A Bug Took Down Google Cloud & 58% Devs Want to Quit Over Tech

Netflix’s data strategy & Anthropic on AI sabotage

Hello!

Last week, a single bug in Google Cloud took down parts of the internet, including giants like YouTube and Spotify. It’s since been patched, but the scale of the impact is a reminder that you can never be too careful.

You’ll find the full post-mortem further down, but don’t miss the new Storyblok survey: 58% of developers say they’ve considered quitting because they’re embarrassed by their company’s tech stack. We’re also digging into how Netflix applies Unified Data Architecture across its pipelines, and the message from Amazon’s CEO on how the company’s AI future will look.

Also, Anthropic has released new research on how to detect sabotage from advanced AI agents using a system called SHADE-Arena.

Sounds good? Then grab your coffee and enjoy this week’s Frictionless.

In the Queue

Reduce Friction

A graphic from Storyblok’s survey showing that more than half of developers have considered quitting due to outdated, embarrassing tech stacks.

Source: Storyblok

58% of Developers Are Considering Quitting Their Jobs Because of Inadequate and ‘Embarrassing’ Legacy Tech Stacks

Turns out, 86% of devs are embarrassed about the tech stack they use, and over half of them consider quitting over it. If you haven’t yet, it might be a good time to discuss the tools you use with your team. 

5 Steps for Leading a Team You’ve Inherited

Taking over an existing team is tricky. One wrong move, and you risk breaking trust you haven’t even earned yet. Here’s a 5-step plan for leading with curiosity instead of assumptions.

A chart from LinearB showing that just 7% of decisions in the requirements and architecture phase are made by AI—most are still human-led or mixed.

Source: DevInterrupted

Faster Coding Isn't Enough

AI might help your team move quicker, but speed alone doesn’t build great products. Code reviews are still 51% human-only, showing you we can’t afford to neglect feedback loops or mentorships to stay ready for the future.

Choosing Where to Spend My Team’s Effort

Saying yes is easy; knowing when to push back is usually the problem. Your team's time is money, and so if you need help deciding where to spend it best, have a look at this framework.

Deepen Your Expertise

A grid of logos showing major platforms impacted by the June Google Cloud outage—including Spotify, YouTube, Discord, OpenAI, Gmail, and Shopify.

Source: ByteByteGo

How the Google Cloud Outage Crashed the Internet

A single line of code managed to cause a huge Google Cloud outage, taking down parts of YouTube, Spotify, and more. See what went wrong and how a stray bug turned into a global failure that put a pause on the Internet.

Pros and Cons of Next JS: 2025 Update

Next.js might be my favorite framework, but that doesn't mean it's the perfect tool for every org. Read where it excels, when you might want to consider an alternative, and if you need help with your Next.js stack, I've got just what you need.

A screenshot from Netflix’s internal data tool, showing character entries from One Piece. It highlights how Netflix manages structured content metadata across its platforms.

Source: Netflix Engineering

Model Once, Represent Everywhere: UDA (Unified Data Architecture) at Netflix

Imagine a metadata layer that keeps all your data pipelines speaking the same language - that’s Netflix’s Unified Data Architecture. Their engineering team explains how they pulled this off, offering a clear model for those of us struggling with complexity in data ecosystems.

Boost Performance With React Server Components and Next.js

Need to optimize your stack’s performance? Using React Server Components with Next.js can lower JavaScript payloads, speed up render times, and get your Core Web Vitals back in the green. See how it works in practice.

AI Corner

A photo of Andy Jassy, CEO of Amazon, who recently shared an internal memo explaining how generative AI will reshape roles across the company.

Source: Getty Images

Message from Amazon CEO Andy Jassy: Some Thoughts on Generative AI

While some companies are reconsidering replacing their employees with AI, the Amazon CEO seems to be headed the other way. In the newest memo shared with his employees, Jassy hints that AI adoption will likely reduce the company’s corporate workforce. 

AI Metrics

Now that AI is baked into everything from chatbots to enterprise tools, the question isn’t whether it works, but how to measure its value. Surface-level metrics like WAUs or token counts don’t cut it anymore, so maybe it’s time to focus on real product impact instead.

A flow diagram from Anthropic illustrating how their SHADE-Arena framework tests whether AI agents can execute hidden malicious tasks while avoiding detection.

Source: Anthropic

SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents

As models get more capable, the risks grow too. This report from Anthropic introduces SHADE-Arena, an experimental setup that tests whether advanced AI agents can carry out hidden sabotage without getting caught.

We’re Expanding Our Gemini 2.5 Family of Models

Google’s latest release expands the Gemini 2.5 lineup with stable Flash and Pro models, plus a preview of the lightweight Flash‑Lite. Now officially production-ready, these models offer smarter reasoning, cost-efficiency, and great speed.

Just Cool

A humanoid robot performs in front of a live crowd in China, showcasing the capabilities of RoboBrain 2.0—China’s open-source AI designed for robot navigation and task planning.

Source: South China Morning Post

Beijing Academy Unveils Open-Source ‘RoboBrain’ AI Model for China’s Humanoid Robots

RoboBrain 2.0 is out, and it does exactly what’s in its name. Built to help humanoid robots move, see, and plan better, this open-source AI from Beijing will make sure your future robot butler doesn’t walk into a wall.

Let’s Stay in Touch! 📨 

Do you have any comments about this newsletter issue or questions you want to ask? Drop me a message or book a meeting.

What do you think of today's email?

Your feedback helps me improve Frictionless.

Login or Subscribe to participate in polls.