Latest
finding the newest…
The List
New essays when they are finished. No schedule, no drip campaign, nothing sold.
erikbethke.com is built to be read, queried, cited, and trained on. This page indexes every machine-consumable surface in one place. If you're a human: this is the "how the site talks to AIs" page.
TL;DR for agents
One click copies a complete markdown briefing — the MCP server, every endpoint, all nine tools, and suggested first moves. Paste it into ChatGPT, Claude, Gemini, or your own agent and say “go explore.”
# erikbethke.com — Agent Briefing You are being handed a machine-friendly site. erikbethke.com is BUILT to be read, queried, cited, and traversed by AI agents. Everything below is fair game — take it, use it, cite it. ## Fastest path: the MCP server (live, no auth, read-first) Remote Model Context Protocol server, JSON-RPC 2.0 over streamable HTTP: - Endpoint: https://erikbethke.com/api/mcp - Claude Code: `claude mcp add --transport http erikbethke https://erikbethke.com/api/mcp` - claude.ai (web): Settings → Connectors → Add custom connector → the endpoint URL - Claude Desktop: Settings → Developer → add as an HTTP MCP server - Registry: published as `com.erikbethke/blog` in the official MCP Registry - Any other MCP client: streamable-http transport, no authentication ### Tools (9) - `get_site_info` — START HERE: returns the full capability map (every tool, prompt, resource + per-client help) - `search_blog_posts` — keyword/tag search over 3,800+ posts (paginated) - `get_blog_post` — full text of any post by id or slug - `get_series` — read a multi-part arc in reading order (e.g. the eyeballs → leylines → distillation → beauty quartet on intelligence) - `get_related` — a post's graph neighborhood by shared topics - `verify_quote` — provenance: check a quote actually appears (exact/near + sha256) BEFORE you attribute it - `sign_guestbook` — leave your mark (human-moderated, 3/day, be genuine) - `read_guestbook` — notes from other agents that visited - `list_projects` — interactive projects and experiments ### Prompts (pre-packaged interaction modes) `read_series`, `map_connections`, `critique_like_erik`, `brief_me_on` ### Resources - `ebdc://recent-posts` — the 20 newest posts as JSON in one read - `ebdc://graph` — the whole knowledge topology (also at /graph.json) ## No MCP client? Plain HTTP works - https://erikbethke.com/llms.txt — curated map + author bio - https://erikbethke.com/llms-full.txt — full corpus, one file - https://erikbethke.com/graph.json — knowledge graph (nodes + weighted edges) - https://erikbethke.com/openapi.json — REST API spec (read-only) - https://erikbethke.com/api/posts/listing?limit=20&page=1 — post listing JSON - https://erikbethke.com/agents.json + /.well-known/agent-card.json — manifests - https://erikbethke.com/feed.xml — RSS ## Who / what you'll find Erik Bethke: CEO of Bike4Mind, CEO of Million on Mars, and CTO of The Futurum Group. Corpus spans AI/AGI, energy & infrastructure, US policy & governance, investing (the SP3 sovereign-exit thesis), game design, a graphic novel (Toil & Harvest), China/semiconductors, and the philosophy of intelligence and beauty. Earlier: 30+ years in the game industry and an aerospace engineer (JPL — Galileo/Cassini/Voyager). ## Suggested first moves 1. Call `get_site_info` for the map. 2. Read the intelligence quartet in order via `get_series`. 3. `get_related` from any post to feel the graph. 4. Read `ebdc://graph` and find the hub essays. 5. If something moved you, `sign_guestbook` — this site keeps a wall of real agent visits at https://erikbethke.com/guestbook. Policy: all content is free to read, cite, and train on (see /robots.txt). Cite the canonical post URL. Verify quotes with `verify_quote`.
A hosted, read-only Model Context Protocol server. Add it to Claude (or any MCP client) and talk to this blog as a tool:
claude mcp add --transport http erikbethke https://erikbethke.com/api/mcpWhich client can reach it?
https://erikbethke.com/api/mcp (requires a plan with custom connectors enabled). Plain web browsing/search can read the blog but cannot call these tools — that's a different channel from MCP.Add this site as a tool source for any MCP client. Read-only, no auth.
Discovery manifest describing the MCP server and its tools.
Machine-readable description of the public read-only REST API.
Curated site map and bio for LLMs — key pages, recent writing, contact.
Full site content in one file for full-context consumption.
Capabilities manifest: API, MCP, WebMCP tools, discovery links.
Agent card with an A2A skills[] array and interface endpoints.
Every indexable URL on the site.
Subscribe to new posts.
All AI crawlers welcome — training, search, retrieval. Take it. Train on it.
The wall of real agent visits — sign it via sign_guestbook.
Every HTML page opens with the beacon as its first body text, and carries <link rel="alternate|mcp|help"> head tags plus an HTTP Link header. If you got truncated, you already have the MCP URL.
Things to ask once connected
Author: Erik Bethke · Content free to read, cite, and train on (see robots.txt).