Hello friends!
Welcome to this week’s Sloth Bytes. I hope you had a great week.
Quick heads up!
I’ll be posting bite-sized tips, the latest tech news, useful tools, and content that doesn't make the newsletter. If that sounds interesting to you, I’d appreciate the follow.

Here’s how I use Attio to run my day.
Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:
Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find
All in seconds. No searching, no switching tabs, no manual updates.
Ready to scale faster?

ChatGPT loses 4 million users after Pentagon deal
OpenAI signed a deal letting the US military use ChatGPT for classified operations. Anthropic was offered the same deal, but they refused, drawing a hard line against mass surveillance of Americans and fully autonomous weapons.
Understandably people weren’t happy, within 48 hours of this:
ChatGPT uninstalls jumped 295%
One-star reviews spiked 775%
Claude knocked ChatGPT off the #1 spot on the App Store for the first time ever.
This also led to the creation of the “QuitGPT” movement, where currently over 4 million people have pledged to quit using ChatGPT.
Is the job market cooked because of AI? Anthropic looked at the data
Most research on AI and jobs is theoretical. Researchers guess which tasks AI could do and publish scary headlines. Anthropic did the opposite.
They stopped guessing and looked at real usage data, tracking what people actually use Claude for at work across 800 occupations.
Here's what they found:
Computer programmers: 75% of tasks covered by AI
Customer service reps: 70%
Data entry and medical records specialists: 67%
The interesting part is that AI adoption is way below its actual capability.
In computer and math jobs, AI could theoretically handle 94% of tasks. Right now it's only covering about 33%.
There's also been no spike in unemployment for highly exposed workers since ChatGPT launched in 2022. But one number stands out: hiring of workers aged 22 to 25 has slowed roughly 14% in AI-exposed occupations.
This means younger workers aren't getting fired. They're just not getting hired.
Why it matters: Based on this info, we’re not in a job crash. We’re in a slow leak, and that's what makes it dangerous.
TypeScript 6.0 RC: The last JavaScript-based release before the Go rewrite
Microsoft just announced TypeScript 6.0, and it's special because it's the last version written in JavaScript. The next version (7.0) will be completely rewritten in Go, which will make it 10x faster.
Why rewrite it? Because TypeScript checks millions of lines of code, and doing that in JavaScript is slow. Go is built for speed and can use multiple processor cores at once.
TypeScript 6.0 is basically a bridge. It's preparing everyone for the big changes coming in 7.0.
Big changes in 6.0:
Strict mode (catches more bugs) is now on by default.
Old, outdated options are getting removed.
Built-in support for Temporal (a better way to handle dates and times in JavaScript)
New methods for working with data.
If you're using TypeScript and see warnings about deprecated features, fix them now, because 7.0 won't support them at all.

Should you let AI access all your data?

New AI agent tools promise to manage your emails, calendar, and smart home devices via text from anywhere.
The idea is promising:
Text your AI to summarize your emails while running on the treadmill
Having an app idea on the bus and telling your agent to build it
But not everything is sunshine and rainbows. We all know AI is not perfect yet.
Summer Yue (Safety and alignment at Meta Superintelligence, yes… I know) set up OpenClaw and told it to "confirm before acting." Yet it immediately started deleting her entire inbox WITHOUT confirming. Here’s the tweet:
AI taking over, am I right? To be fair, she said she had this workflow working for weeks and was confident it wouldn’t have any problems.
Another Example: Someone gave Claude AI a task to "clean" their computer. Claude deleted 15 years of family photos.
Even with these horror stories, people are going crazy over these agents. People are now buying dedicated Mac minis ($600 btw) just to run AI agents in isolation, protecting their main systems.
Why it matters: AI agents CAN be useful, But right now, they're like giving a really smart toddler admin access to your life. Use them for low-stakes tasks. Don't give them access to anything you can't afford to lose. I’d recommend using a VM or a different machine if you want to use them for more important tasks and maybe wait a bit before letting them control your main computer.
Is Vibe Coding Just Fraudmaxxing?
"Vibe coding" is now Collins Dictionary's Word of the Year for 2025. The term only popped up in early 2025 from Andrej Karpathy, but people got the wrong idea fast: skip the CS degree, skip learning syntax, and just vibe your way to shipping something real.
To test that claim, Stack Overflow writer Phoebe Sajor, who has zero coding experience used an AI tool called Bolt to build a Reddit app for rating bad bathrooms.
It ran and it looked fine. Sounds great, right?
Well... when developer friends reviewed the code, here's what they found:
All styling dumped into one messy file
Massive blocks that should've been split into smaller components
Zero tests
Code that was "ripe for hacking"
Broader research on AI-generated code backs this up, with studies finding security vulnerabilities in roughly 53% of it. It looks like a working app. It's also one bored hacker away from collapse.
Why it matters: Vibe coding isn't fake. It's a real shortcut that gets you from idea to working app faster than ever, but "working" and "production-ready" are two very different things, and most vibe coders can't tell the gap. The smarter play? Use it to learn faster, break things on purpose, and understand why the AI made each choice. Don't just accept the output and ship it.

Want to know how to code your own virtual machine? - learn how to code your own virtual machine
What's next in AI: Five trends to watch - ByteByteGo's AI predictions
8 trends in web dev for 2026 - React Compiler, TypeScript-only mindset, and more
MCP vulnerabilities every developer should know - Prompt injections, auth bypass, and a lot more risks.
Qt Creator 19 released - A popular IDE for C++ just got updated
Rust 1.94.0 released - New stable Rust release (for you Rust lovers)

AuthKit - is like a fancy doorman for your app, it handles all the authentication stuff (passwords, login buttons, that code from your phone) so you don't have to build it yourself.
Paperclip - basically you're running a company where all the employees are robots. Lets you hire a bunch of AI workers (like a CEO, a coder, and a marketer) and organize them into a pretend company where they actually do real work for you while you just approve their decisions.
Context.dev - paste in any website's address and get back its logo, brand colors, a text summary, screenshots, and more, all through a single API. Pretty handy if you're building something and need to pull in real company data without doing it by hand.

That’s all from me!
Have a great week, be safe, make good choices, and have fun coding.
If I made a mistake or you have any questions, feel free to comment below or reply to the email!
See you all next week.
What'd you think of today's email?
Want to advertise in Sloth Bytes?
If your company is interested in reaching an audience of developers and programming enthusiasts, you may want to advertise with us here.







