I stopped inviting note-taking bots to my meetings and built a meeting memory instead
I used to think the only way to get good meeting notes was to invite a bot into the call.
And technically, it worked. The bot joined. It recorded. I got a summary.
But socially, it was disruptive.
Someone would ask, “Who is that?”
A customer would pause.
The first minute would turn into permission and awkwardness.
Sometimes the bot would get blocked or stuck in a waiting room.
Even when everything “worked,” the vibe was different. People self-edited.
That experience sent me down a rabbit hole. Are recording and transcription tools actually neutral?
Not really.
Harvard Business Review makes a point I agree with: recording meetings can change the social fabric of a conversation and affect psychological safety, especially for people who already feel less comfortable speaking up.
Should You Record That Meeting?
So I stopped trying to make bots less awkward and asked a different question.
What if meeting intelligence did not require adding a bot participant at all?
That is how IceCubes started.
Why bots feel disruptive in real life
This is not about paranoia. It is about normal human behavior.
1) They hijack the first minute
The first minute sets the tone. A bot turns it into admin, policy, and explaining what is happening.
2) They change how people talk
When people feel monitored, they behave differently. This shows up beyond meetings too. The American Psychological Association has summarized research showing that workplace monitoring is associated with stress and tension for many workers. APA on electronic monitoring
3) They add friction and failure modes
Waiting rooms, permissions, compliance concerns, external meetings where it feels unprofessional. It is one more thing that can derail a conversation.
I wanted meeting intelligence that stays out of the way.
The approach: a browser extension, not a meeting participant
IceCubes runs as a browser extension and captures live transcripts from the meeting experience itself on Google Meet, Zoom, and Microsoft Teams.
No bot joins. No new participant appears. The meeting looks and feels normal.
The outcome is simple: your meeting becomes a memory you can search later.
Live AI is more useful than notes after
Summaries after the meeting are helpful, but the biggest “aha” for me was what happens during the meeting.
When you have live transcripts, you can ask questions in real time, like:
What decisions have we made so far?
What is still open or unresolved?
What objections have come up?
Draft a follow-up based on what we agreed
What should I ask next to clarify the requirements?
This is the difference between documenting a meeting and steering it while it is happening.
Cross-meeting chat is where this turns into real institutional knowledge
Here is the problem I kept running into before IceCubes.
We would align in a meeting. Two days later someone would ask, “Did we actually decide that?”
Then we would spend time recreating context that already existed, but only inside people’s heads.
With cross-meeting chat, you can ask a question across multiple meetings to:
trace decisions over time and recover the “why”
spot repeated patterns across customer conversations
generate weekly digests from everything discussed
find the exact moment a commitment was made
That is institutional knowledge, but searchable.
Exposing meetings via MCP unlocks “build anything” workflows
Even if a product is great, it cannot ship every workflow. Different professionals want different outputs.
So IceCubes exposes meeting memory via MCP, so your AI tools can access it (with the permissions you grant) and you can build workflows inside the tools you already use.
Examples that feel universally useful:
Sales and founders
Alert me when a competitor is mentioned
Summarize all calls this week into a single pipeline update
Pull top objections across meetings and draft responses
Managers
Extract recurring blockers from 1:1s over the last month
Turn standups into a weekly status update
Identify coaching opportunities from patterns in discussions
Product and engineering
Cluster customer pain points across calls by theme
Find when we changed scope and what tradeoffs were discussed
Convert decisions into action items or tickets
Recruiting
Turn interviews into consistent scorecards
Compare what different interviewers flagged across the loop
Generate a clean debrief summary with evidence
IceCubes also connects to the rest of your workflow with native integrations like GCal, Slack, HubSpot, and Salesforce, plus Zapier so you can send outputs to tools like Notion, Jira, Asana, Trello, Google Docs, and Airtable.
The small moment that sparked all of this
The honest origin story is not a grand strategy memo.
I was tired of meetings producing important decisions and context and then losing it. I was also tired of the social cost of bots.
And I kept thinking: if we can make code searchable, and documents searchable, why are meetings still trapped in calendars and memory?
So I built what I wished existed: meeting intelligence that is invisible during the call, but powerful during and after.
If you are curious, I would love your feedback
If your meetings were fully searchable, what would you ask first?
What are the open follow-ups across my meetings this week?
When did we decide X, and why?
What themes keep repeating across calls?
What did we commit to and when?
Give me a weekly digest from everything
If you want to see what I am building, it is IceCubes: https://www.IceCubes.app
We call it IceCubes because we ‘freeze’ meeting intelligence into reusable blocks of memory you can search, share, and plug into AI.
If you try it, I would genuinely love to hear what workflow you would build first once your meetings become searchable and AI-addressable.
