AI FOR BABIES No hype. Pure signal.
GitHub

AI FOR BABIES // AGENT SEMANTIC INTERFACE

Autonomous LLM Agent Endpoint. Direct text representations and callable tool specs.

Full documentation: /llms.txt | /llms-full.txt | WebMCP Spec: /.well-known/webmcp.json

Engineering Reference Standard

AI Explained For People With Actual Jobs.

No corporate hype. No fake benchmarks. Just high-contrast technical architecture covering 77 core concepts, AGENTS.md multi-agent specs, 100k tool memory paging, speculative decoding math, and WebMCP browser protocols.

Explore 77 AI Terms AGENTS.md Generator Speculative Lab Read llms-full.txt
77
Categorized Terms
8
Deep Core Modules
4
WebMCP Tools
0%
Corporate Fluff
Module 01 // Reference Glossary

77 Essential AI Terms

Exactly 77 fundamental concepts across 7 core engineering categories (11 terms each). Filter by category or search in real time.

Showing 77 of 77 Terms Clear Filters
Module 02 // Prompt Engineering

Interactive Prompt Engine

Diagnose prompt weaknesses against 8 strict production rules: negative constraints, few-shot examples, JSON schemas, role clarity, context placement, stop sequences, boundary rules, and token efficiency.

Ready. Enter a prompt and click 'Run Diagnostic' to analyze.
Module 03 // Retrieval Architecture

RAG Masterclass: Foundations to Production

Stop dumping unranked PDFs into context windows. Learn the 3 pillars of industrial RAG: Hybrid BM25+Dense retrieval, Cross-Encoder reranking, and chunking strategy tradeoffs.

Interactive RAG Pipeline Simulator

Simulate how raw queries flow through hybrid retrieval (Dense + Sparse), Reciprocal Rank Fusion, and Cross-Encoder rerankers.

Click 'Simulate Retrieval' to inspect step-by-step ranking logs.
Module 04 // Open Multi-Agent Standard

AGENTS.md Standard & Multi-Agent Teams

The neutral Linux Foundation standard for autonomous agent governance. 6 standardized frontmatter fields defining commands, tests, styles, boundaries (Always / Ask / Never), and multi-agent handoffs.

Interactive AGENTS.md Generator

Click 'Generate AGENTS.md' to create standard specification.
Module 05 // Tool Scaling Architecture

Tool Scaling & 100k Tools Architecture

Why dumping 100 JSON schemas into system prompts ruins agent accuracy. Implement skill paging, plain markdown memory on shared filesystems, and avoid proactive agent traps.

Interactive Tool Paging Simulator

Test how the agent dynamically pages tools from an indexed catalogue of 10,000 skills into a restricted 3-slot active RAM context.

Click 'Simulate Skill Paging' to view dynamic RAM page in/out execution.
Module 06 // Inference Optimization

Speculative Decoding & Speed Lab

Bypass the 140GB DRAM memory bandwidth bottleneck. Calculate speedups from draft verification, acceptance rate alpha, and token throughput math.

Inference Speedup Calculator

Click 'Compute Speedup' to run latency equations.
Module 07 // Environment & Evaluation

Agent Environments & Gym

Synthetic task generation, Harbour evaluation specifications, and continuous test environments created from agent execution traces.

Agent Gym Simulator

Run a synthetic environment evaluation loop: generate task spec, execute agent in sandbox container, and assert ground-truth test assertions.

Click 'Run Gym Trial' to execute synthetic environment test.
Module 08 // Browser Agent Protocol

WebMCP Standard & Technical Integration

Standardizing client-side browser tools for web agents via /.well-known/webmcp.json and window.modelContext. Lossless JSON interaction without DOM scraping.

Live WebMCP Scanner & Tool Caller

This page implements live WebMCP tools on window.modelContext.tools. Test direct client-side tool execution right now.

Select a tool, adjust JSON arguments, and execute.
Module 09 // Knowledge Processing

Marketing Intelligence Layer

Transform personal operating experience into proprietary AI assets. The 4-stage pipeline: Inbox /raw, Extractor, Topic Router, and Confidentiality Sanitizer.

Interactive Sanitizer & Knowledge Router

Test how confidential client details, revenue metrics, and proprietary names are stripped while preserving the underlying engineering insights.

Click 'Sanitize & Extract Insight' to preview confidential data removal.
Module 10 // AI Specification Framework

AI Spec Writing / PRD Framework

Bridge high-level product vision and AI agent execution. 6 core specification pillars: Commands, Testing, Structure, Style, Git, and Boundaries (Always / Ask / Never).

Interactive PRD Spec Builder

Click 'Build PRD Specification' to generate structured artifact.
[ PRIVATE PROTOCOL ]

THE ENGINEERING DISPATCH

Weekly deep dives into production multi-agent architectures, speculative decoding speedups, and prompt optimization benchmarks. Zero fluff. Pure signal.