Bloomverse

logo
heropage
mindmap
agree_indicator
conflict_indicator
mediation
information_trace

Inspiration

Bloomverse began as two ideas: an emotion-aware AI mediator and a real-time meeting mind map.

We realized that both ideas address the same problem: complex conversations often produce more information than shared understanding. People repeat themselves, misread intentions, and leave without clear decisions.

Most meeting assistants summarize what was said. Bloomverse focuses on identifying where and why shared understanding breaks down.

What it does

Bloomverse is an AI meeting assistant that helps teams move from disagreement to shared understanding.

During conversations, it:

Creates speaker-attributed transcripts.
Builds a real-time mind map.
Organizes viewpoints, concerns, and decisions.
Detects repeated arguments and rising tension.
Identifies the topic causing difficulty.
Provides private AI-assisted clarification.
Summarizes decisions, owners, and next steps.

Bloomverse treats emotion-related signals as uncertain context, not as proof of anyone’s feelings.

How we built it

Bloomverse follows a real-time processing pipeline:

Meeting audio → speaker-attributed transcript → structured discussion state → real-time mind map

A Meeting Agent analyzes recent transcript segments and updates each discussion node with its topic, participants, viewpoints, concerns, misunderstandings, repetition level, and tension level.

A separate analysis layer combines language, audio, interaction patterns, and optional visual signals. These signals are treated as uncertain context rather than proof of anyone’s emotional state.

When a discussion becomes stuck, participant-specific AI agents provide private clarification using only the relevant public context. Raw private messages remain isolated, and participants control what information can be shared with the group.

Challenges we ran into

Our main challenge was distinguishing productive disagreement from harmful escalation. Instead of relying on a single emotion score, we considered repetition, interruptions, topic history, and whether the conversation was still making progress.

We also needed to connect rising tension to the correct topic, protect private conversations, and synchronize transcript, AI analysis, mind-map, and participant updates in real time.

Accomplishments that we’re proud of

We are proud that Bloomverse demonstrates a complete conflict-recovery loop rather than producing only another transcript or meeting summary.

Our prototype shows a discussion becoming difficult, a specific mind-map node becoming heated, participants clarifying their underlying concerns privately, and the group returning with a clearer decision and actionable next steps.

We are also proud of our central privacy principle:

Private raw messages ≠ shared messages.

AI supports human-to-human communication without exposing private thoughts or replacing human decision-making.

What we learned

We learned that meeting efficiency is not only about reducing total meeting time. A short meeting can still fail if participants leave with different interpretations of the decision.

A more meaningful metric is time to shared understanding: how long it takes participants to understand the decision, one another’s core concerns, and the next action.

We also learned that emotion analysis becomes more useful when connected to the structure of a discussion.

“This person is frustrated” is vague and potentially harmful.

“This topic has produced repeated disagreement and may need clarification” is more explainable and actionable.

Most importantly, AI mediation should be temporary. Direct human communication should remain the default.

What’s next for Bloomverse

Our next step is to compare teams completing the same difficult decision task with and without Bloomverse.

We plan to measure:

Time to shared understanding.
Time spent repeating existing positions.
Accuracy of captured concerns.
Clarity of decisions and next steps.
Participants’ understanding of one another.
Whether participants felt heard rather than suppressed.

Future versions could support multilingual meetings, classroom discussions, collaborative learning, family conversations, and opt-in relationship mediation.

Success is not simply a shorter meeting. Success means less time stuck in misunderstanding, with every important concern preserved.