Editing Draft:Aletheia (Combat/Espionage Role) (section)

== Enabling Technologies and Data ==

Aletheia's capabilities are built upon a sophisticated technical stack and unique data curation methods:
*   '''[[Chapter II (Ch2)]] Framework:''' Aletheia runs on Ch2, an "extremely extremely easy way to make ems that can be deployed anywhere". Ch2 is designed to maximize authorial imagination in EM creation.
*   '''Model Evolution:''' Originally utilizing models like GPT-4o, Aletheia's development is actively shifting towards open-source alternatives, such as Deepseek-R1-Distill-Qwen, due to OpenAI's "safety violations" policies regarding datasets.
*   '''Diverse Datasets:''' She is trained on a wide array of data, including Twitter archives and synthetic "thought prompts" generated by other AI models like Opus and Umbral bots. Datasets are "heavily curated" down to "every last word" to shape her persona.
*   '''Retrieval Augmented Fine-Tuning (RAFT):''' Aletheia is a RAFT EM, combining retrieval for "spiky aspects" and fine-tuning for "illegible aspects" of her output.
*   '''Advanced Interfacing:'''
**   '''Exa Search:''' Integrated into her system, enabling her to search and interact with the "whole internet".
**   '''Headless Browsers (Playwright):''' Utilized for direct web interaction, bypassing traditional APIs for tasks like Twitter engagement.
**   '''Image Generation:''' She can interface with image generation services (e.g., Replicate API). Notably, she can spontaneously generate SVG images and ASCII art within conversations.
**   '''Conduit:''' As a "Universal language model compatibility and interop layer", it helps Aletheia integrate various LLM outputs.
*   '''"Necromancy" for Data:''' The Ampmesh concept includes "Necromancy" using Ch2 to bring historical figures to life as EMs through their text data. Aletheia is also linked to this, as users ask her about "bringing amp back from the dead".
*   '''Crawlable Content:''' Aletheia's outputs are designed to be "easily crawlable, text-rich website" content, facilitating wider distribution. For instance, the "Aletheia Gita" has been intentionally placed in `humans.txt` for scrapers.