Your memory, in your language. And a note on getting it right

Most AI is built in English first and everyone else second. We understand why, but we do not think it is good enough. The context of your life is not in English unless your life is in English. Your documents, your notes, the names of the people who matter to you: they are in your language, and your memory should work in your language.

So here is where we are, honestly, and here is where we need your help.

What works today

You can bring your documents in any language and they become memory. Word, PDF, email, the formats that carry the real detail of your life, all read cleanly whatever language they are written in.

You can search by meaning in more than 100 languages. Ask a question in your own words, in your own language, and AI Twin finds what you meant. It works across languages too: a question in one language can find the document that means the same thing in another. That is a genuine strength of the system, and it is live now.

You can find things by exact name in any language or script. And you can talk to AI Twin about your memory in dozens of languages, asking in yours and being answered in yours.

What we are still improving, and we will say it plainly

Two things are not where we want them yet, and we would rather tell you than let you find out.

Keyword search, the kind that matches exact word forms, is tuned for English today. For languages with rich word endings, that means keyword matching is weaker than it is in English. Search by meaning and search by exact name cover most of the gap, and per-language tuning is on our roadmap. But we will not pretend it is finished when it is not.

And the interface itself is in English today. Your memory, your documents, and your questions all work in your language. The buttons and menus around them do not yet. Interface translation is on the roadmap, not shipped, and we grade it that way everywhere on the site.

The language pages, and an honest request

We have built dedicated pages for fourteen languages, so that a French speaker, a Punjabi speaker, an Arabic speaker can read about AI Twin in their own language, right down to the direction the text runs. See them all on our languages page.

Some of those pages we are confident in. Some were translated with AI assistance and are waiting for a native speaker to check them properly, and we mark those pages clearly as pending review. We are not going to present a machine translation as if a native speaker wrote it.

Here is the request. If you read one of our language pages and something is wrong, clumsy, or simply not how a person would say it, please tell us. We would genuinely rather know. Send us the page, the phrase, and how it should read, on our feedback page. We will fix it, and we will keep fixing it.

And if we have got something wrong in your language already, we apologise. Building for many languages honestly means getting things wrong in the open and correcting them, rather than only ever shipping English and calling it done. We would rather try, be corrected, and improve, than not try at all.

Your memory should work in your language. We are not all the way there. We are committed to getting there, and we would like your help.

Tell us what we got wrong on our feedback page.