LoginSign Up
← Back to all sparks
T

Tana

COLLAB
Velocity7.5

AI-native workspace combining notes, tasks, and knowledge management with a flexible graph-based structure

tana.inc ↗

Tana's voice agent starts doing concurrent background work inside live meetings.

meeting-agentsvoice-agentsai-chatsearchknowledge-management
◆Current state
Tana has spent the visible window building around the meeting rather than the note. The 28 August release turns the voice agent from an input method into a background worker: several requests run at once while the conversation continues, it acts in connected tools, and it reads the day's agenda, pinned items, captured outcomes and the shared screen for context. Around that sit a natural-language search that resolves plain questions into filter pills, a saveable search document, and a long tail of meeting-surface refinements.
◆Where it's heading
The through-line since July is Tana claiming the meeting itself — first as the default place to hold the call, then as the system that notices unscheduled ones, and now as an agent doing work during them. Each release has moved further from recording what happened toward acting while it happens. The supporting work is consolidating too: one transcript, one digest timeline, one filter bar, which is the housekeeping that follows a surface being used seriously.
◆Prediction
Expect the concurrent agent to be given more tools to act in and the daily brief to grow beyond Slack, since both are described as user-editable rather than fixed. The open question these entries do not settle is how Tana handles an agent that edits documents mid-meeting when it gets one wrong.

◆Recent moves

  1. 11d ago

    Multitasking voice agent, natural language search and daily briefs

    ⚡ SPARK

    The voice agent stops being a way to talk to Tana and becomes something that works while you talk to other people, which extends the meeting-ownership arc the 20 July and 19 August releases opened. The natural-language search and the Slack daily brief shipping alongside it are continuations of existing threads rather than new ground.

    View source ↗
  2. 18d ago

    A clearer transcript, and proposals you can fix in place

    Proposal fields become editable inside the comparison view, and the transcript consolidates into one selectable, speaker-grouped list. This is refinement of the review surfaces around AI output rather than an extension of what the AI does — the maintenance layer between the larger meeting releases.

    View source ↗
  3. 20d ago

    Tana notices the meetings you didn't schedule

    ⚡ SPARK

    The release that decoupled capture from the calendar, detecting meeting-like activity and offering to record it. It established the always-on meeting context that the 28 August voice agent now reads from, making it the hinge between passive capture and active participation.

    View source ↗
  4. 1mo ago

    AI chat goes further before hitting usage limits

    Token optimisation that pushes back AI chat usage limits, shareable links to documents the AI references, and accent-folding person search. Cost and reliability work on the AI surface, not new capability, though cheaper chat is what makes an always-running agent affordable later.

    View source ↗
  5. 1mo ago

    Ask the AI to pin things for you

    Pinning by voice or chat, applied immediately with no proposal to approve, alongside two desktop fixes. Narrow in scope, but an early instance of the pattern the August voice agent generalises: the AI acting directly rather than proposing.

    View source ↗
  6. 1mo ago

    Start, schedule, or capture external meetings right from Today

    ⚡ SPARK

    The release that made a native Tana call the primary path from Today, with external capture as the fallback. It set the direction every meeting release since has extended, including the agent work in August.

    View source ↗