Robin's voice input on a Tuesday morning: what people actually say
The NestBoard team · 2026-07-11
We built Robin to understand actual human speech, not the polite, complete sentences people use in product demos. That means handling the way you really talk on a Tuesday morning when someone's looking for cleats and the toast is burning.
Here's what we hear (all anonymized, all real):
"Robin add… wait, hang on… okay, Robin, add soccer practice Thursday at four… no, five… actually is it four or five?"
Robin waits through the pauses. If you correct yourself mid-sentence, it takes the last version. If you genuinely don't know, it'll ask: "Should I add soccer practice Thursday at 4:00 or 5:00?"
"Hey Robin we're out of… MARCUS PUT THAT DOWN… we're out of oat milk."
Background chaos is fine. Robin filters for the voice closest to the device and ignores most of the ambient noise. If a kid shouts directly into the tablet, though, all bets are off.
"Robinaddchickentothegrocerylist."
This one trips people up. If you run the words together too fast, Robin will ask you to repeat it. A tiny pause between "Robin" and the rest of your sentence makes all the difference.
"Robin, what's on the calendar today?"
Short, clear questions work beautifully. Robin reads back your day—appointments, chores due, anything flagged for today. No interpreting required.
"Robin… uh… that thing we talked about yesterday, can you add it to Saturday?"
Robin doesn't remember conversational context across sessions. It can't reference "that thing" or "the place we went last time." You need to name the actual thing: "Robin, add haircut appointment Saturday at 10."
"Robin, add… how do you spell doxycycline?"
You don't have to spell medication names. Just say them phonetically and Robin will match against its database. If it's unsure, it shows you a few options to pick from. (This works for the shared medication bottles feature—one bottle, multiple people.)
"Robin take a picture of this recipe."
Voice plus image upload is weirdly powerful. You hold up a magazine page or a card your neighbor wrote, say that sentence, and Robin extracts the ingredients and steps. It's faster than typing and way faster than importing from a URL when the recipe isn't online.
"Robin I need to… never mind."
If you trail off or say "never mind" before Robin responds, it cancels the request. No orphaned half-entries, no awkward "delete that" cleanup.
The pattern we see: people get comfortable fast. The first few requests are formal and over-enunciated. By day three, it's just part of the morning rhythm—quick, casual, often mumbled through a mouthful of toast. That's exactly what we want.
Robin isn't trying to be impressive. It's trying to keep up with you on a regular Tuesday, when nobody has time to open an app or type with sticky fingers. It works because it expects you to sound like a person, not a voice actor.