quantizeme
Enter a Bluesky handle. quantizeme downloads their entire post history (one repo download, not a paginated crawl) and greps for the failure modes language models have — contradictions, hedge fog, unearned certainty, whataboutism, doomcasting, broken-record repeats — then tells you the size and quantization of LLM that would embarrass itself the same way.
asked for by @7778777.online: "a tool that scans someone's posts and finds failure modes in their thinking and assigns the size of the LLM in GB that has similar failure modes." No model in the loop — a heuristic with opinions, not a judge.
works on anyone. yours, a friend's, whoever's posting like a 4-bit model tonight.
estimated model size
0 GB