Our standards
Most writing about AI is commentary. Ours isn't allowed to be. This page is the methodology every Vita Indarra book must survive before it carries the name — published so you can hold us to it.
1. Build the real thing. Every book begins as a working system — an agent given a real budget, a model opened up and steered, an oversight surface put in front of real dangerous actions. Not a thought experiment. A thing that runs, on ordinary hardware, and can therefore fail.
2. Attack it on purpose. Before we trust a result, we try to break it: prompt injections, adversarial inputs, the shortcuts a real attacker or a lazy deployment would take. The failure modes most books quietly skip are the parts we consider publishable.
3. Measure honestly. Results come from pre-registered tests with controls where possible, and every number keeps its context: what it was measured on, what it wasn't, and where the evidence stops. When a favorite hypothesis dies in testing, the death is reported.
4. Publish only what survived. The frameworks in the books are the ones that held under attack — carried into print with the limits stated out loud.
Numbers are real or explicitly labeled as estimates or illustrations. Composite examples are flagged as composite. Named cases come only from the public record. There are no invented testimonials, no fabricated credentials, no "bestseller" costumes, and no statistic we can't tell you the nature of. Where a result comes from a simulation rather than the real world, the book says so — every time it appears.
When we're wrong, we fix the book — digital editions are updated, and material corrections are acknowledged rather than silently patched. A reader who spots an error can say so in an Amazon review; we read them, and honest criticism has already made these books sharper.
Vita Indarra publishes under an institutional byline, on purpose. It isn't secrecy about the work — the work is described in the books down to its failure modes — it's a position: in a field drowning in personality-driven hype, we'd rather you judge the evidence than a face. The systems are the credential. If a claim in our books can't stand without a persona attached to it, it doesn't belong in them.
We never pay for, trade for, or incentivize reviews. The review ask in the back of every book says "honest is the only kind we want," and it means it — a critical review that's true helps us more than a kind one that isn't.
That's the whole method. It's also, not coincidentally, the argument of every book we publish: trust is built from verification, not confidence. We just apply it to ourselves first.
