TerpBot

Hermes Agent setup, migration, LightRAG, Telegram, and skill creation guide

Run any model in Grok Bot — one-command setup, model picker UI, evidence-based provider wire maps, and an update-proof doctor. Not farming you, arming you.

Give Claude Code's ultracode mode to ANY model you already pay for. A tiny local proxy + one config.json. Point your AI at AGENTS.md and it sets itself up.

Kill the tool-call tax: harness tool latency back below model TPS. Fast local lanes, persistent pools, session caches. Scratch lab turned real.

Drop-in prompt-caching fixes for the LLM agent harness you use. Point your AI coding agent at this repo and it ships the patches.

First open-source implementation of Google TurboQuant (ICLR 2026) -- near-optimal KV cache compression for LLM inference. 5x compression with near-zero quality loss.

Paper-trade Solana memecoins on the sites you already use. Real prices, fake money, and a record you can actually learn from.

Every feature Cascade ships that most people aren't using — configured properly

Unlock the true power of DevinCLI with all the of the resources I have gathered for you

The old advice to avoid Windows for local LLMs used to be right. It isn't anymore. The fixes for display-GPU desktop crashes and WSL memory limits, from people who run a 5090 daily.

Caduceus — Hermes-native UltraCode dynamic-workflow mode (Dynamic Workflows)

29 controlled experiments on Nemotron 3 Ultra: what prompting can and cannot do to a frontier model. Voice 4/8 to 7/8, hidden bugs 1/5 to 5/5 — blind dual-judge graded, no fine-tuning.

OpenFlow — local-first multi-engine Electron dictation for Windows

Native multi-model selector and safe updater for Grok Build

Persistent tmux session management with first-class Warp tab integration

Mega-swarm web+X research for Grok CLI harness improvement ideas

A stateful reasoning runtime for OMP coding agents. Evidence ledger, obligation tracking, verification gates, and identity-aware retry blocking.

GPU-native screen vision for AI agents — DXGI Desktop Duplication capture + compositor dirty-rect change feed. 96% less image data than full-frame screenshots.