01 / talk to it
Say it, don't type it
Hold one key and speak. Push-to-talk dictation drops your words straight into the assistant — and talking runs ~2.9× faster than the keyboard.
push-to-talk F8the term
bottleneck
noun · the slowest point in a fast pipe
- the pipe
- AI, ~200 w/sec
- the choke
- your eyes, ~200 w/min
- the fix
- read with your ears
def. / you, at the keyboard
The model ships faster than you can read — so reading became the bottleneck, and it's you.
Your assistant writes walls of text and code in seconds. You skim them at a couple hundred words a minute, miss the part that mattered, and ship it anyway. The gap isn't the AI. It's the channel you're reading through.
the other channel · words per minute
~2.9×
faster, spoken than typed
2.
You can't review what you skimmed. TUNLstack speaks every reply aloud and lights each word as it reads — so you stop rubber-stamping AI code and rest your eyes while you take in the whole answer. Local voice, no cloud.
Three keys. That's the whole learning curve.
Most of TUNLstack is optional depth you grow into. The core is three keys — one to talk to it, one to steer it, one to move through it. Learn them in a minute, keep them for good.
02 / steer it
Play, pause, skip — on command
The operator key runs the read-aloud. Pause to think, cut a reply short, jump to the next one. You set the pace; the voice follows.
the operator key F903 / move through it
Navigate without the mouse
Step between replies and sessions by voice and hotkey. Move through five agents hands-free and keep one train of thought across all of them.
voice-nav F7Everything else is optional. These three aren't.