返回资讯中心
外部精选
软件工程
#开发者实时流

Lego Mindset vs. Woodworking Mindset (2022)

"If you’re efficient, you’re doing it the wrong way. The right way is the hard way. The show was successful because I micromanaged it—every word, every line, every take, every edit, every casting. That’s my way of life.…

Hacker News 最新andsoitis2026/10/17 · 2026/10/1730 分钟阅读

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。

Scott Stevenson

Dec 31, 2022

59

7

3

Share

"If you’re efficient, you’re doing it the wrong way. The right way is the hard way. The show was successful because I micromanaged it—every word, every line, every take, every edit, every casting. That’s my way of life." -Jerry Seinfeld

When we were building Mune, I was lucky to collaborate with an incredible woodworker named Don Taylor. Don built world-class acoustic guitars, among other things.

I spent some time with him in his workshop while we figured out how to bring our mere 3D design into the real world.

One afternoon I asked him, “what’s your secret?”, and his response stuck with me for a decade.

Don pointed to a guitar-in-progress hanging on the wall:

“See that? That guitar broke in half last week!

When the neck broke, I didn’t fret for a second. Not one bit. I just glued the guitar back together without missing a beat.

The secret is to learn not to let that kind of thing bother you at all, and to see it as a normal part of the process.”

Don’s mentality wasn’t just about accepting accidents. Every painful manual step, from measuring, to cutting, sanding, gluing and staining, he accepted without a hint of regret or frustration. And he would keep redoing things until he got them just right. I had never encountered that kind of patience before.

Don calmly accepted the manual toil required to create an amazing product.

I grew up with Lego bricks that snapped together, and with products that behaved predictably. Woodworking overwhelmed me. Every single piece of wood was unique and needed to be treated as such. Nothing was certain.

As an engineer, I loved creating predictable, modular systems. I longed to feel the snap of Lego.

We can expect all sorts of things in life to snap together, with neat little checklists:

Tool Syndrome: “If I use the right tools for the job, I’ll do well!”

RFP Syndrome: “If we buy the software that checks these 10 boxes, our problems will be solved!”

Method Syndrome: “If we follow The 12 Step Process, our organization will run smoothly!”

Fill the Seat Syndrome: “If I hire a VP Sales, our revenue will grow faster!”

But reality doesn’t respond well to this thinking when you’re building something new. Things are always more granular than we hope for. Hidden behind every great outcome lies uncountable hours of manual toil, sanding, gluing and polishing.

Scott Stevenson@scottastevenson

When the tasks are completed, and the desired outcome has not been achieved, the real work begins

It is relentlessly working on the mortar between the bricks, the sanding and gluing of the wood, that really matters

When you take that part seriously, you’ll be taken seriously

1:52 PM · Dec 19, 2022

3 Reposts · 20 Likes

It Shouldn’t be Like This

We can feel absolutely miserable when the world doesn’t respond to us like Lego.

“I must have made a mistake, this is so inefficient.”

“This thing is rotten at the core and there’s no saving it.”

After digesting Don’s words and spending a couple years working on Rally, I realized that those words, “it shouldn’t be like this” were an enormous source of stress. A drain on energy and emotional resources.

But, it was the third startup I had been part of, and I noticed the repetition of the same old problems:

Ad campaigns didn’t work as well as we had hoped

Users had surprising feedback, requiring tons of product improvements

Sales cycles were longer than anticipated

And with that, the gut-churning anxiety of foundership melted completely.

“It should be like this. The manual toil is the game.”

“Rather than playing the game, I’ve been doing everything I could to run away from it!”

When I accepted the inevitability of the manual cutting, sanding and gluing required to make something great, everything became fun rather than anxiety-inducing.

I could finally get good at that game, rather than constantly trying to run from it.

We could become master woodworkers like Don.

Lego Mindset might sound like laziness. Sometimes it is. But it’s deeper than that. It often arises from the belief that legible systems are best. That they are what experts build.

In Venkatesh Rao’s seminal piece on legibility, he discusses how humans have a tendency to look at chaotic systems, determine that they must be flawed, and then desire to re-create them in a more legible form.

He pulls an example of failed “scientific forests”:

He goes on to cite a common failure recipe:

How often have you seen this pattern in software?

George Hotz 🐀@realGeorgeHotz

@lexfridman @elonmusk Would you be okay with not shipping any features for 6 months while we refactor the code?

11:57 PM · Dec 18, 2022

112 Reposts · 3.4K Likes

George Hotz 🐀@realGeorgeHotz

@firasd @blueitserver @bscpredict Wait does anyone really doubt that me and 20 great people could rebuild Twitter in its current state in 5 years? I bet we could do it in one!

4:29 AM · Dec 24, 2022

2 Reposts · 31 Likes

Even tech influencers like George Hotz believe in the “High-Modernist Recipe”. And he’s not totally wrong. Software developers need to refactor and tame complexity to stay productive. And sometimes do full re-writes (that’s part of toil). Yet, despite his confidence, he walked away from Vicarious after 6 months, walked away from his own startup before achieving his goals and walked away from Twitter after a few weeks. Three premature walkouts in a row! I bet I know what he was thinking:

But not only is illegibility an acceptable trait of optimal systems. It is a trait that all systems optimized against complex realities will tend towards. The real world is complex and extremely high-dimensional. When we see a successful company that’s chaotic, we shouldn’t think: “wow, it succeeded despite chaos!”, we should think: “Of course! This is what a complex, optimized system looks like: an illegible forest.”

Scott Stevenson@scottastevenson

As neural networks better at chess, they become more illegible: you can't explain why they make the moves they do in words

This will happen in the world at large as we approach singularity: you won't be able to explain in words why anyone does what they do

4:52 PM · Oct 29, 2022

184 Reposts · 1.82K Likes

We believe in the power of Legos because they make our output and process legible. And we believe legible solutions are better solutions.

Manual toil & massage add a lot of illegibility to a process which is hard to describe in a task list. It’s hard to know if massaging is complete. It’s hard to explain the massaging process to others. It adds fuzziness and non-determinism. It is painful if you think it’s a symptom of bad process.

I’ve probably spent hundreds of hours of my life massaging CSS. I couldn’t really tell you what I was doing or why. It’s mostly trial and error until the output feels great. It’s always tempting to believe that a Lego-like Design System can totally remove this toil. It is more legible and feels professional to build components in a Design System than to do style hacking. But I still haven’t seen a truly great product where a rigid design system was obeyed. Or if it was obeyed, it was built after someone style-hacked their way to something that worked. There are just too many edge cases. The eye is too fickle.

Karri Saarinen@karrisaarinen

This kind of mirrors the philosophy we have for @linear and Linear Method:

Focus your efforts on the actual task on hand (building software), not on side quests (building process or management systems).

9:22 PM · Aug 27, 2022

7 Reposts · 189 Likes

Since object-level woodworking problems are annoying and illegible, many of us perform a clever sleight of hand:

“Rather than wasting time on one-off problems, I design scalable systems to solve many problems.”

I’ve interviewed a few hundred job candidates in sales, marketing, design & engineering, and over one third of those have led the interview with this as their core value proposition. I don’t think people realize that this skill is so popular—and I don’t think they realize how extraordinarily rare it is to celebrate manual toil.

There is, of course, nothing wrong with designing scalable systems. It can be brilliant. But it is hard to do right unless you’ve built unscalable solutions first. And at the end of the day, someone needs to passionately solve nitty-gritty object-level problems. That part is important. And few people want to do it.

Scott Stevenson@scottastevenson

Only focusing on system-level problems is a coping mechanism to escape from much more annoying object-level problems

12:01 AM · Dec 11, 2022

2 Reposts · 25 Likes

A lot of people think that nitty-gritty object-level problems are only for juniors, that to “level up” they must migrate to the systems level. But, in my experience, it’s the most senior people that are thinking about the nitty-gritty details. Because they have a greater responsibility to force outcomes. Forcing outcomes requires painful woodworking:

If we accept the inevitability of toil, how do we then get good at it?

At the risk of being overly legible, there are two main components for me: flow and taste.

The micro-actions required to perfectly sand a piece of wood are too numerous to coordinate methodologically. You can’t put them in a todo list or a Kanban board. The only way to do 1000 tiny actions is to lose yourself in the task. To not be there. To be in a state of No Mind, organically responding to what the object of your creation requires. Whether the object is an entire company, or a UI component.

For me, I need to schedule big chunks on my calendar for these “sanding tasks” that are too numerous and tiny to put on a list. I commit fully to getting inspired and to seeing what happens.

It also helps to exercise, to sleep well, eat well, to drink coffee. It takes a lot of raw energy and inspiration to execute 1000 micro-actions.

The second component is developing good subjective taste. Polishing is inherently subjective. How do you know when to stop? How do you know what the goal is? Taste is the answer. You can develop it in two ways:

  1. Consume great work & analyze it: Why do you like that song? What kind of music gets popular? What does the CSS look like for that app you love? If you can afford it, use Tier A, design-first products to set your bar high.

  2. Practice and evaluate your own work: Learning to make good things takes a lot of practice. You will make a lot of bad things at first, and then you will notice why they are bad later. Keep trying, keep iterating!

After working with large language models like GPT-3 at Spellbook, I’m convinced that software development is about to become even more like woodworking. LLMs are non-deterministic, especially once you start chaining calls together. Exact correctness gets traded off for flexibility and nuance. API calls become paragraphs. UI becomes conversation.

LLMs are anything but nice predictable lego bricks. They’re organic components that need to be sanded until they feel right. Building prompt chains feels more like CSS massage than programming—with some added non-determinism.

It’s hard to logically test these systems, or to configure them using purely quantitative methods.

Those who master woodworking will be at an even bigger advantage in the next decade of software development. These systems are incredibly powerful, but they require Don-like patience to polish.

It does. The reality is, most customers will not develop the patience for manual toil. That patience will become increasingly rare.

That creates an opportunity to create Lego-like products for customers. Consumer electronics will tend towards Airpods + iPhone + Macbook style interoperability. Users will still demand correctness from non-deterministic AI products. The appetite for legibility will grow.

Makers will never live in a world of Magic Lego Bricks, but, through tireless toil, we can create the illusion of Magic Lego Bricks for our customers.

Lastly: this post itself is over

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。

打开原文
去社区讨论 →