// FIELD NOTES
Guides from the
workbench.
In-depth articles on Claude Code, workflows, and building with Claude. Written from real projects, not documentation.
The Complete CLAUDE.md Guide: Build Context That Survives Real Projects
A field-tested guide to writing CLAUDE.md files that actually work. Covers structure, content, common mistakes, and how to iterate as your project grows.
Mini Claude Code · Episode 02: Teaching the Agent to Use Tools
Episode 2 gives our REPL its first three tools — read_file, list_dir, run_bash — and walks through the tool-use loop with the Anthropic SDK. We handle stop reasons, tool errors, and the surprisingly tricky part of what to put back in the conversation.
Mini Claude Code · Episode 03: Editing Files Without Wrecking the Repo
Episode 3 adds the apply_patch tool — the one where most agents quietly destroy your work. We build a strict unified-diff applier with dry-run preview, workspace guardrails, and a confirmation gate that survives a Claude in a bad mood.
Context Engineering for Coding Agents: A Deep Read of the 2025–2026 Research
A practitioner's synthesis of the frontier work on context management for coding agents—context rot, lost-in-the-middle, sub-agent isolation, prompt caching, and the patterns that actually survive production.
Mini Claude Code · Episode 01: A 40-Line REPL That Actually Remembers
Start of a hands-on series where we build a working coding agent from scratch in TypeScript. Episode 1 gets the smallest possible loop running — stdin, streaming, and a conversation that survives turns.
Building a Rationale Dataset with Claude: A Practical Playbook
How to collect, structure, and filter Claude's reasoning traces into a rationale dataset you can actually use to distill smaller, faster models.
Distilling Reasoning: What Focused Distillation from Explanations Means for Builders
A practitioner's read of the FDE (Focused Distillation from Explanations) line of research—Distilling Step-by-Step, CoT distillation, and what it means when you're shipping with Claude.
When Distilled Models Lie: The Rationale Faithfulness Problem
Student models often produce rationales that sound right but don't reflect their real computation. Here's what faithfulness means, how it fails, and how to ship anyway.
When to Distill Claude: A Decision Framework for Cost-Sensitive Teams
A practitioner's guide to deciding when distilling Claude into a smaller model is worth it, when it isn't, and how to route between them in production.
Claude Code vs Cursor: An Honest Comparison for 2025
A detailed, no-hype comparison of Claude Code and Cursor. Different philosophies, different strengths. Here's how to choose—and when to use both.
Claude Code Hooks: Complete Setup Guide with Real Examples
Learn how to use Claude Code hooks to automate quality checks, trigger tests, and extend Claude Code's behavior with shell scripts. Includes working examples.