MAX-MAXXINGby HPRFX

Your next hyperfixation

technologyestablishedDocumented

CODEX-MAXXING

The practice of using Codex as a persistent workspace for long-running, multi-step work rather than as a one-prompt code generator.

KEEP THE WORK ALIVE.

Put it on my phoneiPhone Lock Screen · PNG downloadTrack it for 30 daysPrintable A4 calendar

What is CODEX-MAXXING?

Codex-maxxing treats the agent as an environment rather than a single exchange. Context is preserved outside one prompt. Work is decomposed into plans, drafts, artifacts and next actions. Long-running threads, external memory, repeated checks and parallel tasks allow a project to continue after the first request has ended. The term is unusual because it moved from practitioner language into official OpenAI product language.

Supposedly maxes

PERSISTENT CONTEXT / LONG-RUNNING WORK / AGENTIC CONTINUITY / ARTIFACT PRODUCTION

Adherents believe

  1. The useful unit is not the prompt but the project.

  2. Context should survive the conversation that created it.

  3. A good agent leaves behind plans, files and next actions.

  4. Long work requires external memory and verification.

Typical practices

  • PERSISTENT WORKSPACES
  • EXTERNAL PROJECT MEMORY
  • PARALLEL TASKS
  • AUTOMATED CHECKS
  • ARTIFACT HANDOFFS

Spotted in the wild

OpenAICodex-maxxing for long-running work

Official OpenAI use of the term for persistent workspaces, preserved context and complex workflows extending beyond one prompt.

OpenAICodex-maxxing for long-running work — white paper

Detailed guide describing persistent context, project memory, artifact creation and long-running work with Codex.

Not your hyperfixation?