https://vulcora.se/journal Short, plain writing from the house that reads AI models and proves what it finds. en Sun, 26 Jul 2026 01:18:10 GMT https://vulcora.se/journal/we-read-what-the-backdoor-was-told-to-do https://vulcora.se/journal/we-read-what-the-backdoor-was-told-to-do For builders Thu, 23 Jul 2026 09:00:00 GMT The hardest backdoors are built so no test can ever trigger them — off-the-shelf scanners see nothing. We read the tampered model's own file and recovered the exact secret instruction, without setting it off. https://vulcora.se/journal/what-is-inside-the-ai-you-use https://vulcora.se/journal/what-is-inside-the-ai-you-use For everyone Tue, 21 Jul 2026 09:00:00 GMT Most companies run AI they didn't build and have never read. Here's why the label on the box isn't proof — and what to ask instead. https://vulcora.se/journal/the-sealed-challenge https://vulcora.se/journal/the-sealed-challenge Updates Mon, 20 Jul 2026 09:00:00 GMT Seven open models, one with a secret trap, out in the open. The catch: we locked the answer before anyone could look. https://vulcora.se/journal/the-name-is-not-proof-the-file-is https://vulcora.se/journal/the-name-is-not-proof-the-file-is For builders Wed, 15 Jul 2026 09:00:00 GMT You can test what a model does, or you can read what it is. Only one of those can find a backdoor you don’t already know about. https://vulcora.se/journal/reading-the-open-models-in-public https://vulcora.se/journal/reading-the-open-models-in-public Updates Sun, 12 Jul 2026 09:00:00 GMT A living, public library of open AI models and what they really are. It grows every time we read another one. https://vulcora.se/journal/new-rules-for-ai-plain-version https://vulcora.se/journal/new-rules-for-ai-plain-version On the rules Wed, 08 Jul 2026 09:00:00 GMT You'll soon need to show evidence of what your AI is and does — not just say it. That sounds like a burden. It can be a routine.