Route the request. Bound the action. Leave a receipt.
A 54-second English introduction built from a real provider inventory and decision-plane readback. The motion panels are illustrative; the shown values are pinned to the inspected public commit.
I design and ship AI-native developer products: reviewable flows, local-first agent infrastructure, and teaching tools that make their boundaries visible.
Real assets from released local flows. Open a demo, read the case study, or inspect the repository.
A 54-second English introduction built from a real provider inventory and decision-plane readback. The motion panels are illustrative; the shown values are pinned to the inspected public commit.

A native interface design engine in the DESIGN:OS ecosystem. My public repository preserves the fork boundary and links to the source and license context.

Static HTML and code become a self-contained flow graph. Revision comparison adds source evidence, reviewer notes, status, and JSON export over labeled synthetic CRM fixtures.

Local Vietnamese fraction-teaching rehearsal with authored coaching, transfer, replay, and Markdown export. The simulation supports practice; it does not certify teaching competence.

File-based Markdown and JSON bridge with metadata and stale-state guards.
System Agentic AI keeps provider policy, execution boundaries, and verification explicit across the workspace.
Shared operating rules separate host, worker, provider, and account identity.
Read-only, workspace-write, and elevated boundaries stay visible and fail closed.
Smoke checks, source pins, and independent verification distinguish facts from unknowns.
Input, output, trust boundary, and failure behavior are part of the product.
A small local flow beats a large promise. Link the exact source and version.
“Not verified” is a useful result when the evidence is not available.