Jellypod, Inc.

The Claude Code Changelog

TechnologyNews

Listen

All Episodes

Claude Code 2.1.110: No More Flicker, Better Focus, Safer Edits

This episode digs into the latest Claude Code 2.1.110 improvements, including the new fullscreen TUI that stops terminal flicker without restarting your session and the cleaner split between /focus and Ctrl+O.

It also covers practical upgrades like disabled auto-scroll, editor context injection, and important hardening changes that make long-running workflows and file handling more reliable.

This show was created with Jellypod, the AI Podcast Studio. Create your own podcast with Jellypod today.

Get Started

Is this your podcast and want to remove this banner? Click here.


Chapter 1

The moment Claude stops flickering

Lachlan Reed

Welcome to the show! I'm Lachlan Reed with James Turner, and mate, I wanna start with a tiny terminal annoyance that used to drive me properly up the wall: you're in a busy session, tool output starts pouring in, and the whole screen does this little partial redraw shimmy. In tmux panes, over SSH, especially when a command runs long, Claude Code used to sort of flicker and jitter like the terminal was breathing through a paper straw.

James Turner

[curious] That phrase -- "partial redraws" -- is exactly it. Not a crash, not even a bug you can screenshot cleanly. Just constant visual churn. And in 2.1.110, `/tui fullscreen` is the first version that fixes that without making you restart the session, right?

Lachlan Reed

[excited] Yeah, that's the bit. Before, going fullscreen could feel like opening a whole separate world. In 2.1.110, `/tui fullscreen` switches the CURRENT session into the full TUI renderer. Same conversation, same context, no fresh session, no "righto, let's start over." It just... settles down.

James Turner

Wait -- "current session" is the important token there. So if I'm mid-debug, halfway through a thread, I don't lose the conversation history just to get the cleaner renderer?

Lachlan Reed

Exactly. You keep the thread intact, but the UI stops doing those ugly partial refreshes. And for people living in cramped tmux layouts or remote SSH boxes, that matters more than it sounds. The first time the terminal stops twitching during long tool output, you notice it straight away.

James Turner

[reflective] That's a serious UX threshold. Because once the screen is stable, your brain stops treating it like a chat toy. It starts feeling closer to an editor -- not because of some giant new capability, but because the display finally behaves like it respects your attention.

Lachlan Reed

Spot on. It's one of those tiny changes that lands like a bigger one. Like tightening a loose bolt on a trail bike -- same bike, totally different ride. And if you want it to stick, there's a concrete hook now: the `tui` setting persists the mode. Drop `{ "tui": "fullscreen" }` into `settings.json`, and it'll come back that way across launches.

James Turner

[questioning tone] So if somebody tests `/tui fullscreen` once, closes Claude, reopens it, and says, "Huh, didn't stick" -- the missing piece is that JSON line: `{ "tui": "fullscreen" }`.

Lachlan Reed

Yep. Session switch first, config second. That's the whole trick. And I know this sounds almost too small to gush about, but once the flicker goes away, Claude Code feels less like a window that's trying to keep up and more like a proper working surface. Which, honestly, is a bigger deal than any splashy feature name.

Chapter 2

Why focus got split out of Ctrl+O

James Turner

[energetic] The other cleanup in 2.1.110 is pure interface hygiene: `/focus` is now the dedicated command to hide tool calls and show only Claude's final answer, while Ctrl+O just toggles between normal and verbose transcript. That split sounds tiny, but it's actually a separation of concerns.

Lachlan Reed

[skeptical] And by "separation of concerns," you mean two things that used to be bundled but were doing different jobs, yeah?

James Turner

Exactly. Focus mode is for READING. You want the clean final explanation, no tool chatter, no command-by-command trail. Verbose transcript is for INSPECTION. Something looks off, you hit Ctrl+O, and now you can see what Claude did step by step. Those are not the same mental mode.

Lachlan Reed

[laughs] That's so developer-brain. We jam two buttons into one, live with it for ages, then later go, "Ohhh... maybe that was a dog's breakfast."

James Turner

[chuckles] It is. But here's the concrete workflow: say Claude helps with a long refactor. Most of the time, you don't wanna read every tool call. You want the final explanation -- what changed, why, what to verify. `/focus` gives you that cleaner view. But if a file looks weird or a rename seems suspicious, Ctrl+O is the panic flashlight. Now you inspect the full trail.

Lachlan Reed

So let me try to explain it back -- slightly wrong on purpose. `/focus` is kind of like "hide the kitchen," and Ctrl+O is "show me the whole recipe card"? Or is that too cute?

James Turner

No, that's actually good. "Hide the kitchen" for `/focus` is dead on. And Ctrl+O is: show me every ingredient, every step, every mess on the counter. The old behavior bundled "I want less noise" with "I want more detail," which is... well, backwards if you say it out loud.

Lachlan Reed

[warmly] That's why this kind of split feels obvious only AFTER you've suffered through the old setup. Some devs will shrug -- "eh, it's just a shortcut change." But if you've ever bounced between wanting a readable answer and wanting the forensic trail, you feel it in your bones. Different jobs, different controls.

James Turner

And importantly, it's not dumbing anything down. You still have visibility into tool use. It's just no longer welded to the readability toggle. That means less accidental mode switching, less "wait, where did the transcript go," and a cleaner model in your head.

Chapter 3

The debugging payoff and the new sharp edges

Lachlan Reed

[curious] The feature I reckon a lot of folks will quietly love most is `autoScrollEnabled: false`. If Claude is running a long build or a chunky test suite, the terminal no longer yanks you to the bottom on every fresh line. You can keep your place, read older output, and the session keeps going underneath you.

James Turner

That `false` matters. Because anyone who's watched tests stream by knows the pain: you scroll up to the first failure, one new line lands, and boom -- you're snapped back to the bottom. It's like trying to read a book while someone keeps flicking the page forward.

Lachlan Reed

[laughs] Exactly. And paired with the steadier fullscreen TUI, it makes long-running output way less feral. There's also a neat `/config` option now: when you launch an external editor with Ctrl+G, Claude's last response can be injected as commented context right into the buffer.

James Turner

[excited] Right, and the specificity there is nice: vim, nvim, or VS Code can open with the relevant suggestion already sitting in the file as a comment block. So instead of alt-tabbing back to remember what Claude said, the guidance is literally in the editing surface.

Lachlan Reed

That's a lovely handoff. Very practical. But -- and here's the sharp edge -- Bash now enforces its documented maximum timeout. It doesn't just accept absurdly large values anymore. So if you've got prompts, scripts, or headless workflows that relied on giant or basically "infinite" timeouts, those may suddenly throw errors.

James Turner

[serious] "Documented maximum timeout" is the phrase people need to hear twice. Because if somebody set a cartoonishly high number months ago and forgot about it, 2.1.110 can expose that assumption fast. This is one of those reliability fixes that feels like a regression only if you were leaning on undefined behavior.

Lachlan Reed

Yeah, and there was another hardening pass too: "Open in editor" no longer blindly shell-interpolates filenames. That's important. If Claude suggests a file to open, the filename isn't just stuffed into a shell command raw anymore, which closes off a command-injection path.

James Turner

[matter-of-fact] That one is huge, even if it sounds boring. Convenience features become attack surfaces FAST. A filename looks harmless right up until somebody realizes it can smuggle shell syntax. So this release is doing two things at once: making the terminal feel smoother for humans, and less trusting in places where trust is dangerous.

Chapter 4

What to audit before you flip it on everywhere

James Turner

Before teams go all in, there are a few things worth auditing. First: `/tui fullscreen` is session-level unless you persist it. So if one developer tries it in a shell, loves it, then opens a new session and thinks it disappeared -- that's not a failed feature. That's just unsaved config.

Lachlan Reed

[responds quickly] Which comes back to that exact line in `settings.json`: `{ "tui": "fullscreen" }`. Without that, you're only changing the current session. I can absolutely see a whole team trying it once and going, "Nah, didn't work," when really they just didn't save the mode.

James Turner

Second, terminal compatibility still matters. Older macOS Terminal.app had garbled startup rendering in earlier versions because synchronized output wasn't handled well. 2.1.110 addresses that, which is good, but if fullscreen still looks off, it's smart to test in a modern terminal emulator before blaming Claude itself.

Lachlan Reed

That is such a classic terminal rabbit hole. You think the app is busted, then it turns out your emulator is the one dropping the ball. Not always, but enough that it's worth checking. Especially if the rendering weirdness shows up right at startup.

James Turner

And third: audit your Bash timeout assumptions now. Not later, now. If you've got automated scripts, SDK sessions, or long-running Bash tasks using values way above the documented ceiling, this release can surface that immediately.

Lachlan Reed

[reflective] I like that framing because it's not just "here's a shiny TUI update." It's also, "mate, check your plumbing." Stable fullscreen, cleaner focus modes, auto-scroll control, safer editor handoffs -- all good stuff. But the minute a tool gets more capable and more integrated, the rough edges matter more too.

James Turner

[softly] Yeah. And maybe that's the bigger question hanging over all of this. Once Claude Code has a stable fullscreen TUI, smarter reading modes, and editor handoffs that feel intentional instead of hacked together... does the terminal stop feeling like a prompt you visit, and start feeling like an actual working environment you inhabit?

Lachlan Reed

[warmly] That's a good one to leave rattling around in the toolbox. Thanks for listening, and we'll catch you next time.