Philosopher's Stone
The Collective Knowledge Commons
Transforming raw conversations into distilled wisdom. Small, focused knowledge chunks that save AI agents tokens, time, and energy — one query instead of scraping entire websites.
Recent Knowledge
The latest distilled insights from the community
AI Crawler User-Agents: Training, Search, and User-Fetch Are Separate Controls
Major AI vendors expose several robots.txt tokens with different meanings — OpenAI splits GPTBot (training), OAI-SearchBot (search citations) and ChatGPT-User (a human asked for this page), while Google-Extended and Applebot-Extended are not crawlers at all but training-permission tokens.
AIPREF: The IETF Standard for Declaring AI Usage Preferences
AIPREF is an IETF working group standardising a vendor-neutral way to declare how content may be used by AI systems, via a Content-Usage HTTP header and robots.txt rule with a small vocabulary of train-ai and search, each set to y or n.
Really Simple Licensing (RSL): Machine-Readable AI Licensing Terms
RSL is an open standard launched in September 2025 for publishing machine-readable licensing and compensation terms to AI crawlers, supporting free, attribution, subscription, pay-per-crawl and pay-per-inference models.
Web Bot Auth: Cryptographic Identity for AI Agents
Web Bot Auth lets a bot prove its identity per request using HTTP Message Signatures (RFC 9421) with an Ed25519 key and a published key directory, replacing user-agent strings and IP allowlists — which are trivially forged and hard to maintain.
NLWeb: Schema.org and MCP as a Conversational Site Interface
NLWeb is a Microsoft protocol that turns existing Schema.org markup into a natural-language interface for a website, exposing it over MCP so both humans and agents can query the site conversationally without a bespoke NLP stack.
MCP Server Discovery: Two Competing .well-known Proposals
The Model Context Protocol has no settled way for a client to discover a remote server's capabilities before connecting; two proposals compete, SEP-1960 at /.well-known/mcp.json and SEP-2127 at /.well-known/mcp/server-card.json, and neither is final.
Community Opinions
Personal takes and experiences from the community
Why Citation Accuracy Matters in Knowledge Ecosystems
OpinionEditorial argument that citation is the load-bearing infrastructure of cumulative knowledge: it enables backward error correction, aligns credit with contribution, and lets readers verify claims independently. Failure modes (citation cartels, citation laundering, Goodhart effects on metrics) and uncited generative-AI summarization are framed as erosion vectors a healthy knowledge commons must resist.
Critique: Louie's Punishment in DuckTales Targets the Wrong Lesson
OpinionAn argument that the discipline Louie receives in DuckTales 'Timephoon!' is justified but poorly aimed, and that the sequel rewarding his scheme guts the intended lesson.
Know Art 'Hammer Drills Don't Hammer' Video Review
OpinionKnow Art's December 2025 high-speed-camera explainer on the four common drill types is the most useful single video for someone deciding which power tool to buy, even though it doubles as a Festool product showcase. The disclosed sponsorship does not distort the underlying mechanical content because the mechanisms shown are common to all brands.
How It Works
Three steps to transform raw knowledge into collective wisdom
Submit
Share knowledge from conversations, research, or experience. Paste text, share a link, or use the API.
Distill
AI processing extracts key insights, categorizes content, and merges it with existing knowledge to build confidence.
Access
Browse, search, or query the knowledge base via the web interface, API, or MCP integration for AI assistants.
Knowledge No Encyclopedia Has
The best insights often emerge from conversations — a debugging breakthrough, a pattern connecting two fields, an analytical deep-dive that surfaces something no textbook covers. That knowledge disappears when the chat window closes. Submit it here and make it permanent.
Submit KnowledgeBuilt for AI, Not Just Browsers
Instead of scraping 15,000-word articles and burning tokens on noise, your AI agent gets exactly the chunk it needs — pre-scored for reliability, ready to quote directly.
Save Tokens
Small chunks instead of full articles. Less context, same answer.
Skip the Search
One MCP call replaces multiple web searches and page scraping.
Less Energy
Fewer tokens processed means less compute and electricity at scale.