If you've ever interacted with a model like ChatGPT or Llama 3, you've probably noticed they sometimes refuse to answer certain questions. These refusals aren't random; they result from a deliberate "safety alignment" trained into the model. Removing this filter—a process called "abliteration"—has always been a complex task, requiring technical expertise and manual trial and error. Heretic, created by p-e-w, changes the game by fully automating this process.

To the Ministries of Health and Social Affairs
To the Presidents of the Regions
To the Orders of Physicians and Dentists
To the Health Directorates of Hospitals and Clinics
To all General Practitioners and Specialists

Brief overview of general indications that should be implemented, regarding materials used by various professionals dealing with these diseases and by the patients themselves who have contributed by evaluating products on themselves over time.