Picking the Right AI Tool

Different tools for different jobs. A browser chat drafts fast. A research notebook works from your documents. A terminal coder touches files and runs commands. Pick the lightest one that fits — move up when the task grows.

Tier 1: Browser chats for quick tasks

No install, no code. ChatGPT and Claude in a tab:

• Drafting emails and announcements.
• Brainstorming angles for a piece.
• Proofreading and tone fixes.
• Broad questions where precision isn’t critical.

Fast and handy. But they don’t know your private files or the latest facts unless you paste them in or connect a source.

Research tools that cite your documents

When answers must come from specific material, use something that takes uploads and shows citations — e.g. Google NotebookLM for reports, papers, transcripts, lecture notes.

Citations are pointers, not proof. Open the cited page and confirm it says what you plan to rely on.

Tier 2: Coding help inside your editor

In VS Code or Cursor, an inline assistant speeds up small changes: autocomplete as you type, side-panel chat over a file you point at. Good for learning syntax and routine edits. You pick the files, review the diff, run the checks.

Tier 3: Terminal coders that cross files

When the task spans files and commands, a terminal assistant works across the project with your permission: reads folders, edits, runs tests, shows errors.

• Oh My Pi (OMP) on Pi — my first reach, and what built this site.
• OpenCode — open-source, terminal/IDE/desktop, connects to many providers.
• Claude Code and Codex CLI — Anthropic’s and OpenAI’s command-line agents.
• Hermes — background tasks and monitoring rather than one focused session.

The short checklist

• Quick question or draft? Browser chat.
• Answer must come from your documents? Research tool with visible sources.
• Multiple files or checks? Terminal coder.
• Runs while you’re away? Background tool.

Simplest first. Move up a tier when the task demands it.