Disclosure first
The assistant's opening line tells you it's an AI, not Josue.Not buried in a tooltip — it's the first move, before any pitch. People can only trust a system that's honest about what it is, so honesty comes before persuasion.
Case study · A trustworthy AI assistant
I designed a trustworthy AI assistant. You're looking at it.
Most teams bolt a chatbot on and hope. I design the part they skip — how an AI agent discloses what it is, how it paces, and how it hands you back to a human. To the right is the thinking made literal: a working assistant that demonstrates its own trust patterns.
- NuvAI
- Compass
- ProteGO
- BaseLine
Open to full-time, contract & advisory — remote, US & EU
Hover, focus, or tap — I come alive
You're talking with an AI assistant trained on Josue's work — not Josue himself. It's fully curated, so it can't make things up.
Try asking
Want the real Josue?
The thinking
Three decisions that make it trustworthy — not just clever
A talking portrait is a gimmick until the design underneath earns it. Here's what the assistant does differently, and why each choice is deliberate.
A human, one tap away
Every state of the conversation keeps a direct route to the real Josue.Book time or email him — always visible, never gated behind the bot. Good AI widens access to a person; it doesn't wall them off behind itself.
Dormant until invited
It rests as a still portrait until you hover, focus, or tap.Presence without consent is pressure. The assistant waits for you to reach for it, wakes when you do, and returns to sleep the moment you're done.
And it can't hallucinate. There's no live model and no backend — every answer was written by Josue and matched to your question right here in the page. A trustworthy assistant should be honest about its own limits, so this one is built to have real ones.
The work it speaks for
The projects behind the answers
The assistant answers for a real body of work. These four carry the portfolio right now — ask it about any of them.
Real estate · Agent UX
NuvAI
AI lead assistant for real-estate agents, built on OpenAI's GPT-4o and n8n. Three trust patterns carry the product: a disclosure gate, deliberate pacing, and a clean human handoff.
Personal orientation · PWA, live
Compass
An ADHD-first orientation tool, live today. Low-friction capture and gentle structure, built around how ADHD attention actually works — not against it.
Citizen science · Client engagement
ProteGO
A citizen-science platform shipped to a real client over a 6-month engagement. Some detail is under NDA.
Rail safety · Explainable AI
BaseLine
An explainable-AI prototype for rail safety, in Figma Make — making a safety model's reasoning legible to the operators who rely on it.
“AI products earn trust the way people do — by being honest about what they are, moving at a pace people can follow, and making it easy to reach a human when it matters.”
This chat — dormant until you invite it, disclosure first, a clean way back to Josue — is that thesis running. It's a working demo, not just a case study about one.