I don't think an AI desktop robot should have just one job.
That might sound obvious, but I think it changes how we should think about the entire category.
A robot can be cute.
It can be funny.
It can talk to you.
It can sit there while you're working.
It can help you with something.
It can remind you about something.
It can push you when you're procrastinating.
It can listen when you want to talk.
It can even teach you something.
The strange thing is that we keep trying to give these robots a single label.
companion
assistant
pet
productivity device
educational robot
I don't think that makes much sense for a robot that is supposed to stay on your desk every day.
A person doesn't have only one role in your life.
Why should your robot?
You don't want a different robot for every problem
Imagine having four things on your desk.
One is for companionship.
One is a secretary.
One is a coach.
One is an assistant.
That sounds ridiculous.
And yet that's effectively how much software already works.
You open one app to talk.
Another to organize your work.
Another to learn.
Another to track your habits.
Another to remind you about something.
Each one knows a tiny piece of your life.
None of them really knows you.
A desktop robot has an unusual opportunity here.
It can potentially remain the same presence while changing what it does for you.
That's a much more natural relationship.
In the morning, you may want a companion
Imagine you sit down for breakfast and the robot is already there.
You don't need it to give you a productivity report.
You might simply say:
"Morning."
And it knows you.
Maybe it remembers that you had something important today.
Maybe it asks:
"Still worried about that meeting?"
Maybe you talk for a minute.
Nothing productive happened.
That's fine.
The interaction still had value.
This is the part of desktop robots that I think people sometimes dismiss too quickly.
Companionship isn't supposed to justify itself by producing output.
Sometimes you simply want another presence in the room.
That's one reason personality matters so much to owners of robots such as LOOI. Long-term users have described the personality and humor as the part that makes them continue to enjoy the robot, even while complaining about limitations in memory and usefulness.
The robot doesn't need to earn every minute of your attention.
It just needs to be pleasant to have around.
At work, the same robot can become your helper
Then your day changes.
You start working.
Now you don't want the robot behaving like a pet.
You want it to understand that you're busy.
But maybe you need help.
You say:
"I have a meeting in twenty minutes."
The robot knows enough about your recent conversations to help you prepare.
Later:
"What did I say I needed to follow up on?"
It knows.
After the meeting:
"What were the things I said I'd handle?"
It helps again.
The personality doesn't disappear.
The robot doesn't suddenly become a corporate software assistant.
It's still the same little character sitting there.
It just knows what role is useful right now.
That's a very different idea from building a "work robot."
Then it can become a coach
This is where I think things get even more interesting.
Suppose you're practicing a presentation.
You don't necessarily need an assistant.
You need someone to challenge you.
You say:
"Ask me the questions I'm most likely to get."
The robot does.
You answer.
It pushes harder.
You try again.
That could also apply to:
studying
interviewing
learning a language
practicing a speech
preparing for a difficult conversation
The robot doesn't need to become your teacher.
It can simply become the person who helps you practice.
Reachy Mini already has community applications that move toward this kind of role. The current ecosystem includes language-learning, storytelling, office and meeting-oriented experiences, while a community-built AI facilitator has been used to guide real Zoom peer-group sessions.
That is much more interesting to me than another robot demo.
The robot isn't just displaying information.
It is participating in what you're doing.
And sometimes it should be the annoying friend
This sounds strange, but I think it could become one of the most valuable roles.
You say:
"I'll finish this tonight."
The robot knows you've said that three times this week.
At some point it should be able to say:
"You also said that yesterday."
Not aggressively.
Not like a manager.
Just enough to make you think:
"Okay. Fair."
That's very different from a reminder app.
A reminder app doesn't know whether you have been avoiding something for three days.
A robot that knows your habits potentially does.
And that is where knowing the owner becomes useful.
A good relationship allows different roles
Think about a close friend.
Sometimes they:
listen.
Sometimes they:
joke.
Sometimes they:
give advice.
Sometimes they:
remind you about something.
Sometimes they:
tell you that you're making a bad decision.
Sometimes they:
simply sit with you.
We don't call those separate products.
They're just different parts of a relationship.
I think that's the better mental model for an AI desktop robot.
The hardware stays.
The personality stays.
The relationship stays.
But the role can change.
The important part is that the robot doesn't become fake
There is a danger here.
A robot could have:
Companion Mode
Assistant Mode
Coach Mode
Secretary Mode
and a dozen other switches.
That might work technically.
But I don't think users want to manage all of that.
Imagine having to tell your friend:
"Please enter assistant mode."
It would feel ridiculous.
The robot should eventually infer enough from the situation.
If you're casually talking:
be conversational.
If you're working:
stay out of the way unless needed.
If you're practicing:
challenge me.
If you're struggling:
help me think.
If I need encouragement:
support me.
That's not about having more modes.
It's about understanding what kind of interaction is appropriate right now.
This is where the robot's personality becomes more important, not less
It might seem that making a robot useful would make personality less important.
I think it could do the opposite.
If the robot is just a tool, changing roles is trivial.
It's just another interface.
But if it's supposed to feel like a companion, then changing roles without losing its identity becomes very interesting.
Imagine that the same robot talks to you differently in different situations.
Not because it became a different character.
Because it knows what you need.
The same personality can be:
playful when you're relaxing
direct when you're working
patient when you're learning
encouraging when you're struggling
quiet when you're busy
That's what a real relationship can feel like.
And that's much harder to create than adding five more animations.
The robot should not always agree with you
I also think this is important for the "friend" role.
An assistant that always says:
"Great idea!"
isn't particularly useful.
A companion that simply agrees with everything you say eventually becomes boring.
One long-term LOOI discussion made exactly this complaint: users enjoyed its personality but wanted something closer to a sidekick instead of an endlessly agreeable character.
Imagine telling your robot:
"I'm going to start this huge project tonight."
And it knows from experience that you're exhausted.
A useful response might be:
"You said you wanted an early night. Are you sure?"
That's more interesting than:
"You can do it!"
The robot doesn't need to become argumentative.
But it should be capable of having a little judgment.
Otherwise "knowing you" doesn't mean very much.
This could also make the robot more useful without making it feel like work
There's another subtle benefit.
If a robot can switch between roles, not every useful interaction needs to feel productive.
For example:
You're working.
The robot helps you remember something.
Then you take a break.
It tells you something funny.
Later you ask for advice.
It becomes serious again.
The robot doesn't have to split your day into:
productivity
and:
entertainment.
Those things can coexist.
That's probably closer to how people actually live.
The same robot could help with very different parts of life
Consider a normal day.
Morning
It keeps you company.
Before work
It reminds you about something important.
During work
It stays quiet.
Before a meeting
It helps you prepare.
After the meeting
It helps you remember what needs to happen next.
Evening
You talk about something unrelated.
Before bed
You ask it something you've been thinking about.
That's a lot of different roles.
But it still feels like one relationship.
That's the part I think desktop robots can do especially well.
This also makes the physical body more important
None of this means the body is becoming irrelevant.
Quite the opposite.
The body is what allows the robot to remain the same physical presence through all these different roles.
The same eyes.
The same voice.
The same little movements.
The same place on your desk.
That continuity matters.
If you only changed software roles through a phone app, the experience would be very different.
The robot gives the relationship a physical anchor.
That's why I don't think the future desktop robot should simply be:
"an AI assistant inside a cute shell."
The shell matters.
The personality matters.
The physical presence matters.
But those things become much more valuable when the robot's intelligence can actually adapt to the owner.
This is probably where "my robot" becomes meaningful
Imagine giving ten people the exact same robot.
On day one, they're almost identical.
Six months later, they shouldn't feel identical.
One person's robot might be:
the funny one
Another:
the quiet one
Another:
the strict one who keeps them on track
Another:
the study partner
Another:
the one they talk to at night
The hardware hasn't changed.
The personality has become shaped by the relationship.
And the roles have become shaped by the person's life.
That's much more interesting than simply having ten identical robots with ten identical feature lists.
It also creates a better reason for software to keep improving
Now we finally get back to the software.
But from the user's perspective.
I don't care that the company released version 4.2.
I care that six months later my robot can handle something it couldn't handle before.
Maybe it understands me better.
Maybe it remembers better.
Maybe it can help with something new.
Maybe it is better at knowing when to talk.
Maybe it can take on a new role.
That is what software evolution should mean to the owner.
Not:
"The software got updated."
But:
"My robot got better."
This is the same basic principle I explored in What Should an AI Desktop Robot Feel Like After Six Months?.
The important thing is not that the robot accumulates features.
It's that the relationship becomes richer.
This is why I think the future robot is closer to a sidekick than an assistant
"Assistant" is too narrow.
A traditional assistant is supposed to help you accomplish tasks.
That's useful.
But it doesn't explain why you would want a physical robot on your desk.
A sidekick is different.
A sidekick can:
hang around
make jokes
help
remind you
challenge you
encourage you
learn your habits
talk when you want to talk
shut up when you don't
That feels much closer to the product people actually seem to want.
And it explains why the physical personality matters.
The robot doesn't have to be useful every minute
This may be the most important conclusion.
I don't think the robot needs to justify its existence by constantly doing things.
Sometimes it can simply be there.
That's enough.
The important part is that when you need something, it is capable of becoming useful without making you switch to another tool, explain your entire situation again, or interact with it like a software menu.
It knows who you are.
You know who it is.
And the role changes naturally.
That's a very different product from:
productivity robot
or:
AI pet.
It's something in between.
The future desktop robot may have one personality and many jobs
That's the model I keep coming back to.
One robot.
One relationship.
Many roles.
It can be:
your companion when you're alone.
your secretary when you have things to remember.
your assistant when something needs doing.
your coach when you need to practice.
your teacher when you want to learn.
your friend when you just want to talk.
And sometimes:
nothing at all.
It just sits there.
You know it's there.
That's valuable too.
I think that's a much better vision for AI desktop robots than trying to decide whether they are supposed to be toys, assistants, pets, or productivity devices.
They can be all of those things at different moments.
The challenge isn't giving the robot more identities.
It's making all those roles still feel like the same robot you know.
And maybe that's what eventually makes a desktop robot worth keeping.
Not because it can do everything.
Not because it talks all the time.
Not because it has the most impressive body.
But because after enough time, you have something sitting on your desk that feels familiar enough to be a companion, smart enough to understand what you need, and capable enough to actually help when you ask.
That is a much harder product to build.
