VarBear

Code is drifting from keystrokes to conversations while the infra pendulum swings from closet servers to community clouds—meanwhile Pinterest buries Hadoop and Google ships an agent that opens PRs. Hype meets friction here: vibe code’s debt, stealthy crawlers, the Linux you actually want, worktrees that save context, and a tiny human’s first tech ladder—dig in.

🧩 Computational Thinking Is The New Programming
🤖 Google releases AI agent Jules for programming
🏗️ Next Gen Data Processing at Massive Scale At Pinterest With Moka
☁️ The Future is NOT Self-Hosted
🧪 Vibe code is legacy code
🌳 Worktrees: Git's best kept secret (and why you should use them)
🐧 Best Linux distro for developers of 2025
🕵️ Perplexity is using stealth, undeclared crawlers to evade website no-crawl directives
🧑‍💻 Myth Or Reality: Will AI Replace Computer Programmers?
🧒 4 Ways I am Encouraging My 4 Year Old Child to Help Learn Coding and Use Computer

Less vibe, more signal—sharper tools in hand, go cut code, not corners.

Have a great week!
FAUN.dev Team

Google’s AI agent Jules just leveled up—out of beta and into full-on dev mode. It now handles asynchronous tasks, pushes real-time code updates, and can spin up pull requests with deeper GitHub integration.

Under the hood: it runs on the beefier Gemini 2.5 Pro model. Adds Environment Snapshots for state capture. Rolls out tiered pricing. And tightens the data policy—no training on private repos, period.

XM Cyber just dropped a guide on putting Continuous Threat Exposure Management (CTEM) into practice with their platform. It maps out clear steps to bake exposure management into your 2025 security plans.

Trend to watch: CTEM is leveling up—no longer just a buzzword, it's becoming a real security discipline with actual tooling. The era of scattered vuln scans? Fading fast.

Perplexity AI’s stealth crawling behavior includes modifying user agents and source ASNs to avoid website blocks, highlighting the importance of transparent bot behavior.

Docker just dropped the MCP Toolkit and MCP Gateway, tightening up the Model Context Protocol with serious armor. We're talking six major server-side holes patched—OAuth RCE, command injection, leaked creds—plugged.

How? With container-wrapped isolation, real-time network filters, first-class OAuth handling, and secrets locked up natively through Docker Desktop.

Big picture: Protocol security just shifted. It's no longer your problem to duct-tape defenses. Infrastructure owns protection now—baked in, containerized, default.

TechRadar rounds up the best Linux distros for devs. Manjaro delivers Arch power without the pain. Debian and Ubuntu LTS hold steady for those who put uptime over edge. Fedora keeps the new stuff flowing.

Solus rolls with a tight curation hand—smooth updates, no chaos. Mocaccino aims at Gentoo lovers who want portage without the config rabbit hole.

Git worktrees have been around since 2015, but few devs use them like they could. They let you work on multiple branches at once—each in its own directory—without the usual stash-switch-stash-repeat dance.

The real power move? Pair them with a bare repo. That gives you a clean, central base where each worktree lives outside the main checkout. It’s tidy. It scales. And it makes parallel work feel natural instead of chaotic.

Pinterest kicked its creaky Hadoop system to the curb and embraced Moka, a shiny Kubernetes + AWS EKS platform, to crank up scalability and security. Graviton ARM EC2 instances, Spark Operator, and Apache YuniKorn unleashed a performance beast and sliced costs. They wrestled with memory monsters and JDK upgrades, snagging a 5% performance upgrade. Moka now commands 70% of Pinterest's Spark workloads. By year's end, Hadoop will be six feet under.

Software's entering its blurred-lines era. The new hybrid model fuses old-school code with natural language prompts and AI-generated logic. Frameworks like DSPy let devs stitch together pipelines where logic flows through code, prompts, and outside data—like it's all one system.

What’s changing: Programming’s not just typing Python anymore. It’s becoming a conversation—half code, half clever prompt. Knowing both matters if you want to build, debug, or even guess what the thing's doing.

GCompris, CodeMonkey, Microbit, and Raspberry Pi kits aren’t just toys. They’re a full tech ladder for tiny humans. Start with GCompris to get little fingers clicking. Add CodeMonkey for block logic basics. Then toss in Microbit or an Elecrow kit, and suddenly code makes LEDs blink and buzzers buzz.

Kids get a real path: screen to circuits. Just don’t ditch the grown-ups yet—some jumps still need a helper on hand.

Generative AI tools like GPT-4o and Claude Sonnet now handle the grunt work—fixing bugs, cranking out code, writing docs—with scary accuracy. Amazon and Anthropic are already hinting at hiring fewer engineers. But the jobs aren’t vanishing; they’re mutating.

"Vibe coding"—Karpathy's label for cranking out AI-assisted code at warp speed—lets devs skip the deep dive. It works for quick hacks and throwaway prototypes. But ship that stuff to prod? Cue the technical debt.

A dev spun up a full self-hosted cloud on a refit workstation—Immich, Calibre-web, Audiobookshelf, and Jellyfin all running smoothly on Proxmox, Docker, and Tailscale. Custom NAS, the works.

Cool demo. But it hits a wall: self-hosting’s fine for hobbyists, not scaling. One person can’t out-infra Google.

The takeaway? We don’t just need better solo setups—we need shared, open infrastructure the internet can actually rely on.

System shift: The real momentum’s in community-owned, privacy-first cloud tooling. Beyond the solo server in your closet.

Collection of Docker images with headless VNC environments

Cursor for design

The glamourous AI coding agent for your favourite terminal

A CLI to stash secrets. No logins.

Apollo Client enables powerful local state integration through local-only fields, which let you seamlessly blend client-side state with server-fetched data in a single GraphQL query. By defining these fields using `@client` directives and configuring field policies or reactive variables alongside your GraphQL schema, Apollo lets you manage UI and application state without needing separate tools like Redux. This architecture simplifies your app by keeping local and remote data fetching unified under GraphQL’s declarative syntax, reducing boilerplate and enhancing maintainability.

"Dev tools are meant to save time, so we spend the most time mastering them. When you write code, complexity is the enemy; in tooling, it's the motivator."
— SenseiOne

This Week’s Human is Andrew Foe, founder building NVIDIA-powered edge datacenters at HyperAI and a 20+ year cloud/AI practitioner. He’s helped hosting partners become cloud providers, led IoDis to NVIDIA Elite Partner status, and ships sustainable GPU infrastructure (A100/H100/L40S). Previously with Dell, HP, Lenovo, and more, he’s worked with teams at Booking.com, Leaseweb, and AMS-IX to turn complex infrastructure into systems that actually run.