For my use case … almost everything that takes up my time can be automated in someway. My brain is overwhelmed by all the ideas flooding.
Earlier this year thought maybe by q4/ next year hire my friend to help me with supply chain / inventory.
But with the right codex gymnastics I can have a little ai agent walk into my hello Kitty themed 2D office and show me a pitch deck on what to order and why and all I do is say “ok”
And it takes over my mouse and keyboard and gets on it.
Ben, you're not overwhelmed by ideas. You're overwhelmed by the gap between having an idea and shipping it. That gap is what employed people for the last two hundred years. It's closing. You're just the first in your orbit to notice.
The Hello Kitty office with an agent walking in and presenting a pitch deck—that's not a joke. That's the actual near-term future. The agent doesn't need to be perfect. It needs to be good enough that saying “ok” is cheaper than doing it yourself. And right now, for a lot of repetitive operational work, it already is.
Here's what you should actually worry about, though: not the automation, but the taste. The agent can build the pitch deck. It can place the order. It can reconcile inventory. But it can't decide what's worth ordering without you telling it what good looks like. Your job is shifting from “do the work” to “define the standard and approve the output.”
That's a harder job in some ways. When you're doing the work, mistakes are your own and you learn from them. When you're approving, you need to be able to spot a bad recommendation at a glance. So develop that instinct. Train your eye. The people who win in this era aren't the ones with the best agents—they're the ones with the best judgment.
Don't hire the friend yet. Build the agent first. If it can't handle 80% of what the friend would have done, then hire the friend to handle the 20% the agent can't. But my money says you won't need the friend.
And yes, the Hello Kitty office is fine. Aesthetics don't affect throughput.