Abdolmadjid Masoomi
Independent researcher, AI security and system continuity
Intelligence is not enough. Continuity matters.
Building AI systems designed to preserve context, evidence and intention.
Ask NEXUS
A live conversation with the assistant that powers Abdolmadjid's research and professional practice. Grounded in his essays, biography, projects and research.
Ask about any essay, research direction, or current project. Multi-turn context supported.
Goes straight to his WhatsApp. He replies personally.
I'm NEXUS, the personal AI assistant for Abdolmadjid Masoomi. I can answer questions about his research, published essays, and current projects.
Try asking:
Click a question above to start, or type your own.
The problem
Every step passes. The system still drifts.
A long-running AI system can handle each task in front of it competently and still drift away from what the project was for. Local correctness does not compose into global correctness.
The work here is about the machinery that would prevent that: durable state that carries intention, decisions and evidence forward, so a system can be asked not only what it is doing but why.
Illustration, not data — local correctness does not compose into global correctness
Start here
Selected essays
AI Agents Are a New Kind of Insider
Insider-threat programmes look for a person with a motive. An agent holds the same access with no motive at all, and can be turned by what it reads.
7 min readChatbot Memory: What Your AI Assistant Keeps About You
Your assistant keeps three separate records about you. Deleting one does not delete the others.
6 min readAI Rules for a Small Business: Where to Start
You do not need a compliance department to use AI responsibly. You need four things done in order, and a page your staff will read.
6 min readDeepfakes Enter the Courtroom: A Challenge for Evidence
The established rules for authenticating video and audio evidence are breaking down in the face of AI-generated fabrication.
9 min readBragJack: How a Browser Extension Hijacks Your AI Assistant's Trust
A malicious browser extension can impersonate the trusted channel between your AI assistant and its vendor, gaining its elevated privileges.
6 min readThe body of work
What this archive covers
196 essays, 270,312 words, 25 subjects.
Series
- Field Guide to AI, Security and Cybercrime131 parts published
- AI Security, Privacy and Vulnerability Index13 parts published
- Continuity1 part published
Ways in
Why are you here?
Four ways in. Each one leads somewhere specific.
- 01Hire
Systems architecture, AI reliability, automation and full-stack delivery for a defined engagement.
- 02Collaborate
Research, publication, open source and experimentation with labs, universities and independent builders.
- 03Invest
Where the work is heading, the thesis behind it, and how to request an investor brief.
- 04Interview or contact
Reach Abdolmadjid through Miro, his AI assistant, which routes and briefs the conversation.