Best Practices for Using Whisperer
The tips here are distilled from what actually works for Whisperer users. Read the whole list or jump to the section you need.
Before the meeting
Fill in user context
Write 3–5 sentences about yourself: who you are, what type of task is coming up, and what response style you need. The assistant will take this into account without you explaining it every time.
Example:
"I'm a Python developer going through a technical interview. Answer concisely: lead with the key point, then code. Skip preambles."
Create a template for the meeting type
In Prompt Studio, make a separate template for interviews, client calls, and internal meetings. It takes 5 minutes and saves focus when things get intense.
Fill your knowledge base
Add to the knowledge base:
- Your résumé or bio
- Product / stack description
- Frequent questions with your answers
- Background on the company you're meeting with
RAG will pull the right piece into each response — you won't have to repeat context.
Check permissions
Before the call, make sure the microphone and (on macOS) the "Screen Recording" permission are granted. Without them, the assistant can't hear the other side.
During the meeting
Ask short, specific questions
"How do I handle the 'it's too expensive' objection?" — a good question. "Explain sales to me" — too broad; the answer will span several screens.
The shorter the question, the faster the streaming and the more accurate the answer.
Use empty Enter for "from what was heard"
When the other person just asked a question, press Enter in the empty field. The assistant will pick up the latest transcript fragment and answer it. Your hands stay free.
Take screenshots for diagrams and code
If a diagram, database schema, or code appears on screen — take a screenshot (with the hotkey). The assistant will analyse the image and answer in the context of your question.
For long screens, accumulate several frames (multi-screenshot) and send them as one batch.
Watch the transcript
If the transcript "drifts" — errors or gaps appear — check the volume level and your distance from the microphone. A good transcript = accurate answers.
Use System Design mode for architecture questions
Ask a question in System Design mode and get a structured response with a Mermaid diagram: components, data flows, scaling approach. Ready to paste into your notes or show immediately.
After the meeting
Review the mind-map
The mind-map is generated automatically within a minute of completion. Check it while the meeting is fresh: fill in nodes the AI missed and click "Enrich" if the map feels incomplete.
Move action items to your tracker
Action items from the meeting map are ready-made tasks. Copy them to Jira, Linear, Notion, or any other tracker right after the meeting, while the details are still fresh.
Save useful answers to the knowledge base
If the assistant produced a particularly accurate explanation or a strong argument — add it to the knowledge base. Next time it will appear in context automatically.
Export the transcript
For client calls and important meetings, export the transcript (text or Markdown) immediately after completing — for the archive, a follow-up email, or later analysis.
For specific scenarios
Technical interviews
- Template: brief answer → code → complexity
- Knowledge base: design patterns, common LeetCode / System Design questions
- System Design mode — for architecture questions
- Empty Enter — when the interviewer just asked a question out loud
Sales and negotiations
- Template: acknowledge → reframe → next step
- Knowledge base: client cards, common objections with answers, product description
- Multi-screenshot — if the client shows a presentation or competitor comparison
Teaching and tutoring
- Template: explanation → example → check question
- Knowledge base: course syllabus, typical student mistakes
- Translation (TranslationStrip) — if the student speaks a different language
No-logs for sensitive meetings
For confidential conversations, enable no-logs mode: transcript and responses won't be saved. Remember: minutes are still spent.