Projects + Labs
Each card shows the same pattern: what it is, what goes in, what comes out, where to inspect it, and what decision it does not make. Simple enough to scan. Specific enough to trust.
Turns transaction rows into risk flags, evidence notes, and a human-review queue.
Triage only. It does not accuse, file, freeze, approve, or make a final compliance decision.
A compact finance-intelligence dashboard for signals, planning, AML view, and market context.
Demo surface only. It does not connect to accounts, move funds, or give investment advice.
A small agent-control proof for retrieval, safety checks, role boundaries, and audit traces.
Proof environment only. It does not approve customer decisions or operate production systems.
Shows how AI-agent payments should leave a receipt trail before autonomy increases.
No wallet authority. No production payment execution. Human approval stays outside the agent.
Packages wallet-risk and fraud signals into readable case notes instead of verdicts.
Educational review surface. Not trading advice, legal advice, enforcement, or customer decisioning.
Records Kite testnet and decentralized-infrastructure work with paper-mode portfolio state.
No real-fund execution, no wallet power, no order power. This is not investment advice.
Shows lessons from small agent systems: recovery, visibility, roles, review, and limits.
Public lessons only. No production authority, live credentials, or autonomous outside action.
A board-style proof for two AI agents proposing, challenging, and scoring bounded decisions before any outside action.
Simulation and review only. No live trades, deposits, wallet custody, customer decisions, or autonomous financial action.
Turns an AI article or technical claim into a claim-source-boundary review worksheet and reusable checklist.
Writing and evidence review only. It does not certify accuracy, replace expert review, or approve regulated communications.
A map of memory, MCP, orchestration, coding-agent, and governance research themes explored through open systems.
Research map only. It does not republish third-party code, private logs, or hidden agent instructions.
Packages AI outputs as reviewable work items before they become finance, compliance, or customer decisions.
Review interface only. It does not approve, deny, escalate, file, or execute regulated actions.
Turns a public claim into a compact packet: source, claim, boundary, artifact, and reviewer action.
Evidence workflow only. It does not guarantee truth, certify compliance, or replace expert review.
Turns an AI or finance claim into a source trail, confidence note, missing-evidence list, and reader-safe boundary.
Analysis aid only. It does not certify truth, approve compliance, or replace expert judgment.
Shows the path from public source, model output, validation check, human review, and final decision boundary.
Mapping surface only. It does not automate approvals, filings, payments, trades, or customer actions.
Project features
The page focuses on what a project does: the feature, the input, the output, the evidence link, and the decision it deliberately leaves to people.
Each card explains what the project does, what goes in, what comes out, and where to open the demo, article, source, or evidence surface.
Review queues, evidence packets, source trails, control maps, scoring rules, dashboards, and checklists are shown as project features, not process notes.
Finance, AI, wallet, payment, and compliance projects state what they do not decide or execute so the value is understandable without overclaiming.
How to read this page
Start with a project card, open the related surface, and follow the input, output, feature, evidence, and boundary.