I'm going back to writing code by hand

Plus: the case for local AI, Bun's Rust rewrite hits 99.8%, and Node.js v26 lands

May 11, 2026

Hey there.

Welcome to Full Stack Focus. 👋

Here's everything you need to know this week in the world of full-stack development.

Your competitors already read this every morning.

The AI Report keeps 400,000+ executives ahead of every major AI move — in 5 minutes a day. Trusted by leaders at the world's top companies. The question isn't whether AI is changing your industry. It's whether you'll see it coming.

Front-End

🐭 Using Claude Code: The Unreasonable Effectiveness of HTML
Turns out the most powerful tool in your front-end arsenal might just be plain old HTML. A developer using Claude Code shares a surprising discovery - that leaning into raw HTML produced results that left heavier approaches in the dust. It's a fun reminder that sometimes the simplest solution is the most effective one. Worth a quick read if you've been reaching for frameworks out of habit.

👀 Why React Developers Are Leaving Next.js for TanStack
The Next.js vs TanStack debate is heating up, and this video breaks down exactly why developers are making the switch. If you've felt frustrated with Next.js complexity, server component quirks, or the ever-shifting App Router paradigm, you're not alone. TanStack is offering a more flexible, composable alternative that's winning people over fast. Watch this before you start your next project.

🐭 From Your Experience, Do You Work Less, More, or Evenly After Using AI?
The community weighed in on a question everyone's been wondering about - has AI actually reduced your workload? The answers are messy, funny, and surprisingly honest. Some devs are shipping faster, others say AI just moved the complexity around, and a few swear they're busier than ever. A quick, fun snapshot of where the developer community actually stands on the AI productivity promise.

Programming

💔 I'm Going Back to Writing Code by Hand
One developer is ditching AI-assisted coding and returning to writing code manually - and the reasons are more compelling than you might expect. The argument isn't about nostalgia; it's about understanding, ownership, and craft. When you let tools generate everything, do you actually know what your codebase is doing? A thoughtful read that will make you reflect on your own workflow and what you might be trading away for speed.

😧 I Returned to AWS and Was Reminded Why I Left
A developer goes back to AWS after time away and immediately rediscovers all the pain points that drove them off in the first place. From the overwhelming console to confusing pricing and opaque abstractions, the frustrations are laid out in refreshingly honest detail. If you've ever had a love-hate relationship with cloud infrastructure, this one will feel very familiar. Cathartic and genuinely useful as a gut-check before committing to a platform.

⚠️ COBOL Is the Asbestos of Programming Languages
COBOL is everywhere - buried deep inside banking systems, government infrastructure, and financial institutions that keep the world running. Wired's piece draws a striking comparison: like asbestos, it's deeply embedded, hard to remove, and carries serious long-term risk. The people who understand it are aging out of the workforce, and nobody wants to replace them. This is a fascinating look at technical debt at civilizational scale.

👀 Programming Still Sucks
Despite decades of tooling improvements, better languages, and AI assistants, programming is still kind of a mess - and this post isn't afraid to say it. The author takes an honest look at why software development remains frustrating, fragile, and full of unnecessary complexity. It's not a doomer rant; it's a grounded take that most working developers will nod along to. Sometimes it helps to just have someone say the quiet part out loud.

⚡ When 'if' Slows You Down, Avoid It
Branchless programming is one of those performance techniques that sounds almost too clever to be practical - until you see the benchmarks. This post walks through how conditional branching can tank CPU performance and how to rewrite logic to avoid it. It's a great low-level deep dive that'll sharpen how you think about performance-critical code. Even if you never write branchless code in production, understanding why it matters is worth your time.

✒️ You Should Read "Programming as Theory Building"
Peter Naur's classic essay argues that programming isn't primarily about producing code - it's about building a mental model, a "theory," of the problem being solved. This write-up makes a compelling case for why that idea is more relevant than ever in the age of AI-generated code. If the developer who built something leaves, so does the theory - and that's a real organizational risk. One of those reads that quietly shifts how you think about your job.

AI

⚠️ Local AI Needs to Be the Norm
Every prompt you send to a cloud AI is a data point someone else owns - and this post makes the case that local AI is the only responsible default. The argument covers privacy, data sovereignty, latency, and cost, and it's hard to disagree with once it's laid out clearly. As local models get more capable, the justification for routing everything through third-party servers gets thinner. A strong opinion piece that's worth sitting with.

🛠 Running Local Models on an M4 with 24GB Memory
If you've got an Apple M4 chip with 24GB of unified memory, you're sitting on surprisingly capable AI hardware - and this guide shows you exactly how to use it. The author walks through setup, model selection, and real-world performance benchmarks for running local LLMs without a cloud subscription in sight. A practical, no-fluff post for anyone who wants to actually try local AI rather than just read about it. Great companion piece to the article above.

🤖 How LLMs Work Under the Hood - Explained with Runnable JavaScript
Ever wanted to truly understand tokenization, embeddings, attention mechanisms, and text generation without drowning in academic papers? This deep dive explains the core machinery of large language models using actual runnable JavaScript code - so you can poke, tweak, and experiment as you learn. It's one of the most accessible technical breakdowns of how LLMs actually work that we've come across. Bookmark this one.

More Tools & Stories

🕺 Mood UI

Have a great week, see you the same time next week!

Full Stack Focus team :)