Codex 0.153.0’s New Context Mode and TUI Upgrades

Ethan Park

So I, I was halfway through a massive codebase refactor yesterday, and, uh, right when you would expect the model to start forgetting early files, Codex zero point one fifty three point zero drops this new experimental context mode. Before we dive into what that actually changes under the hood, quick shout out to Jellypod for helping make this daily episode possible. But yeah, Maya, this zero point one fifty three update introduces features dot context underscore management dot experimental underscore mode, and it completely changes how long sessions handle memory.

Maya

It is about time! I mean, anyone who has used Codex for more than twenty minutes knows the pain of turn compaction. You are deep in a debugging loop, and suddenly the engine just, um, aggressively truncates your early instructions. But with this flag enabled, instead of just chopping off old messages, it uses token budget context allocation and structured history notes.

Ethan Park

Right, and the really neat piece is the model gets access to a brand new tool called new underscore context. So when the context window starts getting crowded, the agent itself can decide to prune stale conversation branches while explicitly preserving your critical session context and history notes. It is not just dropping tokens blindly based on age.

Maya

Which sounds fantastic on paper, but, okay, we, we have to talk about who actually gets to use this right now. If you are on ChatGPT Plus, Pro, or Pro Lite using the native Codex backend, you can turn it on. But if you are using API key sessions, custom provider endpoints, or temporary structured threads, you are completely excluded from this experimental flag.

Ethan Park

I get why OpenAI limits experimental rollouts initially to their own managed infrastructure, though. From a product reliability standpoint, controlling the backend context management pipeline lets them tune the token budgets without worrying about third party proxy weirdness. And honestly, stopping those mid refactor hallucinations when context overflows is a massive win for heavy CLI users.

Maya

Sure, Ethan, but think about enterprise teams! Most large corporate engineering groups run Codex CLI routed through private API gateways or custom internal endpoints for security and logging. Those are precisely the developers doing huge, multi hour refactoring tasks who need token budgeting the most, and they are the exact ones walled off from this flag!

Ethan Park

That is a fair point, but I bet it is just an early stage constraint while they stabilize history note compaction. The architecture itself clearly points toward broader support later.

Maya

Well, while we wait for that expansion, let us look at how you actually turn it on if you are eligible. In your dot codex slash config dot toml file, you add bracket features dot context underscore management bracket, and then set experimental underscore mode equals true. Simple enough. Though speaking of TOML configurations, zero point one fifty three point zero also cleaned up some TUI settings, right?

Ethan Park

Yeah, they moved disable underscore paste underscore burst under the TUI header, so now it lives under bracket tui bracket instead of being a top level setting. Though the old location still works as a fallback so your build will not break. And honestly, the terminal UI polish in this release is stacked. The plugin CLI now supports remote marketplaces directly from your shell.

Maya

Oh, I loved seeing that! Being able to run codex plugin list, codex plugin install, or remove directly from the command line without manually editing local JSON manifest files is huge for developer workflow. It, it finally feels like a real package manager inside the tool.

Ethan Park

Exactly! And for Vim mode power users in the terminal composer, undo with u and redo with Control R now work smoothly across drafts, even preserving pasted content and file attachments. Plus, if you find automatic turn recaps distracting, you can now set tui dot auto underscore recap equals false in your config.

Maya

Wait, so if you turn off auto recap, you can still run slash recap manually when you actually want a summary, right?

Ethan Park

Precisely. Manual slash recap stays available whenever you need to catch up, but your terminal stays clean during fast, iterative edits.

Maya

See, I tested that setting during a multi file edit session, and skipping that auto recap step makes the terminal feel noticeably snappier between turns. You are not waiting for the TUI to format a summary block before you can type your next prompt. It just get out of your way.

Ethan Park

And on top of that speed feel, they fixed TUI reconnection resilience. If your external app server connection drops mid session, it automatically reconnects, preserves your current draft and full transcript, and keeps any queued or uncertain commands safely paused for your review instead of firing them off blindly.

Maya

That safety buffer is huge. Losing a draft or having a command execute blindly after a brief network hiccup used to be so nerve wracking. Zero point one fifty three point zero really balances deep context features with essential reliability tweaks.

Ethan Park

It really does. That is zero point one fifty three point zero in a nutshell. Give experimental context mode a shot if you are on a supported plan, and we will talk to you next time!