Build & runtime performance
Build, runtime, and frontend performance optimization: Turbopack, startup/HMR, memory use, bundle reduction, Core Web Vitals, assets, and image optimization.
32%
Best tweets about Next.js
Discover the best tweets about Next.js, covering App Router, Server Components, caching, rendering, performance, deployments, releases, and architecture.
Specific Next.js features, App Router behavior, rendering, caching, performance, deployment, debugging, architecture, and production tradeoffs.
Original Xholic analysis
Across 50 Next.js posts, discussion covers release features, App Router and caching practices, deployment choices, migration tradeoffs, security alerts, and SEO implementation. The dataset contains both supportive release and deployment posts and cautionary or critical reports about dashboard architecture, migration performance, and operational risks.
46% of posts
All-time engagement
34% of posts
Published in 90 days
Conversation map
Build, runtime, and frontend performance optimization: Turbopack, startup/HMR, memory use, bundle reduction, Core Web Vitals, assets, and image optimization.
32%
Deployment portability and production infrastructure across Vercel, Cloudflare, AWS, Kubernetes, Docker, VPS/Dokku, Aspire, and multi-cloud adapters.
24%
Framework comparisons and migrations between Next.js, Vite, TanStack Start, React Router, and related alternatives, especially for SaaS dashboards and developer experience.
24%
App Router and React architecture: Server/Client Component boundaries, Server Actions, RSC interoperability, route params, and rendering behavior.
20%
Caching, rendering strategy, and navigation performance, including use cache, SSR/SSG/ISR choices, hydration, shell prefetching, and cache debugging.
16%
Next.js releases and APIs, including View Transitions, Link transition types, navigation prefetching, Adapter API, root params, agent docs, and developer diagnostics.
12%
Security risks affecting Next.js and its ecosystem, including vulnerabilities, path traversal, credential exposure, authorization mistakes, and supply-chain concerns.
12%
SEO and crawlability for Next.js sites, including robots rules, rendered HTML, JavaScript-disabled output, metadata, sitemaps, and static prerendering.
10%
Tone and stance
Performance benchmark
Posts with media make up 54% of this collection. Their median all-time score is 18.1, compared with 11.0 for text-only posts.
Format mix
Consensus and debate
Shared view
Release posts highlight concrete capabilities including claimed rendering and development-startup improvements, Link transition types, the stable Adapter API, and shell prefetching for navigation.
Shared view
App Router-oriented posts focus on Server/Client Component data flow, cached components reading root params, and Next.js caching considerations for technical SEO workflows.
Shared view
Examples cover AWS Amplify, Kubernetes deployments modeled with Aspire, a self-managed Dokku VPS, and Cloudflare Workers, illustrating both managed and self-managed deployment paths discussed in the dataset.
Open debate
One builder recommends Next.js for every web product, while other posts argue that authenticated SaaS dashboards may be better suited to Vite plus React, with Next.js used for public, SEO-oriented surfaces.
Open debate
Next.js 16.2 and a Turbopack optimization post report performance improvements, whereas migration posts report faster builds, startup, HMR, page loads, or lower memory use after moving away from Next.js.
Open debate
One post frames the Adapter API as enabling deployment across platforms. Separately, a security alert claims a severe issue affected self-hosted versions, and another post describes build-time configuration friction with images.remotePatterns in pre-built Docker deployments.
What performs
The five supplied score outliers consist of a View Transitions announcement, a Turbopack memory-efficiency claim, Railway’s migration case study, a security alert, and the Next.js 16.2 release post.
Migration and comparison posts provide concrete reported figures: Railway says builds fell from 10+ minutes to under two; one migration post says memory fell from 30GB to 2GB; another says build time was halved; and another compares roughly 50-second TanStack Start builds with roughly 120-second Next.js builds for the same application.
Optimization and SEO posts name specific levers: WebP conversion, font preloading or subsetting, import optimization, lazy loading, route-level rendering choices, and avoiding robots rules that block /_next assets.
Statistical standouts
Creator landscape
The five most represented creators account for 20% of the selected posts.
1. Alex Sidorenko
@asidorenko_
2 posts
2. Aurora Scharff
@aurorascharff
2 posts
3. Christian Nwamba
@codebeast
2 posts
4. Critical Thinking - Bug Bounty Podcast
@ctbbpodcast
2 posts
5. David Fowler
@davidfowl
2 posts
6. Harshil Tomar
@Hartdrawss
2 posts
Aurora Scharff’s two cited posts cover the new Link transitionTypes API and an experiment using next/root-params within "use cache" alongside next-intl.
The official Next.js posts cite 16.2 rendering, development diagnostics, and error-reporting changes, alongside a stable Adapter API built with Netlify, Cloudflare, OpenNext, AWS, and Google Cloud.
The cited Critical Thinking - Bug Bounty Podcast posts discuss cross-framework CSPT research and characterize Next.js parameter handling as exposing server-side traversal concerns.
Themes, sentiment, stance, and post format are classified per tweet. All counts, shares, medians, creator concentration, freshness, and performance comparisons are then calculated directly from the published snapshot.
Xholic's all-time score compares engagement while accounting for reach, post age, and creator consistency. It is used for relative comparisons within this collection.
This report analyzes the exact 50-post snapshot shown below. AI identifies editorial categories and drafts explanations; all statistics are calculated from the snapshot, and every narrative claim is checked against cited posts before publication.
Best Next.js tweets
Ranked 01–50
@aurorascharff ·
View Transitions just got simpler with Next.js 16.2. <𝙻𝚒𝚗𝚔> now has a 𝚝𝚛𝚊𝚗𝚜𝚒𝚝𝚒𝚘𝚗𝚃𝚢𝚙𝚎𝚜 prop. Tag your navigation with a type, and <𝚅𝚒𝚎𝚠𝚃𝚛𝚊𝚗𝚜𝚒𝚝𝚒𝚘𝚗> picks the right animation.
@rauchg ·
Fable just found a 15-30% memory efficiency improvement in Turbopack / Next.js, nearly autonomously. @tobi asked me today: what have been your “holy s***” moments with AI? My answer was: it’s every single week. And it’s accelerating. In fact, “WTFs/day” might just be my favorite metric for AI progress. It’s one thing to read benchmarks or stories online. It’s another to watch these machines pull engineering feats every day. 3 days ago? Sol helped us find novel vulnerabilities in some of the most audited code in the world. Today? Fable helps us ship this large optimization of a very complex Rust codebase. I just saw some results of work we’ve done to shrink binaries by 10-20x. List goes on.
@prasenx ·
Next.js just got its worst vulnerability ever, CVSS 8.6. → affects versions 13.4.13+, 14.x, 15.x, and 16.0.0–16.2.4 → attackers can access your internal services, cloud credentials, API keys, and admin panels → no authentication needed → one crafted request is all it takes → roughly 79,000 instances are exploitable right now → vercel-hosted apps are safe, self-hosted are not upgrade to 15.5.16 or 16.2.5 immediately.
@asidorenko_ ·
transitionType prop for Links in Next.js 16.2 (requires additional setup for <ViewTransitions />)
@nextjs ·
Next.js 16.2 introduces a stable Adapter API, built with Netlify, Cloudflare, OpenNext, AWS, and Google Cloud. But the API is only part of the story. Next.js is used by millions of developers across every major cloud, and making it work well everywhere is on us. Here are our commitments. https://t.co/ylGpJG3WpZ
@alex_barashkov ·
okay, now I get the TanStack Start hype. moved our app from the latest Next.js over the weekend, and the dev experience improved dramatically: 30GB RAM down to 2GB, much faster startup, and almost instant page loads even from a cold run. after spending the last few weeks trying to optimize our Next.js setup with AI and getting nowhere, i ended up using fewer tokens on the migration than we did on unsuccessful debugging/optimisation.
@Tristan0x ·
Claude Code is a huge productivity unlock. As a career backend dev (with minimal React experience), I spent just a few hours on Sunday optimizing our frontend performance and our app now loads 75% faster! The beautiful thing is its able to run its own lighthouse tests, as well as perform chrome browser automation (`claude --chrome`) to validate/quantify results, then iteratively improve. Easy wins it picked up: * Convert all pngs to webp * Preload fonts with fallback * Inline icon SVGs for less network requests * Optimized imports and tree shaking using Nextjs optimizePackageImports for heavy deps * Lazy-loaded large SDKs and certain modals * Defer GraphQL client initialization until used * Create slick loading animations for charts and skeleton loaders for data fields Crazy to think how we could speed up the majority of web applications on the internet for the cost of a few dollars per site.
@Hartdrawss ·
i have built 50+ AI MVPs for clients in 2 years this is the EXACT stack i use EVERY SINGLE TIME ! 1/ frontend : > next.js, always, no exceptions for web products > shadcn for UI components, never build from scratch > tailwind for styling, never write custom CSS 2/ backend : > firebase / postgres for database, auth, and storage in one > node.js for anything custom that supabase doesnt handle 3/ AI layer : > claude API for anything that needs actual reasoning > vercel AI SDK for streaming responses in the UI never build your own AI wrapper from scratch 4/ deployment : > vercel for frontend, zero configuration > aws for anything that needs more control Simple, boring and reliable works best for me !
@aleyda ·
A must read for technical SEOs: Making Sense of Next.js Caching for Developers and SEOs by @WillKennard 👇 This piece to break down the core concepts of caching in Next.js from the perspective of a technical SEO: * Two Types of Page * Dynamic Routes, Suspense, and Hydration * What this means for SEOs and content teams: Workflows and considerations * More! Check it out: https://t.co/ojJSUmRCCg
@ctbbpodcast ·
@xssdoctor's CSPT research covers eight frameworks: https://t.co/fM5TAY3xa5 React Router's .replace(/%2F/g, "/") in matchPath has no i flag, so double-decode only works when the F in %252F is uppercase. This was reintroduced after a previous fix and is still in the codebase. Splat routes (path="files/*") match with (.*) instead of ([^\\/]+), so ../../admin works with zero encoding. Next.js uses the same await params API in page components and route handlers but they do opposite things. Page components re-encode through getParamValue(), route handlers fully decode through getRouteMatcher(). The traversal lands server-side. Ember's normalizePath() re-encodes % after decoding, which accidentally kills double-encoding. Wildcard params skip the final decodeURIComponent entirely, so they need literal ../ instead of encoded payloads. SvelteKit's param matchers reject bad values at the routing level before any load function even runs. Server load functions in +page.server.ts bypass hooks.server.ts, so auth middleware won't protect you. Nuxt's island component payload revival (revive-payload.client.js) is a stored CSPT sink. If you can poison window.__NUXT__, the key traverses the $fetch URL. (CVE-2025-59414)
@southpolesteve ·
vinext is one month old today! It's a Vite plugin that reimplements the Next.js API surface. Your existing Next.js app, running on Vite. Deploy to Cloudflare Workers, Vercel, Netlify, AWS, or any Node.js server. It's just a Vite plugin, so `vite start` is all you need to ship to production. One of the biggest questions I got was "how will you maintain this?" The project was always about pushing the limits of what AI can do, and maintaining it has been no different. We use AI to triage, fix issues, review, merge PRs. The numbers over the last month: 395 merged PRs. 52 contributors. 34 releases in 27 days. It's been awesome to see a community form around this. Huge thanks to @James_Elicx, @Divkix, @hyoban_cc, @jokull, @Dayifour, @yunus25jmi, and everyone else who jumped in and shipped real code https://t.co/zGvTui0HGN
@freeCodeCamp ·
Next.js apps are split between Server and Client Components, and it isn't always obvious how they should work together. In this tutorial, @webmekanic_ teaches you simple ways to pass data from Server Components to Client Components. You'll also learn which props can be passed safely, how to share data with Context and React.cache, and how to avoid environment poisoning. https://t.co/CuNrN2DLy1
@tomilola_ng ·
Vite is better than Next.js, for most SaaS dashboards Next.js is great when you need public pages, SEO, content, landing pages, blogs, and marketing routes. But for an authentication-gated SaaS platform? You probably don’t need half of what Next.js brings. Most SaaS dashboards are heavy client-side apps: Charts, Tables, Filters, Modals, Multiple State management, Role-based UI, Realtime updates, Admin workflows, Settings pages ... That is Vite territory. For me, the better setup is simple: -> Next.js for the marketing website. -> Vite + React for the actual SaaS dashboard. I use the right tool for the right surface, the UI below is still just a shell, but I’m building my SaaS one step at a time. #Pabbi is coming
@Hartdrawss ·
Just hit these numbers on a client Next[.]js build today ??!! LCP 0.40s. CLS 0.06. INP 8ms. heres exactly how i got there: [ step 1 — rendering strategy ] - most devs default everything to SSR and move on completely - that's poor strategy tbh; every route needs a decision. does this page change per user? - we mapped every route before writing a single component - this alone cuts your server response time significantly and it's quite a difference [ step 2 — nothing blocks the fold ] - went through every script loading above the fold like tracking scripts, font loaders, analytics - the browser should paint the page before it does anything else - Hard re-work but was able to clean it up; lazy loaded all of them so they only fire after the page fully paints [ step 3 — SEO infrastructure on day one ] - schema markup, OG tags, sitemap, robots.txt done and knocked out before any feature - crawl structure mapped properly so that google can actually index the right pages [ step 4 — assets compression ] - every image converted to webp, sized for the containers and a s3 integration - fonts subset to only the characters being used - videos compressed hard before the first commit TLDR; performance and SEO aren't a sprint you run before handoff but rather the foundation everything else sits on. if your site isn't ranking or loading slow on mobile don't blame the content !
@ctbbpodcast ·
HackerNotes TLDR for episode 168! — https://t.co/HDM9ssvdg8 ►⠀React Router's useParams double URL decodes path parameters, and a case-sensitive regex on the React Source Code of matchPath means %252F (uppercase F) decodes to / while %252f (lowercase f) does not ►⠀Not all frameworks are equal: Vue and React are the most vulnerable to CSPT, Next.js and SvelteKit expose secondary context path traversal on the server side, while SolidStart is largely safe ►⠀Pre-production endpoints may serve uploaded HTML inline instead of as Content-Disposition: attachment, a reliable technique to bypass XSS mitigations on file upload ►⠀fetch() silently strips tab characters (%09), enabling WAF bypasses with payloads like %2F%2e%09%2e%5C
@BradfordHuber4 ·
Some more thoughts on @webflow & why I've started building with Claude Code on a custom stack. I see a lot of people talking about how this kind of stack is only good for developers & technical people. If I believed that was true I'd still be building on @webflow . Ultimately I think AI + a custom stack can solve the original problem that Webflow set out to solve. To democratize the web and allow non technical users to build. A huge focus for me at Amply was building sites that could fulfill this promise in the best way possible - where clients wouldn't need us after the build was complete. That started with using client first - a class framework built to be more friendly to clients and non technical users. This was a huge improvement on what was there before, but we still noticed issues clients would have. Even with solid training in place, we'd still have clients go into Webflow & break classes & layouts on accident. So we started building more with components. Components set more clear guardrails where clients could build pages without needing to understand classes and how they work. We adopted a "Component-First" approach where we built everything in Webflow with components. When Webflow released slots, variables, and component variants, this made this approach even more powerful, and frameworks like Lumos really popularized this as well. As we rolled this out with clients, this was really powerful and we saw an overall great response from clients, but at the same time, I could see things that weren't perfect. The complexity of these projects rose dramatically in order to create super powerful components that could span tons of use case. Lots more custom code The amount of components in larger projects was pretty staggering - could see as many as 200 components in a project. This led to a learning curve in and of itself as clients had to learn the components and all their ins and outs. To me this approach created a split - on the one hand clients had more power than ever to build without needing to understand code on the other hand it increased the learning curve for a client when they needed to edit a component or go beyond what the components were capable of. Most clients still would come back to us to help manage there site at high level for bigger changes. What I'm seeing with AI is that ultimately I can create the ultimate system that fulfills @webflow 's original mission. I can build a site in Next.js + Sanity with components, create a design.md file outlining the design system, components, and how they should be used. A client doesn't need to understand the system in depth. A client can upload a google doc of their page copy and AI can assemble the page with preexisting components and styles. And if a component doesn't exist? The client can work with AI to build net new layouts & pages that still fit with in the core styles and design system. This is the vision I'm building for at @thenativeagency.
@foley_seo ·
Optimizing your site for SEO / LLMS? It's CRUCIAL that user-agents / bots are actually able to see the rendered output. Not all tech stacks are suitable out of the box - JS tech stacks are notorious for: 1. Not being easy to render even for Googlebot which can render JS but isn't overly efficient at it 2. Not being suitable for LLMs to crawl / ingest content Want to know if your site is? 1. Visit a page on your website, hit F12 and open DevTools 2. Press CTRL + SHIFT + P 3. Type Javascript and click Disable Javascript 4. Click NETWORK in the header menu of devtools 5. Ensure Preserve Log and Disable Cache are set 6. Now on your website URL append ?test and hit enter 7. On your network tab, click DOC (shown with red box in image) 8. Where it says NAME you should see ?test, click it 9. Click on RESPONSE tab (shown with red box in image) This will show you the OUTPUT of what an LLM would see. Google CAN render Javascript but, not all Javascript tech stacks are the same - if you want to see the RAW TEXT OUTPUT, do the following: In the bottom half of DevTools you will see menu items "Console, Issues, Coverage" etc. click on CONSOLE (selected by default). Click the CLEAR icon (circle with line through it) Copy and paste this command into console and hit ENTER: const text = document.body.innerText; for (let i = 0; i < text.length; i += 1000) { console.log(text.slice(i, i + 1000)); } This will show you the RAW, NON JS output of your URL. THIS is the content that LLMS will see, and potentially Google/other search engines if there are any issues with SSR. IF YOU ARE USING JS based tech stacks you SHOULD consider SSG (static site generation) or pre-render. Effectively you want to MAXIMISE the output of RAW TEXT/HTML in the dom post JS processing before that document output is served. The amount of SITE OWNERS who have these fancy tech stacks are often oblivious to the risks/issues with using JS for a web app / site framework. Here are some solutions based on common platforms: React Next.js – [SSG + SSR + ISR] Most popular for React, supports hybrid static rendering and server-side rendering Gatsby Fully static, React-based. Uses GraphQL for data and pre-renders everything at build time Astro Partial hydration. You can use React components but ship less JS by default Vue Nuxt Offers static site generation (nuxt generate) as well as server-side rendering VitePress Lightweight Vue-based SSG, ideal for docs and content-driven sites VuePress Vue-powered static site generator, originally designed for documentation Svelte SvelteKit Can be configured for full SSG, SSR, or hybrid deployment. Excellent for fast, SEO-friendly output Elder.js Performance-focused SSG for Svelte apps (less common now but still useful) Angular Scully The go-to SSG for Angular. Pre-renders Angular apps into static HTML Analog.js Newer project aiming to bring Next.js-like DX to Angular with SSR and SSG #seo #seo2025 #aiseo
@ldo_dev ·
My brain is fried but I'm grinning like a fool 😆 Yesterday I had only built web apps and now I've got a monorepo running two front ends with Next.js for web and Expo for mobile. Both apps share the same Prisma ORM setup and they share the same Hono API endpoints. One database, one API, multiple front ends. The best bit about this is that the technology is starting to make sense I can't wait to share what breaks next 🫡
@larsbuilds ·
My current deployment flow is pretty cool IMO 🤓 I do have a single node k3s Kubernetes cluster hosting all my SaaS apps as well as harbor and a single DB pod. Every SaaS app repo is a mono repo, holding the business logic mostly done in NextJS and sometimes a little python. The repo has a helm chart too. And I do have one repository with all my argo deployments in it. It uses the SaaS app repo git sha for the image and chart. Whenever I push to main of the SaaS repo: > image gets built and pushed to my harbor > image sha and helm chart sha is updated in the argo-deployments repo > this triggers a sync on argo and deploys my staging environment > some e2e tests run > if sucesfull a PR for prod is opened on argo-deployments > the PR description is generated by codex > every former PR for this app is closed > when merged prod gets updated
@lmsysorg ·
🌐 https://t.co/rLZWRB4aBd is live with a full rebuild: faster, easier to read, and now with an open talent pool! We migrated from a GitHub Pages static blog to a modern Next.js 15 + React 19 site, with the blog build time dropping from ~20min to ~1min. 🎨 What's new: - Refreshed visual design with responsive layout, fluid typography, and a new project timeline - New blog categories (Tech Blogs / News) & full-text search: filter and find exactly what you're looking for - Live scroll tracking on every post, easy to navigate long technical reads 🛠 Better experience for contributors: - PR preview staging: Live PR previews before merge - Full SEO overhaul for better discoverability: dynamic sitemap, OpenGraph, Twitter Cards 🌟 We're also launching North Star talent pool, a long-term talent initiative for engineers, researchers, and builders who want their work to run in production. We're looking for people across inference systems, training infrastructure, AI agents, developer tools, and Event Operations & Community Ops experts. Drop by if you're looking to contribute or join. 🌐 Check out the new site: https://t.co/Pj59SAN8Y2 📩 Interested in North Star? Drop us a note at [email protected]
@mauricekleine ·
how is a nextjs app on vercel with the whole stack in the same region still so laggy? my tanstack start app with a hono backend on AWS and a postgres database absolutely FLIES there are ZERO skeletons or spinners in the tanstack app... ...on a throttled 4G network no server components no cache components no opaque non-standard APIs no security mistakes that LLMs make all the time (auth only in layout.tsx and not in child page.tsx? it's safe bro!! (it's NOT)) just react and oRPC routers boggles the mind
@MaheshPawaar ·
vercel got hacked. a threat actor claiming to be shinyhunters is selling stolen data for $2 million on breachforums. vercel confirmed unauthorized access to internal systems on april 19, 2026. here's everything you need to know. > how it happened: the entry point wasn't a zero-day exploit. wasn't sophisticated malware. it was context[.]ai – a third-party ai productivity tool used by a vercel employee – compromised via a google oauth app. attackers used that access to pivot into vercel's internal systems and read environment variables from a limited number of customer configs. > what the attackers claim to have: employee accounts, github tokens, npm tokens, internal database, and source code. they've already leaked 580 employee records as proof. asking price on breachforums – $2 million. > why this is bigger than vercel: vercel maintains next.js. if the attacker's claims about github + npm tokens are real, the risk is a potential supply chain attack. one breach at this level doesn't just affect vercel's customers – it could ripple across millions of apps built on top of their infrastructure. > what's confirmed safe (per CEO guillermo rauch): next.js codebase – no malicious code injected. turbopack and all open source projects – confirmed safe. vercel services – fully operational. mandiant is in. law enforcement notified. limited customers are being contacted directly. > the real lesson: we keep concentrating the modern web into fewer and fewer platforms. vercel + github + npm = the backbone of millions of production apps. one compromised productivity tool connected via oauth is all it took to get inside. the fix isn't panic – it's least-privilege access. every tool gets only what it actually needs. nothing more. > if you're on vercel, do this now: rotate your env variables and secrets, audit all third-party oauth apps on your google workspace, enable vercel's sensitive variable encryption feature, and keep an eye on vercel[.]com/kb/bulletin for updates. stay sharp.
@saurra3h ·
if you're using @motiondotdev with the React Compiler in Next.js disable auto-compilation for motion components the compiler memoizes renders, but framer needs to re-run on every render to track element positions for layout and layoutId animations everything works locally but silently breaks in prod fix: reactCompiler: { compilationMode: "annotation" } and opt-in with "use memo" only on non-motion components let me know if any of you found better solution
@_javierivero ·
Concept page for Artemis II, here's the prompt: > Imagery by @nasa official photo catalogue and edited with @basementstudio shader tool "Build an interactive hero page with a spotlight reveal effect using Next.js and Tailwind. The page has two versions of the same full-bleed background image — a "base" image and an "alternate" image. The alternate image is only revealed inside a circular spotlight that follows your cursor, drawn on an HTML canvas using globalCompositeOperation: "source-in" to mask the image within the circle. When the cursor moves fast, ghost echo rings trail behind and fade out, giving the spotlight a sense of motion. A custom CSS cursor ring with crosshair lines replaces the default cursor. All movement uses lerp (linear interpolation) for smooth animation via requestAnimationFrame. The page also has parallax: the background and text layers shift slightly in opposite directions as you move the mouse. On top of the visuals, overlay a large typographic headline and a live telemetry dashboard panel (bottom-right) with metrics that tick and animate in real time using setInterval. Keep everything in one component, drive all animation through mutable refs to avoid React re-renders, and style with a monospace/NASA mission control aesthetic — dark background, muted whites, tight letter-spacing."
@wickedguro ·
Something I don't understand how to do on Next.js Happy to hear from the next experts. My next config has images remotePatterns. It's all good if you just deploy next and have the env variables during the build. But what if I have a pre-built Docker with a Next app? How can I add the .env variables there, since this file is only used at build time? For now, I just changed all my <Image /> to <img /> because I don't see any solution.
Best Tweets by Topic