運用AI開發,避免產出劣質品

運用AI開發,避免產出劣質品

Hacker News·

本文主張在AI輔助開發中採取更審慎的態度,強調產品判斷與品質勝過單純的快速建構。文章警示我們將被大量易於產出卻無價值的AI生成內容淹沒,並呼籲轉向有意義的創作。

Image

Image

Build With Attitude ✊ - An Open Invitation To Vibecoders Who Care About Building Things That Matter

AI made building easy. 2026 will decide if we make it meaningful.

Image

Image

Image

Image

We’re about to drown in things that were easy to build and never worth shipping.

Not for the user. Not for the builder.

In 2025, the message around AI was simple:

Look what I built! Here’s how you can build it too.

And that message still matters, it’s key to encouraging beginners to start experimenting.

But as we enter 2026, that’s no longer enough.

It’s time to graduate from:

Look what I built! Here’s how you can build it too.

to:

Look what I built! Here’s how you can build it too, and how I think along the way.

Image

Image

AI Product Manager, builder of StackShelf.app and Attitudevault.dev on a mission to create value, not AI slop.If you’re new here, welcome! Here’s what you might have missed: 2025’s Most Absurd Product Decisions How To Product-think When AI Builds At Lightning SpeedThe Ultimate Vibecoding Guide From Builders Who’ve Shipped

Subscribe 🤗

Why With Attitude ✊

Image

Because building with AI is getting easier, but thinking well is not.

AI can help us generate code, screens, flows, and features at unprecedented speed.

What it can’t do for us is decide what’s worth building, what isn’t, and where value actually lives.

Lovable, Replit and Cursor - while they’re all impressive, high-leverage and value-enabling tools, they’re just that - tools. And as such they can’t (and shouldn’t!) replace our own critical thinking.

That’s where our judgment comes in.

That’s where product thinking comes in.

And that’s what Build With Attitude is about.

In Xian’s words:

Image

What We Mean By AI slop

In this series, AI slop doesn’t mean “AI-generated.”

It means output without judgment: mass-produced, low-quality, and unwanted output (content, code, products, videos) created because it can be generated, not because it should exist.

It’s what happens when execution is automated but responsibility is not.

Image

The Snake-Eats-Its-Own-Tail Loop

A super clear explanation comes from Kurzgesagt, genuinely one of the best learning channels on YouTube, whether you’re 8 or 48.

In their 2025 video AI Slop Is Destroying The Internet, they describe a golden era of soulless slop: low-effort garbage flooding every platform simply because it costs nothing to produce.

Text, images, videos, books, news. All of it padded out with synthetic filler that looks confident, sounds plausible, and means almost nothing.

The most unsettling part is the loop:

Slop gets published ➜ Slop gets scraped ➜ Slop gets fed back into the next generation of models ➜ People start treating it as “knowledge.”

The snake eating its own tail.

I see a similar loop emerging in AI-generated products:

Image

Image

What This Series Actually Is

Each edition centers on a real builder and a real project: not only as a case study, but as a thinking artifact.

We don’t just look at what was built. We examine how decisions were made, where human judgment was key, and what tradeoffs looked like in practice.

The goal is to learn from the hundreds of big and small decisions builders make as they try to create value, rather than contribute to AI slop.

The moment the idea became worth pursuing

The trade-offs that shaped the product’s character

The features that were deliberately left on the cutting board

Where AI accelerated execution, and where it misled

The judgement calls no model could make on our behalf

Image

Who This Invitation Is For

You don’t need a finished product or a perfect story.

You do need curiosity, a willingness to show your thinking, and the courage to talk about what didn’t work.

This invitation is for builders who’d rather be honest than impressive,who believe the most valuable thing they can share isn’t polish, but lived perspective.

Image

Why I See Contributors as the Most Glamorous Builders

This series wouldn’t exist without people willing to admit how they think, and where they got it wrong.

That’s hard enough to do in front of a bathroom mirror, harder in a meeting room, and simply brave in public.

It’s far easier to share a confident takeaway, or one of those “I built this over the weekend and made 10k” stories that travel well on social media and require no follow-up questions.

It’s much harder to share uncertainty. Second thoughts. Or the humbling moment when you realize that the thing you were so excited about matters to no one.

So every time I see a creator doing that publicly, I nod with respect, knowing something important happens:

They help someone who’s stuck realize they’re not broken

They help someone navigating a career shaped by AI see where to think critically

They help us hold the line against AI slop.

Image

The Builders Kicking Things Off

Here are some of the wonderful people who have already expressed interest in joining the series (or whom I’ve personally invited).

Image

Image

How To Participate

This is an open invitation. Anyone building with AI is welcome to participate, just fill out the interview form below:

👉 Interview Questions

Afterwards, I will:

DM you with a suggested publish date

Synthesize your answers into a narrative post

Send you the draft for review

1–3 days before publishing I’ll also send you shareable snippets, so we can share with our network

Once the post is published, I’ll add you to the Builders With Attitude 2026 poster.

Image

Share

You Might Also Enjoy

👉 Vibecoding x Cybersecurity👉 Claude Skills Are Taking the AI Community by Storm👉 Why “it feels worse” is the wrong metric for evaluating ChatGPT-5.2

From The Community

Happy New AI Year by Ruben Hassid 2025 Recap: The Year the Old Rules Broke by Michael Spencer and Jess LeãoNo ChatGPT Wrapped in My Region. So I Reverse-Engineered It by AI Meets Girlboss

New On Stackshelf.app

Image

Jose Parreño Garcia added 50+ Data Science Templates ➜ view shelf.

Karen Spinner build two chrome extensions: Substack Reader and Prompt Collector ➜ view shelf.

Join hundreds of Premium Members and unlock everything you need to build with AI. From prompt packs and code blocks to learning paths, discounts and the community that makes it so special.

SUBSCRIBE 🤗

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

Image

This project will be beneficial to the readers AND the builders contributing.

Going through the questions is helping me clarify my thinking on the project.

They are very thoughtful and solidifying the WHY behind my build.

Image

A few things that really landed for me:

  1. "Execution is automated but responsibility is not." That's the whole tension in one line. The tools don't care what gets shipped. We have to.

  2. The shift from "here's how you can build it too" to "here's how I think along the way." I've been spotlighting VibeCoding.Builders for the past few months, and the stories that teach the most are never the clean success stories. It's the decision points, what they kept, what they cut, where they got stuck and why. That's the stuff people actually need to hear.

  3. "It's far easier to share a confident takeaway... It's much harder to share uncertainty." This is exactly why featuring real builders matters. The messy middle is where the learning lives.

Grateful to be part of this series, feels like we're pulling in the same direction. 🙌

No posts

Ready for more?

Hacker News

相關文章

  1. AI 在撰寫 React 程式碼方面的真實表現如何?

    4 個月前

  2. 機器之前的雜訊:為何AI真實性恐慌忽略了重點

    4 個月前

  3. 我與AI一同深入探索

    3 個月前

  4. 無需逐行審查即可信任AI

    3 個月前

  5. 權威性是AI的瓶頸

    4 個月前