正在加载视频...

视频加载失败

this guy built an app that lets him control multiple claude code instances WITH JUST HIS EYES AND VOICE he looks at the terminal he wants and speaks his command. that's it no keyboard. no mouse. just gaze tracking + voice on a macbook pro absolutely unreal. I guess...

94,583 次观看 • 7 个月前 •via X (Twitter)

34 条评论

Slopware Engineer 的头像
Slopware Engineer7 个月前

A lot of people don't know this but you can achieve half of this with just stock macOS using "head pointer" setting that uses the web cam and your head to move the mouse anywhere on the screen. I use it in combination with normal keyboard bindings for micro adjustments and caps lock for mouse click.

Tech With Matteo 的头像
Tech With Matteo7 个月前

that's wild, never thought about using eye tracking for this kind of workflow. do you think it actually speeds things up compared to just typing?

northStar0x7 的头像
northStar0x77 个月前

Huge respect to @therituallab for building eyeTerm and making the built-in Mac camera actually useful for something beyond selfies Accessibility win, productivity cheat code, and honestly a glimpse of how agentic coding will probably feel in 2–3 years for everyone.Question for anyone who's tried similar: does the gaze latency stay low enough during long sessions, or do you still end up making subtle head adjustments?Either way, keyboard is officially on life support.

ashen 的头像
ashen7 个月前

What's this guy's handle and how can I get this app?

Om Patel 的头像
Om Patel7 个月前

@therituallab

robertus 的头像
robertus7 个月前

gaze tracking + voice as a dev interface removes the biggest bottleneck in coding which isn't thinking, it's context switching. the day your editor just knows where you're looking is genuinely closer than people think

Steven Smith 的头像
Steven Smith7 个月前

I speak to my PM agent. She organizes the backlog. My developer agent takes on anything he can. My reviewer agent reviews and merges low risk. Calls my attention to things that need it. Agent monitors system interactions to proactively report problems or suggest enhancements. None of it requires my laptop open.

MVP 的头像
MVP7 个月前

No cto for pippn's friend...will you take up

Craig Doriot 的头像
Craig Doriot7 个月前

This very thing was warned of in a detailed vision about 15 years ago

Object Orienter 的头像
Object Orienter7 个月前

Naaah, at least for me, it won't happen like this. The reason why I like typing is that I can organize my ideas mentally without speaking the line of thought straight away, because it's just noise most of the time. Typing gives me to control when I consider it ready.

Rob Builds 的头像
Rob Builds7 个月前

Run multiple Claude Code instances daily. The bottleneck isn't typing — it's switching context between them. Gaze tracking solves the right problem. This is an accessibility tool that accidentally became a productivity tool.

Gocha 的头像
Gocha7 个月前

this is where it gets interesting. the bottleneck was never typing speed. it was context switching between terminals. eyes solve that instantly. you look at what you want and talk to it. that's how humans naturally work. keyboards were the workaround, not the solution

orkward ☄︎ 的头像
orkward ☄︎7 个月前

this is another level of madness, wow.

Sharon Daniel ©️🎭 的头像
Sharon Daniel ©️🎭7 个月前

winke wink . its gonna be hard not to wink lol you can't use this in real Real production

Sylvain Dechamps (Sido)🧙‍♂️ 的头像
Sylvain Dechamps (Sido)🧙‍♂️7 个月前

Futur Software Engineers 👇

K Fluffy 的头像
K Fluffy7 个月前

Idk if its actually doing proper tracking versus using the placement of his eyes within the grids. I'd like to see it done without face movement.

Priyanka Pudi 的头像
Priyanka Pudi7 个月前

Looks like next openclaw VR

Democracy Manifest 的头像
Democracy Manifest7 个月前

prob not that hard to do with something like Tobii Eye Tracker

sir rabbit 的头像
sir rabbit7 个月前

him, next year

Matei Gherceanu 的头像
Matei Gherceanu7 个月前

So what the hell you do with it? Eat, fart, fuck and code or work same time? Why not just replace himself, ohh better go full digital and become a desktop pc. :)) then will be the OP agent usage. No body just a pc

Ros Marino 的头像
Ros Marino7 个月前

I mean… nah.

Codie Mullins | AI + Sports 的头像
Codie Mullins | AI + Sports7 个月前

I can’t even wink with both eyes. This guy is such a show off.

Shreya Gupta 的头像
Shreya Gupta7 个月前

But this seems too slow Would rather use mouse

Atul thakur 的头像
Atul thakur7 个月前

the multi-instance approach is where things get interesting. we run 2-3 agents in parallel at Crove too. one for planning one for execution. but controlling them with eyes is next level

Stay Ready 🎺 的头像
Stay Ready 🎺7 个月前

instead of winking he should just snap his fingers. winking all the time is annoying

Dhru 的头像
Dhru7 个月前

The eyes and voice actually help give more input since the model has context what the person is seeing and gets clear execution intent from the voice, more human input better performance

troneras 的头像
troneras7 个月前

Very cool! I wish I had enough tasks to need this setup haha

Boz NZ 的头像
Boz NZ7 个月前

.. and totally logical with hindsight (or if you read sci-fi)

A M 的头像
A M7 个月前

Now make it so you can just do a thumbsup to get him to continue 👍

DavidJBerman 的头像
DavidJBerman7 个月前

Claude filed 47 sexual harassment claims against him

Matt 的头像
Matt7 个月前

this is exactly the direction we went with - voice as the primary input for controlling your whole mac. gaze tracking is a cool addition though, we use macos accessibility apis to know which app has focus. the keyboard/mouse era for AI interaction feels so wrong

Voy Zeglin 的头像
Voy Zeglin7 个月前

I can already see sooo many applications!

Andrey 的头像
Andrey7 个月前

using your voice is the future, so ppl need to use @VoiceOSApp

ᑯƌǨ 的头像
ᑯƌǨ7 个月前

Have you seen the Claude Code / Openclaw butt plug? dude you can control it...

相关视频