Working memory · local-first

As we may flow

Working memory that never loses context.

Meetings, voice notes, tasks and agreements come together in one memory system — and come back exactly when the context is needed.

  • Hears
  • Understands
  • Connects
  • Brings back
Book a demo How it works
flow.AsWeMay
Thursday · 27 August
● Voice context on90 s bufferlocal processing
tasks today
16
decisions
7
awaiting reply
4
review
12
Today’s priorities
01
Confirm the architecture decision
discussed yesterday
02
Get back to Misha on the API deadline
promised by Friday
03
Sort out 2 tasks with no owner
needs review
decision

We keep the second architecture option. The first one is off the table.

task

Misha will check the read-only Agent API by Friday.

  • Local processing
  • Long-term context
  • Tasks only after confirmation

01 · Principle

Information overload → working memory

flow.aswemay remembers not just what was said, but what it changes in the work.

Not another transcription service. flow.aswemay turns the flow of information into connected work objects: decisions, tasks, people, projects, questions and risks.

  1. 01

    Hears

    Meetings, voice notes and the flow of work.

  2. 02

    Understands

    Picks out tasks, decisions, agreements and risks.

  3. 03

    Connects

    Connects what is new to people, projects and past context.

  4. 04

    Brings back

    Gives you the right context before a meeting, in the morning or on request.

Decisions do not dissolve in transcripts. Promises do not stay stuck in chat. Context keeps living between meetings.

02 · How it works

Not a summary — a working model

A line from a meeting becomes part of the memory.

flow.aswemay does not drop yet another transcript into an archive. It turns the conversation into structured context you can keep using.

flow.AsWeMay/ review
meeting 14:03 · 12 min
● 4 items to reviewconfidence 0.92
What was understood from the conversation
task

Maria is preparing the proposal by Friday.

change

The deadline moved to Monday in the 24.08 email.

decision

We are moving the launch to March.

question

When will the budget be approved?

Linked context
PROJECT
Launch
moved to March
PERSON
Maria
proposal
SOURCE
Meeting 27.08
email 24.08

Light theme · the product default

  1. 01

    Extracts the fact

    Recognises a decision, task, agreement, question or risk.

  2. 02

    Understands the links

    Identifies the project, people, deadlines and related discussions.

  3. 03

    Waits for your confirmation

    A finding becomes a task only after you confirm it.

  4. 04

    Puts it to use

    At the next meeting the earlier decision is already part of the context.

03 · Ask the memory

Don’t search. Ask.

Context comes back by meaning, not by file name.

Ask: “What did we decide about the launch in the end?” — flow.aswemay finds the answer along with its history and sources.

●

Sources

04 · Local-first

Your computer and your server

Your working memory stays under your control.

flow does not send your recordings or emails to third-party services. Recording, transcription and analysis of short meetings happen on your computer. Long meetings are analysed by a more powerful model on the analysis server — in a company, that is your own server inside your perimeter. flow does not use third-party cloud AI services.

Architecture

Privacy is not a separate setting. It is part of the architecture.

Audio never leaves your computer. Only the transcript text goes to the analysis server, and only with your consent. After processing, the text is not stored on the server.

If you don’t have a powerful computer, long recordings can be analysed on flow’s analysis server — only with your consent.

01 · Local

Local processing

Speech recognition, first-pass analysis and search across your knowledge base run on your computer.

02 · Two tiers

Two tiers of analysis

Short meetings and emails are analysed by a model on your computer, long ones by a model on the analysis server. Nothing goes to the server without your consent.

03 · Control

Source control

Every important fact can be traced back to the meeting, note or document it came from.

05 · Capabilities

What flow.aswemay can do

One context instead of a set of scattered AI tools.

01 · Voice context

The meeting is over. The work with its context is not.

Voice turns into decisions, tasks, questions and agreements — linked to the source.

02 · Memory

Long-term context

New facts are linked to projects, people and earlier decisions.

  • project
  • decision
  • next step

03 · Review

Review before saving

The system shows where it is confident and where a person should confirm the meaning.

confidence0.92

04 · Daily brief

30 seconds to get into your day.

Tasks, replies you are waiting for, meetings and whatever must not be lost from yesterday’s context.

03key contexts today

05 · Projects

A project’s history, not a folder of files.

Decisions, people, tasks and sources come together in one living context.

06 · Semantic memory

No need to remember where the information lives.

Ask — and flow.aswemay finds the answer along with its history and sources.

Ask memoryWhat has changed since the last discussion?→

Demo · 30 minutes

Not a presentation.
We show it on your own context.

In the demo we can show how the system takes a meeting apart, picks out tasks and decisions, links them to past context and answers from the accumulated memory.

  • We break down a real meetingFrom voice to structured objects, step by step.
  • We test the long-term contextNew facts get linked to earlier decisions and projects.
  • We discuss the local-first setupWhich data and operations make sense to keep local in your particular case.

By sending the request you agree that we may use these contact details to arrange the demo.

As we may flow.