AgentStack
The daily AI show and the systems behind it: OpenClaw automations, fitness-data connectors, local models, hardware, evaluations, and honest build notes.
Listen
AgentStack Daily
Daily episodes on models, hardware, evaluations, security, and the systems being built around them.
Browse episodesBuild
Fitness connector brief
An agent-ready architecture for Garmin, WHOOP, Speediance, Cronometer, 8Sleep, and normalized reports.
Read the implementation briefField report
AI fitness assistant
How the fitness-data assistant was assembled, what it automates, and where the hard edges still are.
Read the build storyWatch
AgentStack videos
Build walkthroughs, model tests, local hardware, failures, and the systems behind the daily show.
Browse videosImplementation notes
OpenClaw and AgentStack field reports
The durable written layer behind the daily show: architectures, failure modes, connector decisions, and working systems.
Build Logs
LLMs Don't Edit Videos: Why the Agent Harness and Tools Matter
Your LLM isn't editing video; tools like FFmpeg are. Understanding the difference between the model, the harness, and the tools is key to building AI agents.
Build Logs
Switching From OpenClaw to Hermes Agents: My Current AI Workflow
Learn how to split Project Aria, Agent Stack Daily, and adaptive workouts across Hermes agents on an M3 and a DGX Spark, using Fable for major architectural changes.
Build Logs
Building BJJ Buddy Live: A Speediance Workout and AI Dev Session
A real Sunday session: shipping Supabase auth to BJJ Buddy, debugging GLM 4.7 offload issues, and grinding Speediance sets while reviewing Whoop and Garmin data.
Build Logs
OpenClaw Fitness Stack: Three Macs, BJJ Buddy, and My Daily Podcast
An inside look at my fitness automation setup—and how I downsized from three Macs and OpenClaw to a streamlined workflow powered by Codex, Claude Code, and Antigravity.
Reviews
Which AI Coding Tool Actually Delivers? Live Showdown Results
Testing Hermes Desktop, Claude Code, Codex, MiniMax, OpenClaw, and Antigravity head-to-head. Here is my definitive guide on which AI harness is worth your money.
Build Logs
OpenClaw Finally Made My Fitness Data Useful
OpenClaw is usually shown as a content machine. My version is less flashy: a local agent system that turns scattered fitness data into daily training decisions.
Use the podcast for the moving story and the field reports for the durable one
Daily episodes track what changed. Written reports preserve the architecture, evidence, and implementation details worth finding later.