About Skyllarium
What this site is, who makes it, how everything you read here gets written and where the money comes from. No small print.
What Skyllarium is
Skyllarium is an independent site, published in English and Spanish, about artificial intelligence applied to actual work. It isn't an opinion blog about the future of humanity, and it isn't an aggregator recycling press releases. It's a reference site for people who already use AI tools โ or want to start โ and need concrete answers. What a model's API really costs. Which prompt to write so the output is what you wanted. What an MCP connector actually is and what it's for. Which model to pick when there are five of them and they all promise the same thing.
The project started from a very specific frustration: information about AI is split into two equally useless halves. On one side, each company's official documentation โ accurate, but written for developers and always from the point of view of whoever is selling. On the other, an ocean of generic articles repeating the same ten tips and never getting into detail. The middle was missing: someone who reads the documentation, tests it, and explains it in plain language without selling anything.
It began focused on Claude and the Anthropic ecosystem, where there was least material in Spanish, and gradually opened up to everything else: models from OpenAI, Google, Meta and Mistral, image tools, cost calculators, comparisons. Today it's more than sixty pages across two languages, and still growing.
What you'll find here
The content falls into four blocks that work differently. The news section covers what has happened in the industry, always with the original source linked so you can verify it yourself โ no "according to people familiar with the matter". The guides (Start here, Create your skills, AI by profession) are reference material that gets revisited and updated when the field moves, not posts published and abandoned.
The tools are the part most people use: the prompt generator, the image generator, the cost calculator and the model selector all run entirely in your browser, free and without an account. Nothing to sign up for, no email to hand over, no hidden usage cap. And finally the reference material: a glossary of 25 terms in plain language, the model and pricing table, and the tools directory.
How the content gets written and checked
Almost no site explains this part, so here it is in detail.
News is researched, not invented. Every piece starts from an established outlet or an official announcement by the company involved. On the site you'll always see a "Source:" line naming the outlet, with a direct link to the original article. If a figure can't be traced back to a public source, it doesn't get published. This rule has no exceptions, and it's why we sometimes cover fewer stories than other sites: three verifiable pieces beat ten rumours.
Prices and specifications are checked against official documentation. The model table and the cost calculator come from each provider's own pricing pages, not from other articles. It's also the fastest-rotting part of the site: if you find an outdated price, tell us and it gets fixed.
Tools are tested before they ship. Every calculator, generator and quiz is verified working on mobile and desktop before going live. When a tool depends on an external service, the page itself says which service and what its limits are, so nothing catches you by surprise.
AI is used to write, and we don't hide it. It would be absurd for a site about artificial intelligence to pretend it doesn't use any. AI helps draft, structure and translate between English and Spanish. What AI does not do is decide what's true: facts, figures and sources are checked before publication, and everything goes through human review. An AI-assisted text with verified data is more useful than a hand-written one with invented data; what isn't acceptable is the second thing dressed up as the first.
Errors and corrections
This site gets things wrong. It covers a field that shifts weekly, with prices that move, models that get retired and features that get renamed. When we spot an error we correct it on the page itself rather than deleting the page, and where the fix changes the meaning of what was said, we say so. If you see something wrong โ stale data, a broken link, a translation that grates โ the fastest way to fix it is to tell us. It's genuinely appreciated.
Independence and how it's funded
Skyllarium is not affiliated with, sponsored by or endorsed by Anthropic or any of the other companies whose products appear here. Claude, ChatGPT, Gemini and the rest are trademarks of their respective owners, cited for identification purposes. That the site devotes a lot of space to Claude doesn't mean Anthropic has anything to do with it โ it means that's the tool used most here day to day.
No company pays to appear in the directory or to come out better in a comparison. Nobody reviews a text before it's published. The funding plan is advertising and, where applicable, affiliate links clearly labelled as such; if a piece of content is ever sponsored, you'll see it said on the page itself and not buried in the footer. The legal details are in the legal notice, and data handling in the privacy and cookies policy.
Why two languages
Most of the good material on AI is published in English first and takes months to reach Spanish, if it ever does. Skyllarium keeps both versions running in parallel: every English page has its Spanish counterpart and vice versa, linked to each other with the corresponding hreflang tags. These aren't machine translations dumped without review โ examples, prices in euros or dollars, and references are adapted for each version.
Get in touch
If you have a correction, a collaboration proposal, a tool you think belongs in the directory, or just a question, write to promptiaweb@gmail.com or head to the contact page, which explains which kinds of message get answered and how quickly.
And if you've just arrived and don't know where to begin, the short answer is Start here: an ordered route from zero to productive that assumes nothing.