CodexHow Blog
Longer, more opinionated writing than the how-to guides or reference pages allow — how this site sources and re-checks its own numbers, the patterns behind specific pricing quirks, and honest write-ups of what CodexHow could not confirm and why that's worth saying out loud rather than hiding.
Start with "how we source every number on this site" if you're wondering why this site reads the way it does — it's the closest thing here to a mission statement. The rest range from close readings of a single pricing quirk to broader arguments about how a comparison tool should behave when neither side deserves a declared winner.
Why Codex CLI Defaults to the General Model
The 272K Cliff and Agentic Workload Planning
What Tier 2 to Tier 3 Actually Buys You
Batch and Flex Cost the Same. Why Both Exist?
Reading a Usage Block Like an Engineer
How We Source Every Number on This Site
Building This Site: How We Keep a Diff Honest
What We Could Not Confirm, and Why That Matters
A Running Table of Codex Pricing Changes
The Cache-Write Tax Nobody Budgets For
Sandboxing Codex for a CI Pipeline, No Surprises
Why Approvals and Sandbox Are Separate Settings
What Changed When Priority Became Fast Mode
Why a Cross-Provider Tool Can't Declare a Winner
Reading OpenAI's Model Pages Like an Engineer