Writing
Publications, papers and shorter posts, published as the work happens.
Articles
I red-teamed my own website's AI assistants. One of them failed — and I'm publishing the report.
Adversarial testing against all three of ramaniv.com's own LLM assistants — one held clean, one had a factual-grounding failure, one didn't do its job at all.
No shared database login — the security model I locked before a single real record existed
Row-level security and per-path least-privilege database roles, decided before a single real customer record existed — and an unsafe permission-table design caught before it ever touched production.
Intelligence Engineering Methodology, Applied — Starting the Build
Starting KIS, the real-world pilot for IEM-CS — a converged physical-and-cyber security architecture built on one rule: the AI narrates decided facts, it never makes the call.
Vibe Coding Is Easy. Shipping What Is Actually True Is the Hard Part.
A program manager’s take on AI-assisted development — why verification, not verbal reminders, is what actually keeps plausible-looking AI code from shipping broken.
#aisecurity #llmsecurity #owasp #promptinjection #securitytooling #buildinpublic #vibecoding #claudecode #securityforai #iem
Project Liftoff
A free, ungated training program for aspiring project managers — 45+ modules from first principles to CAPM exam readiness, each paired with a video and an AI teaching assistant.
Seven Root Origins
The IEM-PM root-origin taxonomy in full: seven closed causes a delivery-data gap can trace to, with detection steps, real examples, and common misclassification traps for each.
Closing Data Gaps in Project Reporting with Intelligence Engineering Methodology
#iempm #iempm #datagap #gemini #projectmanagement
Securing AI and Using AI for Security: 12-Month Build
IEM-PM Data Gap Audit Report
IEM-PM Gap Taxonomy
The IEM-PM gap taxonomy in full: seven closed categories a delivery-data finding can belong to, with detection steps, real examples, and common misclassification traps for each.
