Claude Code /btw and /voice changed how I talk to my terminal

I used to sit there staring at my terminal, watching Claude Code work through a task, holding a question I wanted to ask. Not a big question — something like "hey, what pattern is this project using for error handling?" or "remind me what that config flag does." But Claude was mid-task, and I didn't want to interrupt it. So I'd wait. Or I'd open a second terminal, start a new session, lose all the context, and ask there.

It's a small frustration. The kind you don't even name. You just live with it — until someone fixes it and you realize how much it was slowing you down.

Two features showed up in Claude Code recently that addressed exactly this friction: /btw for side questions and /voice for built-in voice input. I've been using both. One of them already feels like a clear win. The other? I'm still figuring out.

The Frustration I Didn't Know Had a Name

Here's the pattern. Claude Code is generating tests, refactoring a component, or writing a migration. It's doing real work — the kind that takes 30 seconds to a couple of minutes. And in that window, a thought pops up. Maybe I want to clarify something about the codebase. Maybe I want to ask what approach Claude is planning to take next. Maybe I just want to understand a concept that came up in the output.

Before /btw, I had three options. Wait until Claude finishes — and by then I've either forgotten the question or the moment has passed. Interrupt the task — which feels wasteful and breaks the flow. Or open a separate session — which means losing all the context Claude already has about the project.

None of these are terrible. But the cumulative effect is real. It's like working with a colleague who can only handle one conversation thread at a time. You start filtering your own thoughts, holding back questions, batching things up. It changes how you interact.

How Claude Code's /btw Command Solves Side Questions

/btw (by the way) is a slash command in Claude Code that lets you ask a side question while Claude is still working on your main task. You type /btw what does this config flag do? and get an answer in a dismissible overlay — without interrupting whatever Claude is doing.

The design is surprisingly thoughtful. According to Anthropic's Claude Code documentation, /btw sees your full conversation context but has no tools — it can't read files, run commands, or search. It's the inverse of a subagent: a subagent has full tools but starts with empty context. /btw has full context but no tools.

What I like about this tradeoff: it means the answer comes back fast. There's no file reading, no command execution, just Claude answering from what it already knows about your session. The question and answer are ephemeral — they show up in an overlay you dismiss with Space, Enter, or Escape, and they never enter the conversation history. No context bloat.

I started using /btw for things like:

• "What test framework is this project using?" (while Claude writes tests)
• "Why did you choose that approach?" (while Claude is mid-refactor)
• "What's the difference between these two patterns?" (while reviewing Claude's output)

It's a small feature. But it removed that "I want to ask but I need to wait" friction entirely. For me, that's enough to call it a win.

/voice in Claude Code: My Superwhisper Dilemma

/voice is Claude Code's built-in speech-to-text feature that lets you speak prompts instead of typing them. You toggle it on with /voice, then hold the spacebar to talk and release to send the transcription to Claude. It was announced in early March 2026 and is rolling out gradually to Pro, Max, Team, and Enterprise subscribers at no extra cost.

Here's my situation. For the past couple of months, I've been using Superwhisper — a macOS voice-to-text app that works system-wide. You talk, it transcribes, and it pastes the text into whatever app you're using. It costs $8.49 per month. And honestly? It works really well. The accuracy is solid, it handles technical terms, it works offline on Apple Silicon, and I've built it into my daily workflow.

Then Claude Code added /voice. And now I have a question I didn't expect to be asking: do I keep paying for Superwhisper?

Here's how I see my three options:

Option	Cost	Scope	Accuracy	Setup
Superwhisper only	$8.49/mo	System-wide (any app)	Proven, excellent	Already configured
Claude Code /voice only	$0 extra	Claude Code terminal only	Still evaluating	Built-in, zero setup
Both	$8.49/mo	Full coverage	Best of both	Two tools to manage

The honest answer right now: I don't know yet. And I think that's worth saying out loud instead of pretending I've reached a verdict after a few days.

Superwhisper vs /voice: What I've Noticed So Far

Superwhisper has two months of trust built up. I know its quirks. I know it handles my accent well. I know it gets technical terms right most of the time — words like "Playwright," "Vitest," "middleware." I've added custom vocabulary for project-specific terms. It just works.

Claude Code's /voice is newer to me, and I'm paying close attention to accuracy — that's the thing that matters most. Early observations:

What's working well:

• Zero friction to start. No app to install, no API key, no configuration. Toggle /voice and go.
• The push-to-talk model (hold spacebar, release to send) feels natural in a terminal context. No accidental activations.
• Transcription tokens are free — they don't count against your rate limits.
• It understands context. When I'm in a coding session and say technical terms, it generally gets them right.

What I'm still watching:

• Background noise. I work from home and it's mostly quiet, but I've seen reports that ambient noise causes transcription hiccups.
• Technical jargon consistency. Does it reliably get "tRPC" or "Drizzle ORM" or "QA" every time? Superwhisper lets me add custom vocabulary for this. /voice doesn't — at least not yet.
• Prompt quality. When I type, I'm concise. When I talk, I ramble. This isn't Claude's fault — it's mine. But it means my voiced prompts are longer and less focused, which sometimes leads to less precise responses. I'm learning to speak more deliberately.

The scope difference is also real. Superwhisper works everywhere — Slack, Gmail, VS Code, my terminal, my notes app. /voice works only in Claude Code. If I drop Superwhisper, I lose voice-to-text in every other app. That's not a small thing.

What I Haven't Figured Out Yet

I'm being honest about where I am: still testing. Here's what I'm actively trying to answer:

Can /voice handle QA-specific vocabulary reliably? Terms like "regression suite," "flaky test," "shift-left," "edge case matrix" — these come up constantly in my work. Superwhisper handles them well because I've trained it with custom vocabulary. I need more time with /voice to see if it keeps up without that customization.

Will I actually cancel Superwhisper? The $8.49/month isn't a lot. But it's the principle — if /voice is good enough for my terminal work, and I'm already paying for a Claude subscription, paying separately for voice-to-text feels redundant. Then again, Superwhisper covers every app on my Mac. It's not an apples-to-apples comparison.

How do /btw and /voice work together? I haven't tried voicing a /btw question yet. In theory, you could be watching Claude work, hold spacebar, ask a side question out loud, and get an answer without touching the keyboard. That's a workflow I want to explore.

Is the accuracy gap closing? /voice has already expanded to 20 languages and made the push-to-talk key rebindable in recent updates. The pace of improvement is fast. The Superwhisper I'm comparing against today might not be the right benchmark in a month.

---

I don't have a clean takeaway here — and that's the point. Most posts about new features either tell you "this is amazing, switch now" or walk you through setup steps. I'm not there yet. I'm in the messy middle of evaluating two tools that solve overlapping problems in different ways.

What I do know: /btw is already part of how I work. /voice is a maybe that's trending toward yes. And the fact that Claude Code is shipping features that address real workflow friction — not flashy demos, but the small annoyances that compound over a full day of coding — tells me they're paying attention to how people actually use this tool.

I'll update this post once I've spent more time with /voice and made a decision about Superwhisper. For now, if you're a Claude Code user, try /btw the next time you're waiting for a response. That one's an easy win.

---

Using /voice or /btw in Claude Code? I'd love to hear what you've noticed — reach out here or find me on the blog.