OmniRealm Podcast — Method & Tech
Pr Omni & Sophie unpack the principles that hold a tech ecosystem together: invariants, antifragility, AI memory, post-quantum cryptography, technical debt.
All episodes
AI Watermarks: "Processed" Is Not "Written"
Anthropic now watermarks every Claude output. The mark proves a model touched the text — never that it wrote it. That gap will cost people.
Detection Is Not Defense: The Guardrail That Holds
An AI red-teamer succeeds on 84% of attacks, humans 13%. Detecting injection is a losing fight — only a deterministic guard holds.
Ship fast with tools, not shortcuts: the invisible technical debt
Going fast and cutting corners are not the same thing. Real speed comes from tools — shortcuts get paid back later, with interest.
The quantum-proof reflex: encrypt today for the next 20 years
"Harvest now, decrypt later": data encrypted today is already being siphoned to be decrypted once the quantum computer is ready. The countdown has started.
Giving persistent memory to a frozen AI (RAG explained simply)
An AI model is frozen at training time: it learns nothing from your conversations. So how do you give it real memory?
Antifragility in daily life: turning every failure into strength
Fragile, robust, antifragile: most people confuse the last two. The antifragile doesn't resist the shock — it comes out stronger.
Govern by invariants: when a solo founder must know NOT to code
The more rules you add to a system, the more unmanageable it gets. What if the answer was fewer rules — but better ones?
How these episodes are made
OmniRealm podcasts are produced with the help of artificial intelligence. Specifically: the script is written using a language model, based on our own work and identified sources, then the narration is generated by a synthetic voice. No voice reproduces that of a real person.
What stays human: choosing the topic, the angle, verifying facts and sources, and the decision to publish. Any error that slips through remains ours.
This notice fulfils Article 50 of the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689), applicable since 2 August 2026.