CODECK
Get Codeck
← Back to the blog
Max Burleigh

Creator of Codeck

Max Burleigh

Max Burleigh is the creator of Codeck, an independent macOS app for switching between Codex and Claude Code tasks. He writes about the tools, workflows, and verification practices behind building with AI.

Codeck is independently developed and is not affiliated with OpenAI or Anthropic. Articles distinguish runnable demonstrations from conceptual examples and link to the primary references behind their technical claims.

Editorial approach · Contact & corrections

THE READING LIST

Articles by Max

Three translucent evaluation panels converging into a single lens, illustrating the components of a coding-agent benchmark.

Workflows · 7 min read

Artificial Analysis Coding Agent Index v2: what the new scores mean

Artificial Analysis has changed its coding-agent benchmark. Understand the new scoring, API cost tradeoffs and a practical way to evaluate agents on your own repository.

Read the guide
A patched PHP 8.4.27 command line sits above a separate web worker still running upstream 8.4.26, circled with the note Check the web worker.

Security · 10 min read

Two critical PHP flaws: what to patch and how to verify it

PHP’s October 21 emergency releases fix CVE-2026-109618 and CVE-2026-109619. Check affected runtimes, vendor backports, and the workers actually serving your site.

Read the guide
An illustrated task to preserve zero moves through an action and test, then the failing result returns to the next action.

Foundations · 7 min read

How AI coding agents actually work

Follow a coding task from instructions to tool calls, feedback, and a verified result. Understand where the model ends and the software around it begins.

Read the guide
Scattered notes converge into a structured working brief with three clear rows: Goal, Constraints and Checks.

Workflows · 7 min read

Give your AI coding agent better context

Turn a vague request into a useful working brief. Supply the right constraints, examples, and evidence without burying the task in irrelevant material.

Read the guide
One shared contract splits into Library UI and Article content tasks, then both paths rejoin for integration and verification.

Workflows · 7 min read

When to use one AI agent—and when to split the work

Parallel work helps when the boundaries are clear. Use dependencies, shared files, and review cost to decide whether another agent will actually help.

Read the guide
Shared Git history connects three separate folders: project on main, docs-task on codex/docs, and checks-task on codex/checks.

Workflows · 8 min read

Git worktrees for parallel AI coding

Give each task its own working directory. A complete local walkthrough covers creation, inspection, integration, and safe cleanup.

Read the guide
A client sends count_words the text Tools return evidence; the server returns the count three along the opposite arrow.

Foundations · 8 min read

MCP explained through a real tool connection

Trace an actual local client/server exchange: initialize, discover a tool, call it, and inspect the response. No model account or API key required.

Read the guide