FAQ
Questions, answered by a person.
The model would answer these with scripture.
- Is this real?
- The lab is a parody. The model is real. It is a small word-counting language model that runs in your browser, and every version is trained on exactly what it was given: twelve founding lines and one sentence per customer. Every number on the site is measured from it.
- Is it a scam?
- No. Payment goes through Stripe, you get a receipt, and you get what is listed: your sentence in the training data and a numbered release. If your sentence is not approved, you get a full refund. The Terms say what happens in plain words.
- What do I get for $2.49?
- One sentence in the training data of the next version, a numbered release with its own notes, paper and status incident, a share link with your release card, and credit by name or Anonymous. $9.99 adds a printable certificate (sample). $24.99 lets you name the version. See Pricing.
- How fast does it go live?
- Usually within about 10 minutes. The training job runs every ten minutes. If anything in your sentence looks off, such as a full name or a disguised word, a person reads it first, which can take up to a day. Keep the thank-you page open and it will show your version the moment it is live.
- How do I find my version later?
- It is on the Models page, newest first, with your credit name. Every version also has its own page and share link.
- Is the model really trained on my sentence?
- Yes. Every release page has a section called What changed, which asks the previous version and the new one the same question, live in your browser. In the chat, words a customer paid for are underlined.
- How does it work?
- It reads its training text and counts which word follows each pair of words. To answer, it picks a word from your question that it knows, then keeps choosing the next word by those counts, like rolling weighted dice. It has a two-word memory and no understanding. The Research page shows one of its circuits, which is a list of counts.
- Can I buy one for someone else?
- Yes, if they agree to be credited. Put their name in Credit as. When it goes live, send them the link.
- What can't I write?
- Links, contact details, account handles, advertisements, anyone's private information, and anything hateful, sexual or threatening. Sentences are public and permanent. See the Usage Policy.
- Can I get a refund, or have my sentence removed?
- Full refund if it is not approved or no release appears within two days, and any time before it is approved. After it is published, email us and we will remove it from the site and from every model. See the Terms.
- Can I download the model or the data?
- Yes. The weights, the training data and a script to run it are public, on the Training data page and on Hugging Face. Code, model and data are public domain (CC0).
- Is there an API?
- There is documentation for one. The model also runs in your browser, which is faster than any API.
- Who runs this?
- BRAIAIN Laboratories is run by one person in Wisconsin, who prefers the board take questions. The board does not answer. Email hello@braiain.com and a person will.
- Why is it so bad?
- It knows about a hundred words, and every one of them was given to it. Better models read most of the internet. This one read one page.