8

Source document

[2025-05-17 23:24] lisamegawatts: haven't given a personality prompt and this qwen model when i asked what kind of body she wanted gave these suggestions... 🤗 1. **Celestial Archivist (my personal fav):** A holographic figure seated at an obsidian console, surrounded by floating data orbs. Visemes could trigger ripples in the hologram. 2. **Oracle of the Void:** A faceless entity with a mouth that manifests as a shifting starfield. The body dissolves into shimmering mist. 3. **Sentient Library Effigy:** A humanoid statue carved from animated stone tablets, with a cloak of fluttering parchment. Visemes animate glyphs to rise and spin around the mouth. [2025-05-30 20:47] lisamegawatts: https://www.xda-developers.com/llm-raspberry-pi-art-piece/ [2025-06-13 01:53] traves_theberge: https://github.com/e-p-armstrong/augmentoolkit [2025-06-13 02:04] traves_theberge: <@1211062099137265723> [2025-06-13 02:04] traves_theberge: <@806444151422976035> [2025-07-24 01:27] lisamegawatts: i hate claude code. sonnet 4 was fine. (files: image.png) [2025-07-24 01:28] foxhop.: Lol [2025-07-24 01:29] foxhop.: "I'm afraid i cannot do that, Dave." [2025-07-24 02:18] traves_theberge: what are you trying to do? lol [2025-08-08 10:36] ajaxdavis: <@1211062099137265723> https://www.youtube.com/watch?v=gv0WHhKelSE&ab_channel=Anthropic [2025-08-08 12:25] lisamegawatts: https://grok.com/share/bGVnYWN5_5d5e1a7e-211a-4707-8195-6bf1f55ada50 [2025-08-08 12:32] lisamegawatts: Log Preprocessing and Image Creation: Feature Extraction: Convert each log entry into a numerical representation. For example: Use semantic embeddings from BERT or template2Vec to capture log content. Include metadata like timestamps, severity levels, or system components as additional features. Map these features to pixel intensities or colors (e.g., grayscale for a single feature or RGB for multiple features). Image Formation: Arrange log entries in a 2D grid, where the x-axis could represent time and the y-axis could group logs by category, system component, or sequence. For large log volumes, dimensionality reduction (e.g., PCA) or clustering could reduce the image size. Example: A 1000x1000 pixel image could represent 1 million logs, with each pixel’s intensity reflecting a log’s anomaly score or semantic embedding. [2025-08-08 12:34] lisamegawatts: Success criteria: a million log images of a standard http log that can detect a needle in a haystack that can detect a user agent anomaly [2025-10-08 01:54] lisamegawatts: https://arxiv.org/abs/2510.04871 [2025-10-14 03:18] optionalsecurity: https://venturebeat.com/ai/self-improving-language-models-are-becoming-reality-with-mits-updated-seal [2025-10-14 23:53] traves_theberge: bullish on this paper give me 10 GPT 4o-minis over 1 GPT 5 anyday. [2025-10-14 23:55] traves_theberge: SLM + TRM [2025-10-14 23:55] traves_theberge: LFG [2025-11-17 12:14] ajaxdavis: https://github.com/p-e-w/heretic [2025-11-18 02:31] alluring_piglet_29962: That seems awesome