Maxime Mansiet

Blog

Seven Out of Seven, Then One in a Hundred

For five days in October, ChatGPT's links stopped saying where they came from. I built a lab to identify AI assistants from their network calls, and it worked every time. On real sites, it worked about once in a hundred.

AIResearchWeb MeasurementAttributionCiteMe
2026-10-17

Nobody Knows How Paracetamol Works

We trust drugs we do not fully understand because they passed a trial. I think software is heading the same way. On verifying instead of explaining, why Terence Tao is right to worry, and what you still owe.

AISoftware EngineeringVerificationAgentsMathematics
2026-10-14

The Job Moved Up One Floor

I can tell you exactly what my own product does. I could not tell you how it is built. On reviewing instead of writing, what that actually costs, and why I think it still works out.

AISoftware EngineeringCraftAgentsGeneralists
2026-09-23

The Future of AI Agents Is Verifiable, Customizable, and Yours

AI agents are getting autonomous. That's powerful, and dangerous. Verifiable credentials, scoped permissions, and decentralized identity are how we make them trustworthy for everyone, not just developers.

AI AgentsSSIVerifiable CredentialsHologramTrust
2026-04-16

Don't Trust Governments or Big Corps. Trust Yourself.

On trust, ZKPs, AI agents, and why self-sovereign identity is the central paradigm of this revolution.

SSIZKPAI AgentsPrivacyDecentralized Identity
2026-03-15

How Agentic AI Redefines Job Searching, From OSINT to Personalized Outreach at Scale

Job boards are noisy, LinkedIn is saturated. Here's how I used OpenClaw agents, OSINT techniques and email verification to reach recruiters directly, at scale, with personalized applications.

AIAgentic AIJob SearchOSINTAutomation
2026-03-10

Welcome: Who I Am and What I'm Building

A quick introduction to who I am, what self-sovereign identity is, and why I believe it will reshape how we interact online.

IntroductionSSIDecentralized Identity
2026-02-19