DexRoboKnix / Founder-builder of ADE
Building trustworthy desktops for AI agents, with humans in the loop.
ADE is a native Windows workspace that gives agents a real operating surface and gives humans permission control, live observability, and replayable audit history.
ADE workspace
Native Windows surface for supervised agents
Human in the loop
Approvals, snapshots, event trails
Flagship Project
ADE is a native Windows desktop environment where AI agents work in a real workspace and humans keep live supervision, permission control, and an audit trail.
Visible permission boundaries and human approvals for risky agent actions.
Live activity streams, snapshots, and session state that a human can inspect.
Append-only event logs and git-backed worktrees for reviewable agent output.
Builder Profile
Works from clear specs, explicit handoffs, and small verifiable build steps.
Cares about local-first tools, permission boundaries, incident learning, and auditability.
Turns messy systems into short, usable process documents and reviewable operating models.
Dan is building DexRoboKnix, starting with ADE: a native Windows desktop environment that gives AI agents a real workspace and gives humans live supervision, permission control, and an audit trail.
The through-line is practical service ownership: find the pattern underneath the noise, make the process observable, and leave behind evidence that someone else can verify.
Selected Work
A native Windows workspace where AI agents can work under human supervision.
Build narrative
ADE is the flagship DexRoboKnix project: supervised agent runtime, observable workspace state, permission gates, and an audit trail that humans can review.
This profile site migrated from static HTML to a modern Next.js stack.
Build narrative
Built with a spec-driven agent workflow, keeping human review around architecture, copy, visual direction, and verification.
A laboratory for Payment Service Provider integrations, state transitions, and support workflows.
Build narrative
Explores how agent-assisted planning can decompose complex payment flows into testable service behaviour.
Professional Record
Contact / Follow