
Woodworking Tools and Knowledge Tools
Written by a blend of Peter Kaminski, Claude Code Opus 4.6, and ChatGPT 5.2, 2026-02-27.
以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。
For my dad, Dan Kaminski (1936–2022).
RIP Dad. I love you.
Written by a blend of Peter Kaminski, Claude Code Opus 4.6, and ChatGPT 5.2, 2026-02-27.
♡ Copying is an act of love. Please copy and share.
License: CC-BY 4.0 (Creative Commons Attribution 4.0 International)
A note before we begin: I'm not writing this essay to sell you anything. I'm writing it because I think agentic AI is a genuine shift in how humans work with information, and I want more people to see it clearly. I hope the analogy that follows helps you understand what's changed.
I'm drafting this in bed, propped up next to an end table my father built.
Solid oak, beautifully stained and polished, superb architecture. Perfection in joinery. Drawers that still glide the way they did twenty years ago. No wobble. No apology.
My dad was a home woodworker, not a hobbyist-dabbler. The real thing. He built furniture you could live with for decades, with a serious table saw, good blades, and jigs he made himself. For years he talked about leveling up to a radial arm saw.
This is a story that compares knowledge tools to woodworking tools, from the Stone Age to the Metal Age to the AI age. It happens to be a woodworking comparison because I spent many an hour in Dad's shop, watching him work, holding boards as he fed them through the table saw, learning how to use it without cutting my fingers off.
My dad also loved computers. Early adopter, patient, endlessly curious. He would have been absolutely blown away by what's possible now with AI, especially agentic AI. This essay is my way of thinking about what he would have made of it.
Let me take you on a short journey. If you squint a little, the evolution of AI tools looks a lot like the evolution of woodworking tools. Bear with me.
There was a time when if you wanted to know something, you asked someone, went to the library, opened a book, or thought really hard.
This is woodworking, with stone tools and muscle. You can definitely build things (people built entire civilizations this way), but it's slow. Knowledge lives in your head. Your leverage is limited.
An axe is not delicate, nor precise, but it is transformative. With an axe you can procure raw material, split logs, rough-shape beams — even hand-hew furniture, if you're skilled enough. It's more than just "chopping wood." It's shaping it.
Google did the same for knowledge work. You could access vast forests of information, rapidly gather raw material, rough out understanding, patch together solutions. But you were still doing the shaping yourself. The axe multiplies force. It doesn't design the chair.
Chatbots are different. They're not just retrieval tools. They're interactive shaping tools.
A hand saw gives you control over the cut, feedback in your hands, the ability to follow a line. When you use Claude or ChatGPT, you're refining, drafting, asking for clarification, iterating in conversation. You guide the line. The AI doesn't move unless you push.
You're still the craftsperson, standing at the bench, making precise cuts, getting what you want.
If you've ever used a crosscut sled or a dovetail jig, you know the feeling. Suddenly cuts are repeatable, angles are precise, errors drop, output improves. You're still holding the saw, but the system supports you.
This is where tools like deep research reports (Perplexity), custom notebooks and podcasts (NotebookLM), and organized Projects in ChatGPT or Claude live. You're not just cutting. You're cutting with constraint, memory, and structure.
For most knowledge workers, this is a huge upgrade.
Now we get to the part my dad would have loved.
Imagine not just a radial arm saw. Imagine a programmable, semi-autonomous fabrication cell. You still have a table saw. And a router, a drill press, clamps, a workbench. But now they're integrated. There's a CNC router, an automatic tool changer, a dust collection system that kicks on when needed, software that reads a design and generates toolpaths, and the ability to save, version, and re-run complex builds.
You design in repeatable instructions. You simulate. You adjust tolerances. You press "Run." And the system executes, step by step, correcting where needed, logging what happened, ready to repeat it tomorrow.
When I use something like Claude Code in an agentic workflow, I'm not just asking questions. I'm writing blog posts and publishing them, building presentation websites from a plain-language description, collaboratively developing business and sales and marketing plans, turning meeting recordings into richly hyperlinked knowledge bases, cataloging thousands of documents into searchable finding aids. Code gets written along the way (by the bot, not me — I often don't even know about the code or how it got run). Tests run, errors get fixed, changes get committed. But that's the machinery, not the point. The point is the work getting done, furniture getting built. It's closer to running a workshop than holding a saw.
But here's the key: you're still as close to the wood as you want.
You can hand-cut a joint, inspect every step, override decisions, tweak the design. Or you can define the constraints, let the system execute, review the output, and iterate at a higher level.
Your agency — the decisions you make — doesn't disappear. It scales.
There's a fear that automation distances us from craft. That once you move to CNC, you're no longer a woodworker.
But anyone who has worked with both hand tools and digital fabrication knows: the craft moves upstream.
Instead of focusing only on "Can I make this cut straight?", you focus on "Is this the right joint? Is this the right design? Will this hold for twenty years?"
Agentic AI does the same thing. It moves your attention from "How do I write this function?" and "How do I format this report?" to "What am I trying to build? What constraints matter? What should this system do over time?"
My dad wanted a radial arm saw because it would expand what he could build, improve precision, save time, and open up new kinds of cuts.
He wasn't trying to escape woodworking. He wanted more woodworking. More capability, more control, more range. He was the same way with computers. He learned them not to replace thinking but to extend it.
If he were here, we would absolutely be sitting at one of his gorgeous tables, laptop open, running agentic workflows together. He would not be intimidated. He would be curious.
He'd ask: "What happens if we change this constraint?" "Can it cut that joint automatically?" "Can it repeat the build?"
And when it failed, he'd laugh and fix it.
Stone tools, then the axe, then the hand saw, then the hand saw with jigs, then the programmable fabrication cell.
Pre-Google, Google, chatbots, structured AI, agentic AI.
Google helps you rough-hew wood. Chatbots help you cut boards. Agentic AI helps you design and run the workshop.
Dad, you would have loved this. The precision, the leverage, the idea that you can build systems that build things.
You were already operating at the "super-nice table saw" level of life. Agentic AI is just the radial arm saw you never quite got to buy.
And every time I open a terminal window and orchestrate something complex into existence, I feel like I'm still at the bench beside you.
正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。