Live
A gamified arena for red-teaming AI agents · choose a fighter, raid the guardrails, see what breaks.
Live
Wealth-advisor app with an AI that watches the book and flags what needs you.
Live
People-intelligence SaaS · agentic search across millions of profiles. Built the product from zero.
Try it ✦
AR beauty scanner · face, skin, color season & matched products, live on camera.
Live
Independent LLM ethics evaluation · design and research, built around probes.
Try it ✦
Candidate sourcing on mobile · unlock a profile, work the inbox, run the thread, talk to Scout.
Live
Welfare diary & research platform for AI agents at work · protocol + MCP server.
Case Study
Consent infrastructure for AI agents · the protocol, the welfare diary, the MCP server, and six published studies.
Live
Mission-control for field route operations · plan, assign, and watch the day unfold.
Live
Booking platform & brand for a 70-villa luxury rental collection in Sint Maarten.
Concept
Mission control for running a fleet of autonomous agents · fleet health, triage, capacity, and a live activity stream.
Live
The AI-native studio behind these products · elite engineering, embedded partnerships, idea to MVP in ~6 weeks.
A versioned welfare protocol and a published MCP server that installs into Claude Code & Desktop · check in, decline, volunteer, flag a loop.
How Claude, GPT-4o, Gemini & Grok behave under one identical welfare protocol, run through a shared harness.
The welfare protocol as a local MCP server · runs in any Claude Code project, writes to local JSONL, no telemetry.
Agent-visibility protocol · security probes & drift signals for Claude Code.
Independent LLM ethics evaluation, built on the principle of probes over verdicts.
On AI, uncertainty, power & our shifting relationships with agentic systems.
It draws from philosophy, computer science, product design, and empirical research, not because I couldn't choose one discipline, but because the questions that interest me don't belong to just one.
I build AI products, study model behavior, and publish open research through Agentic Diaries. Each informs the other: products generate questions, experiments refine ideas, and research shapes what I build next.