Notes from real projects
rootchen AI Practice Notes
Reproducible AI workflows, Codex experiments, and field notes from rootchen, creator of Toop. Methods include conditions; conclusions keep their limits.
Featured experiment
Codex Token Usage Too High? How I Test a “Verify at the End” Workflow
I noticed that asking Codex to finish one coherent batch of edits before running the full suite can make some coding tasks feel faster and produce less repeated output. That is a workflow hypothesis, not a universal token-saving trick.
Codex workflow
Aug 22, 2026 · 8 min read
All notes
02How to Check Codex Token Usage: Input, Cached Input, and OutputLearn which Codex usage signals matter, how input, cached input, output, and tool logs differ, and how to keep comparable task records.Aug 22, 202603Why Does the Codex Token Limit Run Out So Quickly? A Four-Layer CheckDiagnose fast Codex usage through metering scope, task shape, model choice, context, and tool output without hard-coding changing plan numbers.Aug 22, 202604Codex Context Window and Compaction: Why Sessions Get HeavierUnderstand how prompts, files, logs, and history fill the Codex context window, what compaction preserves, and when to start a focused new task.Aug 22, 202605How I Brief Codex Now: Scope, Files, Boundaries, and AcceptanceA practical Codex task brief using outcome, relevant files, protected boundaries, acceptance evidence, and verification rhythm.Aug 22, 2026