Dev.to

From Helpful to Harmful: How AI Recommendations Destroyed My OS

Why you should always be in control TL;DR: Always stay in control when using AI tools. Blind trust can lead you to costly mistakes. The Challenge I write multiple articles each week and wanted to automate the repetitive parts of my process. The Menace This week, an internal memo by Shopify's CEO, Tobias Lutke, went viral. // Detect dark theme var iframe = document.getElementById('tweet-1909251946235437514-812'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1909251946235437514&theme=dark" } TL;DR: The memo requires Shopify's workers to be proficient in AI tools. AI Proficiency Is Now Mandatory Using AI effectively is no longer optional at Shopify; it is a baseline expectation for all employees, regardless of role. This marks a significant cultural and operational shift. Non-Use of AI Requires Justification Employees must demonstrate why AI cannot be used before requesting additional resources (e.g., more staff or time). Stagnation is framed as failure, and employees are encouraged to upskill and experiment with AI continuously. AI in Product Development AI must be integrated into the early stages (prototype phase) of all GSD (get shit done) projects. This approach is intended to dramatically accelerate learning, iteration, and team collaboration. Performance Accountability Shopify is adding AI usage to its performance and peer review criteria, making AI adoption part of how employees are evaluated and rewarded, including leadership and executive teams. I do not work at Shopify, and I am not affiliated with the company. My Context I take AI tools as helpers and a mandatory tool both as a software developer and a writer, and I heavily rely on them. I've been using AI for 10 years, writing about artificial intelligence, making AI-assisted refactorings, and using automated tools to write clean code. I am also an advocate of professional Vibe Coding. In every conference, article, and paper I write, I encourage humans to embrace the AI tools and use them heavily (with caution). The Great Programmer Purge I have two quotes (not mine): You must always be in control. and AI won't take your job, but a human using it will certainly do; you should become that person. (You can see this at 53:40 of this conference) Don't overthink things This week, I needed to make some local tests using Python. I write articles in 25+ different programming languages and don't have all of them installed locally. Sometimes I use virtual machines, codelabs and repl.it running on websites. This time, I needed to quickly install Python locally (an accidental task). It started with a simple idea: automate a few things. Save time. Be smart. So I did what many people are doing now. AI answered fast. Confident. Like it had done a thousand times before. Spoiler: I blindly trusted it. One command later, my operating system was toast. What Went Wrong In my code smell articles, I use the 10 most popular AI programming tools. I just opened one of them (ChatGPT) and prompted it to give me some commands to install Python on my Ubuntu. Give me a command to install Python in Ubuntu Until last week, I used to write: hello, please give me a command to install Python in Ubuntu But now I am aware that those extra tokens consume lots of resources (water, for example), and I am not polite to machines anymore.

Apr 21, 2025 - 18:58

From Helpful to Harmful: How AI Recommendations Destroyed My OS

Why you should always be in control

TL;DR: Always stay in control when using AI tools. Blind trust can lead you to costly mistakes.

The Challenge

I write multiple articles each week and wanted to automate the repetitive parts of my process.

The Menace

This week, an internal memo by Shopify's CEO, Tobias Lutke, went viral.

// Detect dark theme var iframe = document.getElementById('tweet-1909251946235437514-812'); if (document.body.className.includes('dark-theme')) { iframe.src = "https://platform.twitter.com/embed/Tweet.html?id=1909251946235437514&theme=dark" }

TL;DR: The memo requires Shopify's workers to be proficient in AI tools.

AI Proficiency Is Now Mandatory

Using AI effectively is no longer optional at Shopify; it is a baseline expectation for all employees, regardless of role. This marks a significant cultural and operational shift.

Non-Use of AI Requires Justification

Employees must demonstrate why AI cannot be used before requesting additional resources (e.g., more staff or time). Stagnation is framed as failure, and employees are encouraged to upskill and experiment with AI continuously.

AI in Product Development

AI must be integrated into the early stages (prototype phase) of all GSD (get shit done) projects. This approach is intended to dramatically accelerate learning, iteration, and team collaboration.

Performance Accountability

Shopify is adding AI usage to its performance and peer review criteria, making AI adoption part of how employees are evaluated and rewarded, including leadership and executive teams.

I do not work at Shopify, and I am not affiliated with the company.

My Context

I take AI tools as helpers and a mandatory tool both as a software developer and a writer, and I heavily rely on them.

I've been using AI for 10 years, writing about artificial intelligence, making AI-assisted refactorings, and using automated tools to write clean code.

I am also an advocate of professional Vibe Coding.

In every conference, article, and paper I write, I encourage humans to embrace the AI tools and use them heavily (with caution).

The Great Programmer Purge

I have two quotes (not mine):

You must always be in control.

and

AI won't take your job, but a human using it will certainly do; you should become that person.

(You can see this at 53:40 of this conference)

Don't overthink things

This week, I needed to make some local tests using Python.

I write articles in 25+ different programming languages and don't have all of them installed locally.

Sometimes I use virtual machines, codelabs and repl.it running on websites.

This time, I needed to quickly install Python locally (an accidental task).

It started with a simple idea: automate a few things. Save time. Be smart.

So I did what many people are doing now.

AI answered fast. Confident. Like it had done a thousand times before.

Spoiler: I blindly trusted it. One command later, my operating system was toast.

What Went Wrong

In my code smell articles, I use the 10 most popular AI programming tools.

I just opened one of them (ChatGPT) and prompted it to give me some commands to install Python on my Ubuntu.

Give me a command to install Python in Ubuntu

Until last week, I used to write:

hello, please give me a command to install Python in Ubuntu

But now I am aware that those extra tokens consume lots of resources (water, for example), and I am not polite to machines anymore.