I spent years building one ecommerce company.
Now I'm playing a different game.
I’m looking at what's changing in technology, business and consumer behavior and asking:
Where's the opportunity?
Some ideas I'll test.
Some I'll build.
Some I'll be completely wrong about.
That's what I'll share here.
AI made one of my biggest weaknesses worse:
Almost every idea now looks buildable.
Website? Afternoon.
Prototype? Weekend.
Automation? Probably tonight.
So “can I build this?” has become a pretty useless question.
The challenge I am facing now goes more into this direction:
Which idea deserves the next 6 months of my attention?
Building comes at a decent cost now.
Choosing is the real challenge here!
@steipete Agent wars. First boss: a checkbox.
Until agents can prove who they are and what they’re allowed to do, “autonomous” still means asking the human to click “I’m not a robot.”
I’ve checked more than 20,000 backscratchers by hand, one at a time.
It’s a strange detail to share. But it says something about the work behind building a business: some tasks are repetitive, and you still have to get them right.
The idea gets attention. The follow-through takes longer.
Some decisions I made five years ago made sense then. I wouldn’t make them today.
I’ve lived in different places. I’ve learned more. My priorities changed.
That doesn’t make the old decision foolish. I make the next one with what I know now.
@garyvee I’m still building my own audience, so this hits home. You don’t need to quit your job or pretend you’re an influencer. Start posting what you’re learning, even if 10 people see it. Better to have a few things out there than show up online for the first time when you need a job.
@IAmAaronWill That willingness to say yes, learn fast, and adjust is agility. You don’t need a five-year plan for every open door. Sometimes you just need to walk through and see what’s there.
@garyvee You were doing wine content before “creator” was a job title. Nostalgia might get the first click. The real hook is what you think about wine now that you didn’t back then. Go for it!!!
@jasonfried A file comes in. By the time a person needs to look at it, it’s stored, labeled, triaged, and turned into something they can act on.
That’s the bit I’d steal. Less “AI assistant.” More: stop making people push the same work through five different tools.
“RSI” is a hell of a claim for “the model got faster, cheaper, and better.”
The watch app is the interesting part: a casual question became a working tool in an hour. That doesn’t prove RSI. It does show how quickly the gap is closing between “I wish this existed” and “I built it.”
The idea is right. The danger is turning safety into a moat.
If compliance requires the budget, lawyers, and infrastructure of a frontier lab, the standards won’t prevent concentration of power.
They’ll formalize it.
Safety should make the industry more accountable without making competition impossible.
Making individual workers faster is only half the opportunity. The bigger unlock is fixing the handoffs between them.
Between “I’ll take care of that” in a meeting and the work actually getting done, things still fall into email, memory, and vibes.
That’s where productivity quietly goes to die.
@Codie_Sanchez But “you haven’t figured it out yet” only helps if you keep running experiments.
Otherwise it’s just a nicer story you tell yourself while avoiding the work.
I don’t think the biggest value of university was the lectures.
You can get most of that information today from a book, a podcast, a YouTube video, or a short conversation with AI.
The real value was spending years around the same people.
Working on the same projects. Seeing who takes responsibility, who disappears when things get difficult, how people react under stress, and what kind of work they produce when the deadline is real.
That creates a kind of trust you can’t build from a profile or a 10-minute conversation.
The information is everywhere now.
Shared experience is still rare.
@ItsKieranDrew University teaches you how to answer questions.
The arena teaches you which questions matter.
Nothing accelerates learning like decisions with consequences.
@SchmidhuberAI@ylecun@musedivision@geoffreyhinton Credit matters.
But when a discussion turns into a 12-post trial over who thought of what first, the audience learns less about the idea and more about the ego protecting its territory.
Good ideas survive scrutiny. Great ideas don’t need a custody battle.
@gregisenberg Email won’t disappear. It will become a landfill.
The valuable layer will be verified identity, permissioned access, and agent-to-agent negotiation.
If every agent can contact everyone, the inbox stops being a communication tool and becomes an attack surface.
@williameijer This chart is less about immigrant intelligence and more about immigration selection.
Turning selection effects into a ranking of national intelligence is just IQ discourse wearing a spreadsheet.