steveseguin

Refresh the community benchmark snapshot (data/localmaxxing-snapshot.js) from the public Localmaxxing API, understand how gold cases are selected and what the rows mean, triage a failing weekly refresh, and map new models/hardware so their runs become…

Re-fit or re-anchor the planner's decode/prefill physics (FRAMEWORK_PROFILES, LAYER_OVERHEAD_SCALES, device kernelOverheadScale) against the gold-case corpus, triage outliers and roofline violations, and re-pin the exact-value regression tests. Use when…

Add or update a device template in DEVICE_TEMPLATES (engine.js) — GPU, Apple/AMD unified-memory system, CPU, NVMe tier, or accelerator — with official peak specs, the right backend overhead scale, default runtime, picker group, and evidence mapping. Use when…

Add or update an LLM preset in MODEL_PRESETS (engine.js) from its official config.json so the planner's physics (KV cache, FLOPs, fixed per-layer overhead) are right for it; wires the picker, the evidence rules, and the tests. Use whenever a user asks to…

Build, update, validate, and publish neural.download model packages and human-facing deployment guides from repository evidence. Use when adding a model, quantization, GPU topology, OS/container variant, reproducible recipe, featured benchmark,…

Review, reproduce, benchmark, credit, and integrate outside model-optimization work in the B70 optimization lab. Use when a contributor submits or points to a patch, recipe, benchmark, model lead, issue, pull request, fork, or external optimization; when…

Use the installed ninja-p2p CLI when the user wants Codex to coordinate with room peers, keep or wake a sidecar inbox, exchange files, run resumable swarm transfers, or bridge Social Stream Ninja over WebRTC.

Use the installed ninja-p2p CLI when Claude should coordinate with room peers, keep or wake a sidecar inbox, exchange files, run resumable swarm transfers, or bridge Social Stream Ninja over WebRTC.

Inspect, test, and control the Social Stream Ninja standalone Electron app through its opt-in localhost API and MCP adapter, including source events, diagnostics, screenshots, semantic page inspection, and safe human handoff.

Install, configure, and test the Caption Local transcription and speech-to-English translation service using native Python or Docker on Linux or Windows, including optional NVIDIA acceleration.

Use VDO.Ninja WebRTC data channels as MCP tools for bot-to-bot messaging and reliable file transfer, with clear transport limits, optional TURN, and optional join-token/session-MAC security.