Hello friends!
Welcome to another edition of Sloth Bytes. I hope you had a wonderful week.
Porkbun is the domain name registrar you need.
Still using GoDaddy or Namecheap? There’s a better way with Porkbun!
Porkbun is the domain registrar trusted by creators, developers, entrepreneurs, and folks who want low prices without the nonsense.
Why people are choosing Porkbun:
• Most domains sold at cost
• Low, transparent registration and renewal pricing
• Free features like WHOIS privacy and SSL certificates
• Powerful web and email hosting options
• Real human support 24/7, 365 days a year
• Named the #1 domain registrar by Forbes Advisor and USA Today
For launching a business, building a personal brand, starting a side project, or creating your first website, Porkbun makes it easy.
Get $1 off your next domain registration with Porkbun now.
Know Exactly Who's Spending Your AI Budget.
Every AI request leaves a trail. Mesh gives engineering and finance complete visibility into who used which model, how many tokens were consumed and where your AI budget is going.
Stop guessing. Start governing AI spend.
Connect once, switch between GPT, Claude, Gemini and hundreds more whenever you want, while automatically routing requests for 40% lower costs and 99.99% AI response rate.
Next.js 16.3 is out and wants your app to feel instant
Next.js 16.3 is out, and this release is less about adding another API you forget exists and more about fixing the stuff you actually feel while building and using an app.
The big feature here is Instant Navigations. It’s a set of changes meant to make App Router navigation behave more like a good SPA where you click something, see useful UI immediately, then stream in whatever still needs data.
Other important updates:
Partial Prefetching: instead of repeatedly prefetching nearly identical pages, Next.js can prefetch a reusable route shell once and reuse it across matching routes. Think 20 chat links sharing the same skeleton instead of downloading that skeleton 20 times.
Smarter cached UI: cached content can show immediately while fresh data loads behind it, using Next.js’ Cache Components model.
Less memory usage in dev: in 16.3, Turbopack uses up to 90% less memory when running
next dev.Persistent build caching: Turbopack can save and restore work between runs, which should make repeated local and CI builds faster.
Immutable static assets: hashed framework assets can use long-lived immutable caching instead of being re-requested after every deployment.
Agent-focused tooling: version-matched docs, first-party skills, better error output, and more structured ways for coding agents to understand a Next.js project without hallucinating APIs from 2024.
If you’re already on Next.js 16, you should definitely update to this version. It brings a lot of performance improvements that should make development feel smoother. If you’re coming from an older release, here’s the official Next.js 16 upgrade guide.
Meta made a coding agent too, because obviously
Meta launched Muse Code, a terminal coding agent powered by its new Muse Spark 1.2 model.
Muse Code is built for long-running, repository-scale work:
It can plan, write, test, and verify changes across a codebase.
It can run multiple sub-agents in parallel for harder tasks.
It keeps a persistent action log, so a crash does not mean starting the whole job from scratch.
It is currently in beta and uses pay-as-you-go pricing instead of a normal chat subscription.
The standard model costs $1.25/M input tokens and $4.25/M output tokens, which puts it on the cheaper side of the current frontier coding-model pile.
The more interesting part is the harness. Meta trained the model with Muse Code to ensure it works best with that specific agent. Another sign that “best coding model” is becoming less useful as a standalone question. The agent loop, tools, context management, verification, and prompts can matter just as much as raw model intelligence.
An Open Standard for AI Plugins
Agent Plugins 1.0 is a new open, vendor-neutral standard for bundling Agent Skills and MCP servers into one distributable plugin.
The problem is pretty simple: today you can build the exact same useful agent extension and still have to repackage it differently for every client.
Agent Plugins gives everyone one predictable folder structure:
my-plugin/
├── plugin.json
├── skills/
└── mcp.jsonThat tiny plugin.json manifest tells compatible clients what they’re looking at, while the actual reusable instructions live under skills/ and MCP configuration lives in mcp.json.
AWS, Anysphere, GitHub, Microsoft, OpenAI, and Vercel all helped shape the 1.0 spec, and several major coding-agent clients already support it.
The official compatibility page currently lists:
ChatGPT/Codex
Cursor
GitHub Copilot
Kiro
VS Code
If adoption keeps spreading, this could become the default way developers package reusable agent capabilities instead of every company inventing its own plugin format.
Or, you know… it could end up like this:
Cloudflare is giving AI agents wallets
Cloudflare is building programmable wallets for AI agents. Yep, not sure why but I guess AI needs spending money too.
The idea is to let an agent pay for APIs, data, content, or other resources on its own while still operating inside limits you define. Cloudflare is leaning on x402, a payment standard built around HTTP’s criminally underused 402 Payment Required status code.
A machine can request a paid resource, get a 402 response with payment instructions, pay, attach proof, and retry the request. You don’t need a checkout page, an account, or even payment details for this. It simply uses the balance inside the wallet.
Cloudflare already supports this in its Agents SDK: developers can charge for APIs, pages, and files, or even charge per MCP tool call. Wallets complete the other side of that transaction: now the agent can actually be the buyer.
The useful part is guardrails. Cloudflare’s Agents SDK supports agent-side x402 payments with an optional human-in-the-loop confirmation step, so you can control when an agent pays instead of just letting it control your money. Please don’t go bankrupt because you let your AI buy things for you.
The secure way to release an npm package in 2026 - Evil Martians walks through the security steps you should do if you plan on releasing an npm package.
What Codex Actually Sends to the Model - A developer got curious and recorded what actually gets sent to AI models when you use Codex since it’s not “just your prompt.” It’s a great read to see how complicated coding agents can be.
How I Find Problems to Solve as a Staff Engineer - A very useful article to help you identify problems without you having to rely on someone to assign you a task.
Scanning 7.6 Petabytes of AI Training Data for Secrets - Truffle Security scanned public Hugging Face datasets and says it found 221,303 working credentials across 187 million files. Which means AI models have been trained and have access to those credentials… yikes.
PolicyStack - Small, composable privacy and consent primitives for apps. It’s a really interesting and useful tool. Never realized how annoying it was to make those cookie consent banners.
github/gitignore - Giant collection of useful .gitignore templates. Please use one before accidentally committing .env and becoming next week’s security story.
Cloudflare OS - An open-source AI operating system companies can shape around their own tools, context, and rules. Basically a starter kit for building an internal agent environment without inventing every layer yourself.
TanStack Table v9 - TanStack Table got some major improvements! If you’re already using it, you should see if you can update to this new version. If you haven’t used it and need tables for your app, I strongly recommend it.
Question for you this week: What tech stack do you like to use when creating projects?
For me recently it’s been Next.js and Convex. It’s basic and simple, but has all the essentials and such a great developer experience.
Anyways, 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.
