{
  "lastUpdated": "2026-08-04",
  "currentFocus": "Building multi-agent systems that hold up outside demos: agent architecture, model evaluation, and observability across the coding agents and local inference I actually run day to day.",
  "models": [
    {
      "name": "Claude (Anthropic)",
      "type": "subscription",
      "description": "Primary coding subscription. Claude Code is the main surface for full project builds, refactors, debugging, and architecture work."
    },
    {
      "name": "GPT / Codex (OpenAI)",
      "type": "subscription",
      "description": "Codex subscription for cross-model comparison, evaluation tasks, and the codex CLI harness used by local-model-plugin-cc and agent workflow experiments."
    },
    {
      "name": "Grok (xAI)",
      "type": "subscription",
      "description": "Hosted frontier model via Grok Build CLI: review, adversarial review, and rescue workflows through grok-plugin-cc; also used in model-prompt-tests benchmarks."
    },
    {
      "name": "Local models",
      "type": "local",
      "description": "Ollama, LM Studio, and OpenAI-compatible endpoints for local inference. Powers Hermes agents, llama-swap style routing, and the local-model-plugin-cc review/rescue path without cloud dependencies."
    }
  ],
  "dailyDrivers": [
    {
      "tool": "Claude Code",
      "role": "Primary coding agent. Full project builds, refactoring, debugging, and architecture decisions.",
      "status": "daily"
    },
    {
      "tool": "Codex CLI",
      "role": "Comparison agent and the broker target for local-model-plugin-cc (Ollama / LM Studio / custom endpoints).",
      "status": "daily"
    },
    {
      "tool": "OpenCode",
      "role": "Active daily use. Evaluating how it compares to Claude Code and Codex for agent-driven workflows.",
      "url": "https://opencode.ai/",
      "status": "daily"
    },
    {
      "tool": "Grok Build CLI",
      "role": "Hosted Grok review and rescue path via grok-plugin-cc; headless runs with explicit safety gates.",
      "status": "daily"
    },
    {
      "tool": "Mission Control",
      "role": "Unified usage and runtime dashboard across Claude Code, Codex, Hermes, and related lab sources (sessions, tokens, failures, and machine load).",
      "url": "https://github.com/BenBish/mission-control",
      "status": "daily"
    },
    {
      "tool": "astro-llms",
      "role": "Astro integration that generates curated llms.txt and per-entry Markdown mirrors from Content Layer collections. Powers bshp.io agent surfaces at build time.",
      "url": "https://github.com/BenBish/astro-llms",
      "status": "daily"
    },
    {
      "tool": "omp",
      "role": "Terminal coding agent with IDE-oriented tooling (LSP, subagents, sessions). Evaluating fit against Claude Code and OpenCode.",
      "url": "https://omp.sh/",
      "status": "exploring"
    }
  ],
  "url": "https://bshp.io/stack"
}