My kid said he wanted to do origami.
I sent my Agent a single sentence asking it to handle the origami materials. It searched for tutorials on its own, printed out the origami guide along with sheets of paper. Later I wanted a different color, so I told it again, and it reprinted everything.
From "I want to do origami" to having the materials in hand — under five minutes.
In that moment, I felt that what changed might be more than just efficiency.
My kid said he wanted to do origami.
I sent my Agent a single sentence asking it to handle the origami materials. It searched for tutorials on its own, printed out the origami guide along with sheets of paper. Later I wanted a different color, so I told it again, and it reprinted everything.
From "I want to do origami" to having the materials in hand — under five minutes.
In that moment, I felt that what changed might be more than just efficiency.
The origami isn't an isolated case. Lately I've noticed that the thing my Agent eliminates the most in my life is "middle steps."
Market stall supplies. Before, when I received supply lists for my stall, things were scattered. Now the whole flow is connected: information comes in, gets automatically added to my shopping list assistant (another Agent of mine), and the items are already waiting in the shopping cart.
Two scenarios, same pattern: before, you do N steps to get the result. Now you do 1 step — say "I want this."
I call those eliminated N-1 steps middle management cost.
It's not money. It's your attention and the number of decisions you make.
Every time you open a search box, filter results, compare options, make a decision — your attention gets drained. Individually it's not much, but accumulated over a day, you feel "tired but haven't really done anything." That's the hidden damage of middle management cost.
The logic of traditional tools is: I give you better tools, you do the work yourself. Better search engines, faster printers, more convenient shopping apps. But no matter how good the tools are, you're still the one running around in the middle.
The logic of an Agent is: You tell me what you want, I'll handle it.
Shift from humans adapting to tools, to tools adapting to humans.
Good question. Because the essence of automation is also "you just say it, the machine does it."
The traditional automation I'm talking about here is mainly pre-written rules: if a message contains "market stall," execute script A. But when I had my lobster Agent handle origami, it wasn't executing a script called "origami workflow." It recognized the intent — "the kid wants to do origami" — and searched for materials and printed them on its own.
In my system, Skills are the hands and feet — search, print, add to cart — they execute actions. The Agent decides whether to call them. An Agent adds role positioning, domain memory, and accumulated experience of what worked and what didn't.
What I really want to keep exploring is: when the task changes from origami to something else, can it dynamically combine existing tools on the fly, instead of waiting for me to write a new workflow?
The cost? Token burn. Every instance of understanding, judgment, and decision consumes tokens — more expensive than executing a preset script. But what you get in return is: you only need to make one decision — knowing what you want.
Back to the beginning. My kid wanted to do origami. I sent my Agent one sentence.
This might be my biggest bet on Agents: Middle steps will be eliminated. "I want" will become the new interface.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。