Baxnet Ideas · Founder note
Mimoto Should Be the Blood Test, Not the Doctor
TL;DR: Personal Chat Intelligence should make conversation evidence easier to inspect without pretending that a message count can diagnose a relationship or prescribe what happens next.
Imagine a doctor walking into the room with a clipboard.
The page contains measurements, units, reference ranges and the date the sample was taken. The doctor studies it, then asks about symptoms, medication, sleep, family history and what has changed since the last appointment. The report matters because it gives the conversation something stable to work from. It does not conduct the whole consultation.
I keep returning to that division of labour when thinking about what Mimoto should become.
Mimoto should operate like the blood test, not the doctor.
That is an analogy, obviously, and there are limits to it. Message history is not a biological sample. Mimoto is not a medical device, and a relationship report should never be mistaken for a diagnosis. The comparison is useful because it gives Personal Chat Intelligence a clear job: produce defined, inspectable results, then leave room for the person—or somebody they deliberately choose—to interpret what those results mean.
A correct number can carry the wrong story
Suppose a report shows that the average time between messages has doubled over six months.
That may be a correctly calculated result for the selected conversation and period. It may also be almost useless without context.
Perhaps the relationship moved from messages to phone calls. Perhaps one person started a new job, had a baby, travelled, became ill or changed messaging platforms. Perhaps the friendship genuinely became more distant. The chart cannot choose between those stories, even though the line on it is real.
The same applies to message balance, who usually starts a conversation, how often questions appear, or whether communication becomes concentrated at different times of day. These can be useful measures. None arrives with a universal emotional meaning attached.
A message count can be correct while the story attached to it is wrong.
This is one reason I am wary of products that rush from private data to a confident verdict. The temptation is understandable. People rarely open a relationship tool because they are curious about database rows. They want to know whether somebody cares, whether a friendship changed, whether an argument was part of a pattern, or whether their own memory is being fair.
Those are legitimate questions. They are simply larger than the evidence contained in a message archive.
Relationships do not come with healthy ranges
Even the medical side of the analogy is less absolute than it first appears.
MedlinePlus explains that laboratory tests are important but usually do not provide a complete picture of a person’s health. A clinician combines them with an examination, health and family history, imaging and other information. A result outside a reference range may not indicate a problem, while a result inside the range is not a guarantee that everything is fine.
That is a useful dose of humility for personal data products.
There is no agreed healthy reply time for a friendship. A perfectly balanced conversation is not necessarily a close one. One person sending most of the messages may reflect care, pressure, humour, coordination, personality or the simple fact that the other person phones instead. A quiet month can signal distance or a period in which both people were standing in the same room.
The report can describe the pattern it measured. It should not invent a normal range for human closeness.
What the results layer should contain
This boundary does not make the product timid. It makes the work more exact.
Mimoto can already turn supported message history into reports, trend views, relationship metrics, communication-flow views and structured exports. The aim is to make a long, messy archive easier to inspect than it is in a search box or years of scrolling.
A trustworthy results layer should make several things clear:
- what was measured and how the measure is defined;
- which conversation, participants and time period were included;
- which source material supports an observation;
- whether the output is a direct calculation, a classification or an inference;
- what data may be missing;
- and where the person can inspect, correct or reject the result.
There is judgement in this work. Choosing a metric is a choice. Deciding where one conversation begins and another ends is a choice. A label such as encouragement or apology may require classification rather than simple counting. Even a deterministic calculation can become misleading if the period or denominator is hidden.
So I would not describe every Personal Chat Intelligence output as perfectly objective. The better ambition is structured, defined and reviewable. Reproducible where possible, explicit about uncertainty where it is not.
A report should be easier to share than an archive
The distinction becomes more useful in a future filled with personal agents and specialised advisers.
Imagine somebody discussing a strained friendship with a qualified therapist or another adviser they trust. They remember that communication has collapsed, but memory is doing what memory does: compressing six months into the last painful week.
With the person’s explicit permission, the adviser might ask for a narrowly scoped report covering one conversation and one period. Mimoto could return selected trends and source-linked observations. The adviser would then consider that material alongside the person’s account, circumstances, goals and everything that happened away from the chat.
The report would travel without turning the entire private archive into an attachment.
A future personal agent could help coordinate a similar request, although a software agent and a qualified human professional are plainly not equivalent. In either case, the permission boundary matters. The user chooses the source, scope, recipient and purpose. Producing a report does not grant continuing access to the underlying relationship history, and receiving measurements does not grant authority to act on the user’s behalf.
That separation is valuable for the adviser too. A therapist, coach or agent should be able to see what was measured without inheriting Mimoto’s interpretation as the only possible explanation. That keeps the report in the position of evidence on the clipboard, with the adviser’s judgement still open to question.
Knowing where the job ends
There is a version of personal AI that tries to become the entire room: it gathers the evidence, explains the evidence, gives the advice and perhaps carries out the next action. Sometimes that will be convenient. With intimate data, it also makes errors difficult to locate. A mistaken classification can quietly become a mistaken story, then a mistaken recommendation.
I would rather Mimoto become dependable infrastructure for reflection and advice than compete to deliver the most persuasive verdict.
In practice, much of that is unglamorous work: state the definition, show the period, preserve the source trail and admit what is missing. A person should be able to take the output into a serious conversation without the report having settled that conversation in advance. Different advisers may interpret the same pattern differently, and they do not need unrestricted access to the life behind it in order to do so.
For Mimoto, the job is to put better evidence on the clipboard—and know when to stop writing.