qwenstradamus
Train an AI on your tweets that writes in your voice. Upload your X archive, label a few of your own tweets, and download a model to run locally.
Sign in
Sign in with your X account to create and manage your models.
signed in as · your models · log out
Your models
1Upload your X archive
Get it from X → Settings → Download an archive of your data. Drop in the
whole .zip — your tweets are extracted right here in your browser, so only that small
file is uploaded (never the multi-GB archive). You can also pick just data/tweets.js.
Working…
2Label your tweets
For each tweet, write the prompt it answers — the question or situation that would make you tweet this. The tweet stays your real words.
Goal reached — keep going for an even better model, or train whenever you're ready.
Already have labels from before? Upload a labels file (.jsonl of {prompt, completion}).
3Add a bio (optional)
Current facts your tweets can't convey — grounds the model so it knows who/where you are today. Skip if you like.
4Add eval prompts
Enter up to 10 prompts (one per line). After training we'll run them through your model so you can see it answer in your voice.
Bigger models capture your voice more sharply, but cost more and take longer to train.
5Your model is ready
Run it locally with Ollama — the download includes a Modelfile and instructions.
Only 0 of 300 labeled
More labeled prompts make a noticeably better model — 300 is the sweet spot. You can train now, but the voice may be weaker. You can always come back and label more before training, too.