We don't keep logs of what you ask Robin
The NestBoard team · 2026-06-28
When you ask Robin a question—whether you're saying "What's for dinner Thursday?" out loud or typing "Add milk to the pantry"—we process that request and then delete the transcript.
We don't store what you said. We don't keep a searchable archive of your queries. We don't build a profile of the things your household talks about.
What we do keep
We track whether Robin understood you or returned an error. We log which features get used (calendar lookups, pantry adds, chore assignments). We measure response time and success rates so we know if something breaks.
These are aggregate, technical metrics. They tell us "Robin failed to parse 3% of voice queries on Tuesday" or "meal plan lookups are up 40% since we added recipe import." They don't tell us what you asked.
If you report a bug and opt to share details with support, we'll see the specific interaction. Otherwise, the words disappear as soon as Robin finishes processing them.
Why we built it this way
The easy path would be to log everything. Transcripts make debugging simpler. They'd let us train better models, catch edge cases, build features that anticipate what you're about to ask.
But a family calendar knows sensitive things. Medical reminders. Therapy appointments. The fact that someone's struggling with a particular chore rotation. When you're standing in your kitchen asking Robin to add a medication or move a private event, you shouldn't have to wonder whether that sentence is sitting in a database somewhere.
We've written before about what Robin actually reads—the permissions boundaries that keep it from accessing things like your kids' accounts without explicit sharing. This is the same principle applied to ephemeral queries.
Processing happens in memory. The transcript exists for a few hundred milliseconds while Robin figures out what you need, executes the action, and returns a response. Then it's gone.
The tradeoff
This makes some things harder for us. When someone reports "Robin misunderstood me yesterday," we can't pull up a log and see exactly what happened unless they're willing to recreate it. We catch fewer subtle bugs. Our error reports are blurrier.
We think that's the right tradeoff. You're letting a piece of software into the middle of your household routines. It should know as little as possible, for as briefly as possible.
Metrics tell us if Robin is working. Your actual words are none of our business.