How is AI supposed to fix bullying?
Saw this come up as a request from a nonprofit at a hackathon. Sure, AI can help gather evidence. But the hardest part for the victim is what comes after. That feeling of being completely alone with it. That takes people, not AI.
Love this — it basically tells a PM where they stand in this new world.
So much of what a product can do moves with the model. The moment a model unlocks a capability, a bunch of features you couldn't ship before suddenly open up. So half the job now is knowing what the models you have access to are actually good at, and squeezing the most out of them. Do that well and you iterate way faster and keep pushing your own limits.
My favorite scene in The Shawshank Redemption is everyone drinking beer on the rooftop in the evening sun.
Andy is the link between everyone else and hope.
@lichinlin I really like the post-run summary page. I’d much rather end a run with a photo I took that captures how I felt in that moment, instead of something predefined by the app.
Is outfit styling really the most important thing an AI wardrobe should do?
I think getting dressed is a lot like doing makeup. It’s part of the creative process. AI doesn’t need to do that for you. It can help you figure out what you actually like, refine your taste, and give you suggestions when you need them.
@joshelman I think one of my first apps was Talking Tom. I was a kid back then and obviously don’t use it anymore. Funny enough, I’m now building something kind of like an interactive Talking Tom that manages my Pomodoro sessions and interacts with my todo list.
AI still hasn’t fully translated everything it can do into everyday situations that feel obvious to normal people.
Maybe that’s the real value of agents:
translating AI capabilities into life situations people already understand.
And maybe the harder question isn’t how powerful AI becomes, but how naturally that power can fit into people’s lives.
Maybe agents are helping AI fit more naturally into everyday life.
I recently saw a creator surprised by how many books in Japanese bookstores were teaching people how to use Claude Code.
Her post got around 600K views on X. It made her realize how huge the information gap around AI still is.
Maybe this is one of the things agents can help with.
Helping users realize:
“Oh, I can actually give this to AI.”
A good agent can do a lot of the cognitive work first.
You don’t need to know what a prompt is.
You don’t need to figure out the perfect way to ask.
I’m building an agent for pet owners and just found something pretty wild about pet insurance:
A lot of people only find out during a claim that some random vet note from years ago, even something like tartar, can affect coverage.
By then it’s too late to change it, so you’re stuck appealing and going back and forth with insurance.
Has anyone dealt with this?