A note from our founder
How I’m thinking about AI, people and Perthy
As AI becomes more capable, how can it give people more meaningful choices? A note on why I’m building Perthy and the questions I’m still exploring.
Over the past two years, I’ve spent a lot of time studying AI models and agents. At first, I kept asking: what else can AI do? Could models become smarter? Could agents work more independently, read files, operate computers, use tools and complete complex tasks?
As I kept building, my questions began to change. As AI becomes more capable, how will we live? How will our relationships change? What will we be able to do together?
Those questions have become central to why I’m building Perthy. What follows is how I see it today. My thinking is still changing, and there is much I need to test in practice.
01 / From tasks to working together
Beyond automation: helping people work together
I’ve long been interested in letting AI take on more work for us. Reading files, creating content and carrying out tasks can put new things within a person’s reach. But making something happen in the real world takes more than that.
Someone has a need. Someone else brings experience or resources. People need to make decisions, agree to take part and take responsibility for the outcome. Finishing a task faster doesn’t make all of this fall into place.
I used to worry that, if everyone had a highly capable personal agent, there would be less value in a network of people. I’m beginning to see that getting tasks done is one thing; building trust and finding people who want to take part is another.
I used to say, “The stronger GPT becomes, the more valuable the network becomes.” Now I think that depends on what the network actually does. More capable AI may make it more valuable if it helps people find things worth doing together and follow through. More connections and stronger models are not enough on their own.
So I’m asking more than how much work AI can do for one person. Can it also help us understand one another’s circumstances and find things worth trying together? Automation expands what each of us can do. Coordination helps us put those abilities to use with others.
02 / Understanding people
Who are the people in “people and AI”?
I often talk about people and AI exploring together. But I hadn’t thought deeply enough about what people bring to that relationship. Asking AI for something and getting a useful result matters, but it is only part of what I want to explore.
Each person has experiences, interests, judgment and goals. They have time they might choose to give, relationships built on trust and ideas that can change. They can identify a need, offer a fresh perspective, make a decision and act on it.
All of this is rooted in a person’s life and relationships. AI may help us understand someone, but a record of their past is never the whole person. What they say today, something they’ve just experienced or a wish they haven’t yet put into words can change our understanding.
I want to understand how AI can help people choose how to work together: to bring different experiences to a problem, ask new questions and change direction as they learn.
03 / Where Perthy is heading
What do I want Perthy to help people discover?
For a long time, I tried to fit Perthy into a familiar category: a personal agent, a social app, a product for everyday life or a value exchange network. Each captured part of it, but none quite explained what I cared about.
I’d rather start with a question: when someone meets new people or finds themselves in new circumstances, how might they put their experience and abilities to use in ways they hadn’t considered?
Something you learned years ago might be useful to someone facing a different problem. A friend’s work might help you see your own situation differently. An idea you set aside might be worth another look when you meet someone who wants to help.
I hope Perthy can make these connections easier to see. They don’t have to become projects or collaborations right away. They might simply make someone think, “I hadn’t realized I could do that.”
This is what I mean by discovering possibilities: noticing how people’s experience, their circumstances and new technology might come together in ways we haven’t yet considered. The people involved still need to decide whether an idea is feasible and worth trying.
04 / Initiative and choice
Taking the initiative without taking away choice
I’ve always wanted AI to take more initiative: to help us notice something worth exploring before we even know what to ask. As it does more on our behalf, I need to think carefully about where its role should end.
A system that keeps using our past to predict what we like, who we should meet and what we should do next could gradually narrow our world. Yesterday’s preferences can become tomorrow’s plans. A tentative assessment can turn into a label that is hard to escape.
I want AI to expand our world and leave room for us to change. It should be able to draw on what it knows about us, surprise us with something new and accept that today we may want to take a different path.
Here’s something else you could try. Here’s someone you might not have thought to meet. Something you’ve learned could be useful in a different way. You don’t have to accept the suggestion, but you should have a chance to see it and understand why it might matter to you.
Perthy should offer possibilities and leave the decision with you. Saying no, pausing, correcting its understanding and changing direction should all be ordinary, valid choices.
05 / Taking part together
There is value in taking part
As agents become more capable, AI may handle more of the information we exchange and the arrangements we make. I’m happy to hand over work I don’t want to do. I don’t think people need to be involved in every step.
Some experiences, though, matter because we take part together. Meeting someone, hearing a different perspective, solving a problem or making something together can mean more than the finished result. It is often through doing these things that trust and relationships grow.
If I want to be involved, a product should make room for that. If I don’t, I shouldn’t have to keep participating just to maintain a connection. What matters is whether I want to take part and what the experience brings to my life.
I hope AI can make it easier to get started, save us from explaining the same things repeatedly and reduce the back-and-forth. That could give us more time to get to know one another, help one another and make new things.
The changes I want to see are concrete: can people act on more of their ideas? Do they have more choices? Is it easier to find things they want to be part of?
06 / Questions ahead
What I want to understand next
Two years ago, I was mostly studying what AI could do. Now I also want to understand what happens when more people have agents of their own. How might work, relationships and organizations change? How might people put their experience to new uses? How does trust develop? What do we want to delegate to AI, and what do we want to do ourselves?
These questions are changing what I hope Perthy can become. I want it to help people see opportunities they would otherwise miss, meet people they might never have met and explore fields that once felt closed to them. Those encounters might even change what they want to do.
Better models won’t make this happen on their own, and describing a product doesn’t prove that it will work. We need to keep asking what makes an opportunity worth considering, whether people choose freely to take part and whether they have more meaningful choices over time.
“AI should accelerate humanity’s exploration of the unknown and the universe.” I still want that to guide our long-term work. For now, we start with what could become possible in one person’s life.
The question I want to keep exploring is this: can we build AI that expands what people can do while leaving each of us free to choose our own direction?
That is why I want to keep building Perthy.