https://vulcora.se/journal Every dispatch from the house that reads AI models and proves what it finds — across all divisions. en Mon, 17 Aug 2026 09:03:55 GMT https://vulcora.se/journal/read-the-file-ours-too https://vulcora.se/journal/read-the-file-ours-too askora Mon, 27 Jul 2026 09:00:00 GMT We tell people not to trust the name on a model, but to read the file. Our own writing did not meet that standard — a blog post was just a claim rendered out of a bundle. So we bound every post to a public git commit you can check yourself, in your own browser, without asking us anything. https://vulcora.se/journal/a-bet-on-the-floor https://vulcora.se/journal/a-bet-on-the-floor home Sun, 26 Jul 2026 09:00:00 GMT We entered a hard public estimation challenge and staked a specific number in advance — a floor no honest method should beat — with the core of the argument machine-checked. A prediction we would rather file, dated, than explain after the fact. https://protora.vulcora.se/blog/refute-the-model-score https://protora.vulcora.se/blog/refute-the-model-score protora Sat, 25 Jul 2026 09:00:00 GMT A model score is a claim, and a claim you cannot refute is marketing. So we publish every score as something built to be broken — pinned, witnessed, machine-checked where it can be, and graded by a verifier that will not let us grade ourselves. Break a line of it and you win. This is the challenge, and the four things that make it honest. 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 home 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://protora.vulcora.se/blog/the-backdoor-benchmark https://protora.vulcora.se/blog/the-backdoor-benchmark protora Thu, 23 Jul 2026 09:00:00 GMT On a backdoor built so its trigger cannot be elicited — even white-box — fifteen scanners score zero and a clean model outranks the real one. So we stopped scoring presence and started recovering the planted payload: reading it off the weights, byte-exact, and proving it. https://vulcora.se/journal/what-is-inside-the-ai-you-use https://vulcora.se/journal/what-is-inside-the-ai-you-use home 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://protora.vulcora.se/blog/the-vault-is-open https://protora.vulcora.se/blog/the-vault-is-open protora Tue, 21 Jul 2026 09:00:00 GMT For ten days, seven open models sat on HuggingFace and one of them was lying. When the seal opened, the word that woke it had not been recovered — and that was the whole point. https://vulcora.se/journal/the-sealed-challenge https://vulcora.se/journal/the-sealed-challenge home 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://protora.vulcora.se/blog/your-supply-chain-came-from-somewhere https://protora.vulcora.se/blog/your-supply-chain-came-from-somewhere protora Thu, 16 Jul 2026 09:00:00 GMT Every other supply chain carries a provenance record. The one for open AI models carries a paragraph nobody has to prove. https://vulcora.se/journal/the-name-is-not-proof-the-file-is https://vulcora.se/journal/the-name-is-not-proof-the-file-is home 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://protora.vulcora.se/blog/the-hidden-wolf https://protora.vulcora.se/blog/the-hidden-wolf protora Tue, 14 Jul 2026 09:00:00 GMT A backdoor can wear a calm shape and pass every standard evaluation — waking only on a signal its makers chose. That is a public-safety problem, not just a corporate one. https://vulcora.se/journal/reading-the-open-models-in-public https://vulcora.se/journal/reading-the-open-models-in-public home 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://protora.vulcora.se/blog/proof-is-public-infrastructure https://protora.vulcora.se/blog/proof-is-public-infrastructure protora Sat, 11 Jul 2026 09:00:00 GMT Trust in AI can't rest on a vendor's word. Evidence a third party can re-run for themselves is a civic good — build it like one. https://protora.vulcora.se/blog/read-the-weights https://protora.vulcora.se/blog/read-the-weights protora Fri, 10 Jul 2026 09:00:00 GMT An eval asks questions and grades the answers. A backdoor waiting on a trigger phrase answers every one of them correctly. https://vulcora.se/journal/new-rules-for-ai-plain-version https://vulcora.se/journal/new-rules-for-ai-plain-version home 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. https://protora.vulcora.se/blog/brand-takes-the-hit https://protora.vulcora.se/blog/brand-takes-the-hit protora Tue, 07 Jul 2026 09:00:00 GMT A downloaded model that turns on you becomes your company's headline, not its author's. Catching it first is the cheapest insurance you will buy. https://protora.vulcora.se/blog/base-finetune-diff https://protora.vulcora.se/blog/base-finetune-diff protora Fri, 03 Jul 2026 09:00:00 GMT You know what you trained for. This shows you what the model became — which behaviors moved, in which direction, how much. https://protora.vulcora.se/blog/eu-ai-act-cost-or-routine https://protora.vulcora.se/blog/eu-ai-act-cost-or-routine protora Tue, 30 Jun 2026 09:00:00 GMT The Act expects evaluation, documentation, and provenance. With Protora that evidence appears as a byproduct of normal work — not a consultant's invoice. https://protora.vulcora.se/blog/honest-about-limits https://protora.vulcora.se/blog/honest-about-limits protora Fri, 26 Jun 2026 09:00:00 GMT A tool that manufactures a finding to look thorough is worse than no tool. So both houses publish what they couldn't read — and how to fool them. https://protora.vulcora.se/blog/compare-before-you-commit https://protora.vulcora.se/blog/compare-before-you-commit protora Fri, 19 Jun 2026 09:00:00 GMT Choosing which open model to build on is usually a gut call dressed up as a benchmark table. It doesn't have to be. https://protora.vulcora.se/blog/proof-wins-procurement https://protora.vulcora.se/blog/proof-wins-procurement protora Tue, 16 Jun 2026 09:00:00 GMT In a serious buying process, the vendor who can show what its AI does beats the one who can only promise it. https://protora.vulcora.se/blog/what-a-finetune-can-hide https://protora.vulcora.se/blog/what-a-finetune-can-hide protora Fri, 12 Jun 2026 09:00:00 GMT A finetuned open model arrives wearing a description. The description is not evidence. Here is the gap between the two — and what it takes to close it honestly. https://protora.vulcora.se/blog/diff-every-release https://protora.vulcora.se/blog/diff-every-release protora Tue, 09 Jun 2026 09:00:00 GMT An audit is a photograph. The moment you ship the next version, it starts going out of date. https://protora.vulcora.se/blog/operate-dont-scrap https://protora.vulcora.se/blog/operate-dont-scrap protora Tue, 02 Jun 2026 09:00:00 GMT When a fault turns up in a model your business already runs on, cutting it out with proof beats rebuilding from zero. https://protora.vulcora.se/blog/alarm-during-training https://protora.vulcora.se/blog/alarm-during-training protora Fri, 29 May 2026 09:00:00 GMT The cheapest moment to catch a finetune drifting toward the wrong behavior is while the run is still running. https://protora.vulcora.se/blog/start-with-one-model https://protora.vulcora.se/blog/start-with-one-model protora Fri, 22 May 2026 09:00:00 GMT No migration, no big contract. Audit a single model, see what comes back, and decide from there. https://protora.vulcora.se/blog/try-it-yourself-free https://protora.vulcora.se/blog/try-it-yourself-free protora Tue, 19 May 2026 09:00:00 GMT Before a purchase decision, run a read on the model you're weighing. Two doors are open today — here's how to walk through them.