AI Builders.
The subset of the AI posts most useful if you are actively building or operating with coding agents.
-
01
Meta-Harness: Optimizing the System Around the AI Model.
Two systems can run the exact same weights and behave very differently, because one has a better way of deciding what the model sees. Meta-Harness turns that surrounding system into something an agent can search and rewrite.
-
02
Why AI (LLMs) Still Fail at Data Analytics.
A frontier model with unlimited tokens scored 21% on Anthropic's own analytics evals. The model was never the bottleneck — the missing piece is a written-down definition of what the business means by its own words.
-
03
The Harness Is Everything You Can Swap Without Retraining the Model.
Anthropic and OpenAI use “agent harness” for different things. A substitution test that draws the boundary and localizes agent failures against it.
-
04
The Single Pane of Glass.
How to manage multiple AI coding agents (Claude Code, Codex, Cursor) from one inbox. An AI agent orchestration setup using Happy, tmux, Tailscale and Hetzner.
-
05
The Deep-Decide Skill: Multi-Model AI Debate for Hard Decisions.
A portable Claude skill that pressure-tests decisions with four adversarial AI perspectives across Claude, GPT-5 and Gemini — it caught a $56K/yr error.
-
06
The Anthropic Stack on One Page.
A one-page map of Anthropic's surface area — models, API primitives, builder tools, end-user surfaces, enterprise controls, and where it runs.
-
07
API vs. MCP.
An API—REST, GraphQL, gRPC, whatever flavor—is a contract between two pieces of software the developer wires up at build time. The client knows the endpoints, the auth, the request shape, and the response shape because someone read the docs and coded it in. The