For homeowners: a way to enter a contractor conversation already informed. - For contractors and service partners: a way to enter a job already scoped. - The arbitrage is in collapsing the information gap between both sides before the work begins. #### Engagement fit Partnership and operator conversations welcome. The engine is portable; the distribution channel is the variable. --- ### ComplianceQ status: Approaching launch · Real-world applied AI solution domain: Applied Arbitrage · RegTech sigil: Q ctas: - Partner conversation | contact:engagement | primary #### Why it exists Regulated workflows — BSA/AML investigations, KYC reviews, SAR drafting, sanctions screening — have been specialty work performed by trained analysts at significant cost per case. The intelligence to do this well consistently is now buildable as a system. ComplianceQ is the RegTech expression of that arbitrage. #### What it does An applied AI solution for regulated investigation and review workflows. Built on the AML investigation capsule suite — eight coordinated compliance templates implementing FATF Rec. 10/24, FinCEN CDD, and OFAC screening — with Vault as the evidence substrate underneath. #### Decision factors - For compliance teams who need consistency, auditability, and analyst leverage simultaneously. - For institutions where the cost of a bad SAR is regulatory, not just operational. - Capsule-based deliverables mean every investigation travels with its full reasoning chain by default. #### Engagement fit Partnership and pilot conversations open. Regulated buyers, by definition, do not move fast — but they buy substrate, not features. --- ### PromptScript status: Portfolio experiment domain: Markdown-Native Forms sigil: P ctas: - Discuss the experiment | contact:research | ghost #### Why it exists PDF forms are how legacy systems extract structured data from humans. They are also a terrible interface for LLMs, which already speak markdown fluently and want to operate at the specification layer. PromptScript flips the model: define the form in markdown, let the LLM fill it, let the LLM reply and submit from the same markdown surface. #### What it does A dynamic markdown form concept. Forms are authored in plain markdown with structured fields. An LLM can read the form, fill it out, return responses, and submit — all without ever touching a PDF or a traditional form-rendering layer. The markdown is both the spec and the surface. #### Decision factors - Useful when forms are the bottleneck between LLMs and legacy data capture. - Useful when the form definition needs to be human-readable, version-controllable, and machine-fillable simultaneously. - Sits adjacent to Capsule (capsules can carry the filled form as a portable artifact) and the harness work (a harness can drive the fill). #### Engagement fit Research-stage. Conversations welcome with teams thinking about LLM-native form replacements at scale. --- ### chatsdk status: Frontend library · Open-trajectory as it matures domain: AI Router/Harness Frontend sigil: K ctas: - Talk to an architect | contact:research | primary #### Why it exists Once you have an AI router or a custom harness running, you need a frontend to deploy it against. Most teams either build that frontend from scratch every time or accept whatever opinionated UI ships with their model provider. chatsdk is the frontend library we built so we'd stop solving the same UI problem on every engagement. #### What it does A frontend library for deploying your own AI router or harness. Plug in the routing layer (Rolodex or your own), plug in the harness, get a working interface. Workspace-aware, room-aware, capsule-aware where it makes sense. #### Why it will go open Same trajectory as EveryAI.Chat and on the same principle as Capsule: the carrier layer between intelligence and humans should not be gated. Substrates stay commercial; frontends do not need to be. Open release timed to API stabilization. #### Engagement fit Currently embedded in Virion implementations. Open-source release planned as the API surface stabilizes. --- ## Links evidence-systems → vault: anchor-system evidence-systems → capsule: portable-companion portable-work-products → capsule: anchor-system portable-work-products → vault: substrate agentic-harnesses → rolodex: routing-substrate agentic-harnesses → everyai-chat: harness-implementation supply-chain-decisioning → sherlock: anchor-system human-workspaces → everyai-chat: anchor-system human-workspaces → chatsdk: frontend-library human-workspaces → rolodex: routing-substrate human-workspaces → vault: evidence-substrate human-workspaces → capsule: portable-handoff evaluation-loops → evolve: anchor-system evaluation-loops → capsule: fitness-ledger-format applied-arbitrage → readyset: smb-lending applied-arbitrage → fix-now: home-improvement applied-arbitrage → compliance-q: regtech applied-arbitrage → adryos: asset-world-ecosystem applied-arbitrage → vault: evidence-substrate applied-arbitrage → capsule: portable-deliverables --- ## Initiate Contact intents: - engagement: Commission an implementation, scope a build, or open a partnership conversation. The right intent for buyers, operators, and teams ready to engage Virion's commercial work. - research: Talk to an architect. The right intent for technical conversations, alpha-stage products, peer-level discussion, and harness or substrate design questions. - introduction: General introduction or ecosystem conversation. The right intent for adjacent operators, potential partners, or readers who want to understand the work without a specific project in mind. - press: Media, podcast, or editorial inquiries.