Both pipelines run 100% on the 3090 workstation (192.168.1.143). Your Mac does zero heavy lifting — it only sends the command and collects the finished .mp4.
What it does: Extracts the raw mathematical skeleton from the video, builds a real 3D rig with a weighted mesh, and renders a studio-lit silver mannequin video using Blender's EEVEE engine — all headlessly on the 3090.
Created: 2026-05-04T03:40:01.378182+00:00
Run 090 appears to contain meme images and a caption but not the standard SGFLIX factory package surface (FACTORY_RUN_STATUS.md, research, strategy, audio, CHAI, handoffs, manifests, and QC files). This cycle did not modify or overwrite its generated assets.
Recommended next step: either formalize Run 090 as a lightweight meme package with status/QC/manifest files, or mark it as non-official/aborted so future factory cycles do not treat it as a completed numbered run.
| import torch | |
| import json | |
| import os | |
| import numpy as np | |
| def export_mhr_from_pt(pt_path="mhr_model.pt"): | |
| if not os.path.exists(pt_path): | |
| print(f"Error: {pt_path} not found.") | |
| return |
A pattern for building personal knowledge bases using LLMs.
This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.
Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.
Consolidated: March 25, 2026 Original dates: January 7-8, 2026 Sources: 10 strategic/architecture gists consolidated into one Contents: Master Synthesis, Risk Analysis, ggen Comparison, Code Archaeology, Naming Schemes, Use Cases, Enhancement Plan, System Context, Features Roadmap, Oracle Handoff
A technical journal of building dual-process voice AI on top of Kyutai's Moshi and NVIDIA's PersonaPlex, with Letta as the reasoning backbone.
Build a Talker-Reasoner voice system where:
