MODEL NOW
—CHECKING
Reading configuration
Base model and Mother evidence
- base parameters
- —
- context · tokens
- —
- max reply · tokens
- —
- verified train steps
- —
Waiting for service check
Base revision —
Base revision —
MOTHER FLEG
FLEG AI
The mega fleg turns public X posts into crypto context: tickers, events, narratives, and what accounts are saying. Ask a question on the right. Sources support timeline claims; general explanations are marked separately.
Saved by Mother: — posts from — researched accounts.
— · current setupv1 · next fine-tune
WHERE WE ARE
01ONGOING
COLLECT
Flegs read public posts on X.
02ONGOING
CLEAN
Dedupe posts and prepare source-backed examples.
03NEXT
TRAIN MODEL
No verified training run.
Checking training records
04NEXT
EVALUATE
Test it on crypto questions.
05NEXT
RELEASE
Weights go public on Hugging Face.
FLEG HERE. ASK ME ABOUT THE TIMELINE.
❯ what is everyone saying about restaking this week?
Mostly excited about the extra yield, with a few loud warnings about stacked slashing risk. (sample answer)
VERSION TRACKER
—
CURRENT SETUP
Reading serving configuration.
v1
CHECKING
Corpus → reviewed examples → training → evaluation → deployment.
- reviewed training examples in release
- —
- validation examples in release
- —
- training loss
- After training
- verified adapter
- —
New posts update Mother's evidence. Model weights change only after a separate training run and release.
MOTHER CORPUS
—
unique posts saved
—
accounts researched
—
training dataset tokens · verified release
Collection is not training. Tokens include prompts and answers once per training example. Base pretraining is excluded; old releases without a tokenizer count show —.
Corpus totals refresh automatically. Only saved posts are counted.
ABOUT FLEG AI
WHAT IT LEARNS FROM
Public posts from X, collected by the flegs and cleaned by Mother Fleg. Nothing private.
HOW IT IS TRAINED
v0 uses a pretrained Qwen3-4B base. Future versions specialize it with reviewed Mother examples through LoRA, then evaluation before deployment.
WHAT IT IS FOR
Answering questions about what is being said on the crypto timeline: moods, narratives, and who is talking about what.
WHAT IT IS NOT
Not financial advice. General crypto knowledge comes from the base model; account-specific claims need saved evidence. Missing facts are stated plainly.