← All posts

Tag

claude code

4 posts

A physician-developer choosing among a terminal, an agent-first editor, a cloud workspace, and a persistent messaging agent
AI Tools

The Harness Is the Workplace: Choosing Where an AI Agent Works

Claude Code, Codex, Antigravity, OpenClaw, Hermes, and Omarchy illustrate a durable lesson for physicians: choose the agent's workplace before delegating the work.

· 12 min read
AI AgentsAgent HarnessClaude Code
Four distinct agent workspaces surrounding the same language model, each with different tools, permissions, memory, and review artifacts
AI in Medicine

Same Model, Different System

Claude Code, Google Antigravity, OpenAI Codex, and DeepSeek Harness expose the architecture around an AI model. Physicians should compare that architecture before comparing model rankings.

· 13 min read
agent harnessclaude codegoogle antigravity
Physician-builder at a laptop with a local clinical knowledge graph, binders, stethoscope, and private local server
Technology Featured

The Physician's Stack: Obsidian + Claude Code + Local LLM

How to build a clinical knowledge vault without sending patient data to anyone's server. The two-vaults architecture, ingestion workflows, and the HIPAA-first policy.

· 12 min read
second brainObsidianClaude Code
A physician-developer working at night with an AI coding assistant open beside a software project
AI in Medicine

AI-Assisted Coding: Your Second Pair of Hands

Claude Code, Cursor, and Copilot do not replace judgment. They compress the feedback loop for physician-developers who still read every line before they ship.

· 13 min read
ai codingclaude codecursor