<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Building AI Agents on Kevin Keller</title><link>https://kevinkeller.org/topics/building-ai-agents/</link><description>Recent content in Building AI Agents on Kevin Keller</description><generator>Hugo -- gohugo.io</generator><language>en</language><managingEditor>kellerkev@gmail.com (Kevin Keller)</managingEditor><webMaster>kellerkev@gmail.com (Kevin Keller)</webMaster><copyright>© 2026 Kevin Keller</copyright><lastBuildDate>Sat, 14 Nov 2026 08:00:00 +0000</lastBuildDate><atom:link href="https://kevinkeller.org/topics/building-ai-agents/index.xml" rel="self" type="application/rss+xml"/><item><title>Rixi: Sealed Sandboxes for AI Agents, and Any Project, on Machines You Control</title><link>https://kevinkeller.org/posts/rixi-cloud-native-ai-agent-execution/</link><pubDate>Sat, 14 Nov 2026 08:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/rixi-cloud-native-ai-agent-execution/</guid><description>Rixi ships a project and its exact environment to a remote machine, runs it, and streams the output back. It also gives AI agents sealed, throwaway sandboxes: one command creates a cloud box, egress is locked in the cloud firewall, model access is governed, and agents drive it over MCP or the Python and TypeScript SDKs.</description></item><item><title>Nanocortex Web Agent: An AI Knowledge Management Agent with Zettelkasten-Powered Retrieval</title><link>https://kevinkeller.org/posts/nanocortex-web-agent-production-ai-knowledge-management/</link><pubDate>Fri, 05 Jun 2026 00:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/nanocortex-web-agent-production-ai-knowledge-management/</guid><description>An AI knowledge management agent with a web interface, zettelkasten knowledge graph ingestion, document approval workflow, optional web search, and a skills system — powered by the Snowflake Cortex Agent API.</description></item><item><title>TurboQuant KV Cache — Running 128B Models on Consumer Hardware</title><link>https://kevinkeller.org/posts/turboquant-kv-cache-local-llm-consumer-hardware/</link><pubDate>Thu, 30 Apr 2026 10:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/turboquant-kv-cache-local-llm-consumer-hardware/</guid><description>KV cache is the memory wall that limits context length on consumer hardware. TurboQuant shrinks it 5x with minimal quality loss — here&amp;rsquo;s a ready-to-run build that packages llama.cpp with TurboQuant KV compression into a single conda install.</description></item><item><title>TurboQuant — What 6x Vector Compression Means for AI Agents</title><link>https://kevinkeller.org/posts/turboquant-vector-compression-ai-agents/</link><pubDate>Sat, 28 Mar 2026 10:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/turboquant-vector-compression-ai-agents/</guid><description>Google&amp;rsquo;s TurboQuant compresses embedding vectors to 3-4 bits with under 2% recall loss — no training required. Here&amp;rsquo;s why that matters for AI agent memory systems.</description></item><item><title>Zettelkasten Memory for AI Agents — Semantic Knowledge Graphs That Grow With Every Conversation</title><link>https://kevinkeller.org/posts/zettelkasten-memory-layer-ai-agents/</link><pubDate>Fri, 27 Mar 2026 10:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/zettelkasten-memory-layer-ai-agents/</guid><description>A pluggable semantic memory layer for AI agents inspired by the Zettelkasten method — auto-linking, importance scoring, and graph traversal across CrewAI, LangGraph, and Claude Code.</description></item><item><title>Automate Google Workspace with AI Coding Agents: Slides, Sheets, Docs, and More</title><link>https://kevinkeller.org/posts/google-productivity-skill-ai-coding-agents/</link><pubDate>Mon, 23 Mar 2026 09:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/google-productivity-skill-ai-coding-agents/</guid><description>An open-source skill that teaches any AI coding agent — Cortex Code, Claude Code, or others — to automate Google Slides, Sheets, Docs, Drive, and Forms using natural language prompts.</description></item><item><title>Postgres Agent Orchestrator: Coordinating AI Agents with pgmq, LISTEN/NOTIFY, and ltree</title><link>https://kevinkeller.org/posts/postgres-agent-orchestrator-pgmq-llm/</link><pubDate>Fri, 20 Mar 2026 14:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/postgres-agent-orchestrator-pgmq-llm/</guid><description>An open-source demo that coordinates multiple AI agents using only PostgreSQL — pgmq for task queuing, LISTEN/NOTIFY for event-driven coordination, ltree for lineage tracking. No Redis, no Kafka, no vector database.</description></item><item><title>Postgres Analyst: Natural Language SQL with Ollama — No Cloud Required</title><link>https://kevinkeller.org/posts/postgres-analyst-natural-language-sql-ollama/</link><pubDate>Fri, 20 Mar 2026 10:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/postgres-analyst-natural-language-sql-ollama/</guid><description>An open-source semantic layer that lets you query PostgreSQL in plain English using local LLM inference via Ollama. No API keys, no cloud — just your database and a local model.</description></item><item><title>Nanocortex: A Blueprint for Building Custom Snowflake Cortex Agents</title><link>https://kevinkeller.org/posts/nanocortex-blueprint-snowflake-cortex-agents/</link><pubDate>Mon, 16 Mar 2026 00:00:00 +0000</pubDate><author>kellerkev@gmail.com (Kevin Keller)</author><guid>https://kevinkeller.org/posts/nanocortex-blueprint-snowflake-cortex-agents/</guid><description>How a 1,600-line Python file can teach you everything about the Cortex Agent API</description></item></channel></rss>