Guides: using AI well, Claude Code and Codex
Sourced guides on using AI well. Some cover what the research says about AI and your own thinking; others compare Claude Code and Codex and explain their setup files. Every guide links its sources and says when it was last checked.
Using AI without rotting your brain
Does AI rot your brain? What 8 studies foundMIT, Microsoft, PNAS, Anthropic and Lancet studies on AI and thinking: what each found, how big the effect was, and what none of them show.
How to use AI without getting dumberSix habits for using ChatGPT, Claude or any AI while keeping your own thinking, each tied to a study, plus a ten minute daily routine.
Cognitive debt in AI coding: the evidenceWhat cognitive debt means when an agent writes your code, what Anthropic's 52 developer trial measured, and five habits that keep a codebase yours.
Claude Code, Codex and your setup
Is Claude Code free? Is Codex free?Claude Code needs a paid Claude plan. Codex has limited free access on ChatGPT Free, more on Plus and Pro. Sourced, checked 29 September 2026.
Did OpenAI drop ChatGPT's 5x and 20x plan names?OpenAI now names its ChatGPT Pro plans Pro 100, Pro 200 and Pro 500 and publishes no multiplier for any of them. Checked 30 September 2026.
Claude Code vs Codex: which should you use?Claude Code costs $20 to $200 a month with no free tier. Codex has a free tier and ChatGPT plans up to $500. Sourced, checked 29 September 2026.
CLAUDE.md vs AGENTS.md: the differenceCLAUDE.md is Claude Code's own file. AGENTS.md is the shared format Codex reads. What each does, sourced, checked 29 September 2026.
How to write a CLAUDE.md or AGENTS.md, with a templateWhat to include in a CLAUDE.md or AGENTS.md, what to leave out, and the size limit each tool enforces. Checked 29 September 2026.
Claude Code hooks explainedWhat Claude Code hooks are, the main events (PreToolUse, PostToolUse, Stop) and what each can block. Sourced, checked 29 September 2026.
What an AI daily planner needs to doWhat a daily AI planner actually needs to do beyond writing a to do list, and where a free agent file setup stops short.
Does an AI accountability partner work?What an AI accountability partner needs to do to actually change behavior, not just log it. Reddit-evidenced, 29 September 2026.
Recently added
Why Claude Code forgets the plan after compactionCompaction swaps the chat for a summary, and only what is on disk comes back in full. What survives, and four ways to keep the plan. Read 2 October 2026.Claude Code memory: what carries over between sessionsClaude Code starts every session fresh. CLAUDE.md, which you write, and auto memory, which it writes, are what carry over. Read 2 October 2026.Google Gemini Spark price: AI Ultra at $100 a monthGoogle Gemini Spark is on Google AI Ultra at $100 a month per The Wall Street Journal, and on Google AI Pro, whose price the sources do not give.Grok Bot price: no standalone plan, from $20 a monthGrok Bot has no standalone plan; it comes with SuperGrok, Cursor Pro and Cursor Teams plans, and CellCog lists the cheapest route at $20 a month.Claude Code plan mode: how to turn it on and use itClaude Code plan mode lets Claude read and propose changes without editing; press Shift+Tab or run claude --permission-mode plan to start it.Claude Code process exited with code 1: what to checkWhat Claude Code process exited with code 1 means, which program prints it, and the checks the Anthropic docs give, as read on 6 October 2026.Claude skills vs commands vs subagents vs plugins explainedSkills, commands, subagents and plugins in Claude Code: what each is, where its files live, and when to use which, per docs read 6 October 2026.Claude Code drifts from the plan: why it happensWhy Claude Code drifts from the plan, from Anthropic's memory documentation read on 7 October 2026, and what to check when it does.CLAUDE.md vs memory in Claude Code: two systemsCLAUDE.md holds instructions you write and auto memory holds notes Claude writes; Anthropic's docs list both as Claude Code memory.
Mover OS: $49 once, plain files on your own machine, inside the AI you already use.