Your first conversation with Mirror

After your DNA finishes processing, the most useful next step is to talk to Mirror. The agent already has access to every analytical result, so a good first question is open-ended rather than narrow. Mirror is at its best when it can choose what to surface for you, which means giving it room to summarise rather than asking it to retrieve a single number you could read from the Genome page yourself.

Better first questions, in roughly increasing depth

• "Walk me through the most important things in my genome." Open-ended invitation. Mirror will decide what is most important based on signal strength and clinical actionability.
• "What in my profile would I want to bring to a doctor?" Filters to actionable findings only. Pharmacogenomic alerts, pathogenic carrier statuses, APOE if you ask, and any health-risk findings with established interventions.
• "What does my ancestry actually look like, and where did each component come from?" Routes through the ancestry analyser plus the migration history. Mirror can pull up the Origins globe and walk you through the migrations attached to your dominant haplogroups.
• "Compare my polygenic scores to the population." Gives Mirror licence to surface percentiles, agreement across the six PRS methods, and any traits where the methods disagree (a sign of lower confidence).
• "If I were planning to have children, what should I and my partner know?" Routes into the Lab. Mirror surfaces your carrier statuses, summarises the recessive-condition risk model, and suggests partner-screening genes worth checking.

Bringing a result into the conversation

Wherever you see a number on the platform, you can tap "Ask Mirror about this" to open the agent with that specific number already in context. You do not have to retype anything, and you do not have to remember what tab you were on. The deep link encodes the metric, the surrounding tab, and the value, and Mirror starts the conversation already aware of what you were looking at.

Voice mode best practices

• Speak in complete questions. The transcription accuracy is highest when sentences end with falling intonation, and Mirror responds better to a structured question than to a fragment.
• Use names rather than pronouns for clarity. "Compare my BRCA2 result to the population" beats "compare it to the population" when the prior turn was about a different gene.
• Pause briefly between questions. The four-state state machine needs a clean transition out of speaking before it returns to listening, and rapid-fire follow-ups can clip the start of the next question.
• Mute the microphone (toolbar button, not just turning your head away) when you do not want Mirror picking up ambient conversation. The state machine is designed to ignore short utterances but a long enough side conversation can trigger a turn.

Multi-turn conversations

Mirror remembers the last few turns within a session and the running summary across sessions. You can build on a previous answer ("you mentioned APOE earlier — what about TOMM40?") without restating the context. If a conversation becomes long enough that the model would otherwise lose track, the platform compresses earlier turns automatically and surfaces a notice in the transcript so you know which turns are summarised rather than verbatim.