Skip to content
Frontend Feeds
  • Today
  • Archive
  • Sources
  • Categories
    • The Giants6 Sources
    • Multi Author Blogs18 Sources
    • Top Front-end Bloggers25 Sources
    • More Front-end Bloggers77 Sources
    • Browsers, engines, etc.11 Sources
    • Libraries, Frameworks, etc.12 Sources
    • Company/Startup Blogs12 Sources
    • Developer/Designer News4 Sources
    • YouTube Channels13 Sources
    • Podcasts7 Sources

Category

The Giants

6 Sources
  • CSS-Tricks
  • DEV Community
  • Envato Tuts+ Code
  • Hacker News Front PageUnofficial feed
  • SitePoint
  • Smashing Magazine

Sunday, 22 November 2026

  • DEV CommunityThe Giantsread at source

    Common SEO Mistakes Developers Make in JavaScript Websites

    Most JavaScript websites I audit are not broken because the developer did something wrong on purpose. They are broken because the page Google sees and the page the user sees are two different pages. A founder builds a site with an AI tool like Lovable, Bolt, or v0, and gets a working product in a weekend. What they usually get is a Vite React single-page app, and Google sees almost none of it. The site looks finished in the browser and is nearly empty to a crawler. This post covers the ten mistakes I see most often in JavaScript websites, the fix for each one, one trap nobody warns you about, and a five-minute test that shows you exactly what Google sees. How Google Actually Handles JavaScript Google handles JavaScript in three phases: crawl, render, then index. It indexes the rendered HTML, not the raw source you see when you click view source. That middle phase is where the trouble starts. A page can sit in the render queue for a few seconds, but it can take longer than that, in Google's own words. Vercel and MERJ measured Googlebot on a large site in 2024 and found a median wait of 10 seconds from crawl to finished render, and the slowest 1 percent waited about 18 hours. Google's own JavaScript SEO basics guide says server-side or pre-rendering is still a great idea, because it makes the site faster for users and crawlers, and not all bots can run JavaScript. As Martin Splitt from Google put it, a lot of people are still looking at view source, and that is not what Google uses for indexing. Google uses the rendered HTML. Mistake 1: Content That Only Exists After JavaScript Runs The most common mistake is also the most basic: the server sends an almost empty shell, a div with an id, and everything else only exists after JavaScript runs. All seven guides I read for this post cover it, and it is still the one I find most often. The classic version is an H1 set inside a useEffect after an API fetch. The user sees a headline, the crawler sees an empty div, and the content only appears in Google's delayed second pass, late, partially, or never. Other crawlers and AI bots often cannot render JavaScript at all, so for them the page stays empty. The fix is simple to say and harder to ship: use server-side rendering or static generation for every public page you want to rank, and keep client-side rendering for logged-in areas. In Next.js that means Server Components, getStaticProps, or getServerSideProps. In Nuxt it means SSR, in Angular it means Angular Universal, and for an existing SPA it means prerendering with a tool like Prerender.io. Mistake 2: Empty or Duplicate Meta Tags Every route ships the same title, usually React App, with no meta description and no Open Graph tags. Six of the seven guides I read cover this one, and every shared link ends up showing the same preview. Tags set by React Helmet after the page mounts exist only in the rendered DOM. Google usually picks them up in its delayed second pass, but AI crawlers and link previews on Slack, LinkedIn, and iMessage never run JavaScript, so they see nothing. Give every route a unique title under 60 characters and a description around 150 characters, rendered in the initial HTML. In Next.js use the metadata export or generateMetadata, in React use react-helmet-async, in Vue use @unhead/vue, and in Angular use the Title and Meta services. Mistake 3: Links That Are Not Links A div with an onClick that calls router.push is not a link, and Googlebot will never follow it. This is the mistake that silently deletes a whole site's internal linking. Google can only discover links that are a elements with an href attribute. When navigation lives in click handlers, the internal link graph disappears, and pages that are not linked from anywhere are never discovered at all. Use real anchors everywhere: the Link component in Next.js, Link in React Router, router-link in Vue Router, or a plain a href. If it navigates, it should be an anchor, not a div. Mistake 4: Canonical and Hreflang Injected After Load A canonical tag added by JavaScript after hydration is often missed by Google, and the wrong language page can end up ranking in the wrong market. There is a second version of this trap that only Sitebulb covers: when the canonical in the response HTML and the canonical in the rendered HTML are different, search engines may honour the wrong one. Google's own rule is to never use JavaScript to change the canonical to a different value than the one in the original HTML. Render the canonical and hreflang server-side, in the initial HTML. In Next.js, generateMetadata handles alternates.canonical and alternates.languages, and every page should carry a self-referencing absolute canonical. Mistake 5: Lazy Loading the Content Google Needs First Lazy loading is meant for content below the fold, but I keep seeing it on the hero image and the main copy. That is the content Google needs first, and it is the content that gets delayed. A lazy-loaded hero image delays the Largest Contentful Paint, and images without width and height cause layout shift. Text that loads only when the user scrolls may never be indexed at all, because Googlebot does not click and does not scroll. Lazy-load only below-the-fold images and non-critical resources, and always with explicit dimensions. The hero image should load eagerly with fetchpriority set to high. Content inside tabs and accordions should be in the DOM at load and hidden with CSS if you need it hidden. Mistake 6: No Structured Data No JSON-LD means no rich results, and AI tools like ChatGPT understand the page less well. Four of the guides I read cover this, and it is still missing from most JavaScript sites I audit. Add JSON-LD in the HTML for the types that fit the page: Article, Product, SoftwareApplication, FAQPage, and BreadcrumbList are the common ones. Google confirms that JavaScript-generated JSON-LD is acceptable, as long as you test it, so validate the result with the Google Rich Results Test. Structured data will not fix a page Google cannot render, but once the page renders, it is the cheapest way to win more space in the results. Mistake 7: Soft 404s on Dynamic Routes The single-page app server returns index.html with a 200 status for every path, so a page that does not exist looks like a real page to Google. The app shows a Not found message, and Google indexes the junk URL. Hit a dynamic route with an invalid ID and the page renders an Oops, not found message at HTTP 200. Google logs soft 404s, and if the pattern repeats it can drop the whole URL pattern. Martin Splitt described it plainly: instead of responding with 404, the site just responds with 200, always showing a page based on the JavaScript execution. Return a real 404 from the server or framework: return notFound true in Next.js, or res.status(404) in Express. Google's own workarounds for JavaScript-only sites are a redirect to a server 404 URL, or injecting a noindex meta tag with JavaScript on error pages. Mistake 8: A Missing or Stale Sitemap A JavaScript site with no sitemap leaves discovery to chance, and a stale sitemap wastes crawl budget. Pages that are not linked from the homepage may never be found at all. On a large single-page app the sitemap has to list only canonical, indexable URLs. Error URLs, redirected URLs, and noindex URLs in the sitemap are wasted fetches, and they send Google the wrong signal about what matters on the site. Generate the sitemap at build time from the real routes, include lastmod, refresh it on every deploy, and submit it in Google Search Console and Bing Webmaster Tools. Mistake 9: Infinite Scroll With No Pagination Content that only loads when the user scrolls is content Google never sees. Search engines cannot access it, and Google's guidance says to follow its pagination guidelines and avoid infinite scroll. The fix is pagination with crawlable links: a href links to page 2, page 3, and so on. If you want the scroll experience, add a Load More button that also exposes those paginated URLs, so users get the smooth feed and crawlers get the links. Mistake 10: Blocking JavaScript in robots.txt or Leaving noindex Live Two lines in the wrong file can remove a whole site from Google: a staging Disallow pushed to production, or a noindex tag that survives launch. A Disallow for JavaScript or CSS files stops Googlebot rendering the page at all, because it will not render JavaScript from blocked files. A noindex tag from staging does the same thing in reverse: the page never indexes. Gate staging with an environment variable, so production always gets index, follow, and check robots.txt after every deploy. There is also the inverted version of this trap, which only Sitebulb covers. If noindex is in the response HTML but JavaScript removes it, Google skips rendering entirely, so it never sees that the noindex was removed, and the page is never indexed. Fix noindex in the HTML, never in JavaScript. The Mistake Nobody Warns You About: Hydration Mismatch Turns SSR Back Into CSR Here is the mistake none of the guides I read warns you about: your server-side rendering can silently turn back into client-side rendering, and you will never see it happen. When the HTML the server sends and the HTML the client renders disagree, React throws a hydration error and discards the server HTML, re-rendering everything on the client. The page you thought was server-rendered is, in practice, client-rendered again, with all the same SEO risks: delayed indexing, and missing content if the re-render fails. The usual causes are small and easy to miss: a date rendered differently on the server and the client, a random ID or Math.random in the markup, a browser-only API like window or localStorage read during render, and HTML that changes after an auth check. You can spot it three ways: hydration warnings in the browser console, content that flashes or disappears on load, and rendered HTML in Search Console that is missing content present in the raw HTML. Prevent it by rendering browser-only parts after mount, keeping server and client data identical, and using suppressHydrationWarning only where the difference is expected. One more silent delay works the same way. Google's rendering service may cache your JavaScript and CSS for up to 30 days, ignoring your own caching headers, so a fix you shipped can take weeks to show up in Google's render. The fix is content fingerprinting, a filename like main.2bb85551.js that changes with every build. Ranking on Google and Invisible to AI Search at the Same Time Your site can rank on Google and be invisible to AI search at the same time, and this is the part most developers have not heard yet. Vercel and MERJ tested the major AI crawlers in December 2024 and found that none of them rendered JavaScript. GPTBot, ClaudeBot, and PerplexityBot fetched the raw HTML and moved on. A client-side rendered site gets no citations in ChatGPT, Claude, or Perplexity, because there is nothing in the raw HTML to cite. The fix is the same as mistake one: server-rendered HTML. Bing renders JavaScript with limits, and its index feeds Copilot, so a page that cannot be reliably rendered may not be indexed or selected for grounding there either. Quick Comparison The table below compares the ten mistakes side by side: what breaks, and the fix. Mistake What Breaks The Fix CSR-only content Google indexes late, partially, or never SSR or SSG for public pages Empty or duplicate meta tags No per-page signal, same preview everywhere Unique title and description in initial HTML Links that are not links Pages never discovered, link graph gone Real a href anchors everywhere Canonical and hreflang after load Wrong page or language ranked Render canonical and hreflang server-side Lazy loading critical content Hero and copy missed, LCP and CLS suffer Lazy-load below the fold only, with dimensions No structured data No rich results, weaker AI understanding JSON-LD in HTML, validate with Rich Results Test Soft 404s Junk URLs indexed, trust drops Real 404 from the server Missing or stale sitemap Pages never discovered, crawl budget wasted Build-time sitemap, only indexable URLs Infinite scroll, no pagination Content past the fold never indexed Crawlable paginated links Blocked JS or live noindex Rendering blocked or page never indexed Allow JS and CSS, gate noindex by environment Hydration mismatch SSR silently becomes CSR again Identical server and client render The 5-Minute Test: See What Google Actually Sees You do not need a paid tool to see what Google sees. Five minutes is enough. First, disable JavaScript in Chrome DevTools and reload the page. If the page is blank, that is Googlebot's first view of it. Second, curl the page and grep for your headline: if the headline is not in the response, neither is it in Google's first pass. Third, curl the page with the GPTBot user agent to see what AI crawlers get. curl https://example.com | grep -i "your headline" curl -A "GPTBot/1.0" https://example.com | grep -i "your headline" Fourth, compare the raw HTML with the rendered HTML using Screaming Frog in JavaScript rendering mode or the View Rendered Source extension. Fifth, run the URL through Search Console URL Inspection, click Test Live URL, then View tested page: the HTML tab shows what Google indexed, and the Screenshot tab shows the render. FAQs Three questions developers ask me most often about JavaScript and SEO. Can Google index a JavaScript website? Yes, Google can index JavaScript websites, but only after rendering, and rendering is delayed. Google uses an evergreen Chromium renderer, and content can be indexed late or partially. Other crawlers often cannot render JavaScript at all, so server-rendered HTML is the safe route for any page you want found, ranked, or cited. Is client-side rendering bad for SEO? Client-side rendering is not bad for logged-in pages, but it is a bad default for public pages you want to rank. Behind a login, nobody needs to find the page in search, so render it however you like. For public pages, use server-side rendering or static generation, and use prerendering as the bridge for an existing single-page app. Do I need to rebuild my SPA in Next.js to rank? No, you do not need to rebuild to rank, but you do need server-rendered HTML somehow. Prerendering an existing SPA takes an afternoon, with a service like Prerender.io or Rendertron, and it is the practical first step. Server-side rendering or static generation is the better choice for the next rebuild. Dynamic rendering is a workaround, not a long-term fix, and it is not cloaking as long as the served content matches what users see. About the Author Written by Belayet Hossen Sakib. He is a professional SEO expert . He provides SEO consultancy and search engine optimization services with strategic leadership for SEO client projects, from technical audits and keyword research to content that gets cited on Google and AI platforms like ChatGPT, Gemini, and Perplexity.

  • DEV CommunityThe Giantsread at source

    Five open-source tools for Razer, Logitech and RGB gear on Linux

    I run I Did My Own , a site that collects open-source projects where someone looked at an app and built their own version. This week a little cluster of new listings landed on the Just shipped page, all scratching the same itch: setting up mice, keyboards and RGB lighting on Linux, where the vendor apps (Razer Synapse, Logitech Options+, Corsair iCUE) aren't an option. Here are the five, with the language and licence shown on each project page. OpenRazer: the driver layer Open-source Linux drivers for Razer keyboards, mice and other devices. It exposes lighting, DPI and device settings to user-space apps, which is what the next two projects build on. C · GPL-2.0 Polychromatic: a graphical front end for OpenRazer Controls lighting effects, DPI, polling rate and other device settings through OpenRazer, so you don't have to poke at it from scripts. Python · GPL-3.0 RazerGenie: the Qt option Another front end for OpenRazer, this one a native Qt desktop app for Razer lighting and device settings. C++ · GPL-3.0 Piper: buttons, DPI and profiles A GTK app on top of libratbag for configuring gaming-mouse buttons, DPI and profiles on Linux. It's listed as its own take on both Logitech Options+ and Razer Synapse. Python · GPL-2.0 OpenRGB: lighting across brands A cross-platform app to control RGB lighting on hundreds of devices (many Razer ones included) without vendor cloud software. C++ · GPL-2.0 How they fit together OpenRazer talks to Razer hardware; Polychromatic and RazerGenie are two different desktop front ends for it. Piper comes at it from a different library, libratbag, and focuses on mouse buttons, DPI and profiles. OpenRGB is the one to look at if your desk mixes brands and you mostly care about lighting. All five sit side by side on one product page, and each card links through to the maker's GitHub repo: 👉 Open-source Razer Synapse alternatives on idmo The rest of this week's new listings are on Just shipped . Do you run any of these? I'd like to hear what your Linux peripheral setup looks like, especially if you've hit a device none of them handle yet. And if you've built your own version of a vendor app, you can add it to the site.

  • DEV CommunityThe Giantsread at source

    The Executive Communication Secret That's Spreading Through Silicon Valley

    You dread Fridays. Not because the week was ending, but because you have to write your weekly executive update. Every Friday afternoon, same routine: dig through Slack messages, try to remember what shipped versus what we thought would ship, and craft diplomatic ways to say, "most tasks are delayed." Four hours later, you would have something that was too long, not specific enough, and somehow may still missed the things executives truly cared about. To help you, I built a system to write these updates. Now you can spend 15 minutes reviewing versus 4 hours creating. The Problem Every PM Knows You drive grooming sessions all week. You track tickets in Jira. You have the information executives need scattered across five different tools. But when Friday comes, you are starting from scratch. Looking through chat logs, trying to piece together what happened, explaining why dates moved. The typical PM workflow: Check JIRA for what is open/pending/closed Scroll through Slack for context on blockers Remember (or forget) the important conversation from Tuesday Write everything up in a way that won't get you fired Send it off hoping no one asks follow-up questions This is exactly the kind of work AI handles better than humans. Pattern recognition across multiple data sources, consistent formatting, clear communication of status updates. What Executives Actually Want Having driven leadership meetings for years, I can tell you executives want the same seven things every week: What shipped this week What's blocked and why Key metrics and trends What they need to decide Risks that could affect bigger goals Next week's priorities How this connects to quarterly objectives That's it. Everything else is noise that just makes our job harder. The Simple System Instead of writing updates, collect data all week and let AI assemble it on Friday. Daily data collection (already happening): Team posts updates in specific Slack channels JIRA tickets get updated as work progresses Ensure you keep notes from important meetings Friday processing (15 minutes): Export the week's data Run it through three AI prompts Review and edit the output Send the update AI handles the tedious part (organizing scattered information) while you handle the strategic part (context and priorities). Setting Up Your Data Flow Since you are already creating this information, you just need to organize it slightly differently. Slack channels: Create #product-shipped and #product-blocked channels. Train your team to post there instead of only in DMs. Takes two weeks to become habit. Jira labels: Add exec-update labels to tickets that matter for leadership. Filter by this label when exporting weekly data. Meeting notes: Use this simple template for any important meeting: Decision: [what was decided] Owner: [who's responsible] Deadline: [when it's due] Risk: [what could go wrong] This setup takes 30 minutes and makes your weekly exports clean and useful. The Three AI Prompts Prompt 1 - Organizing the chaos: “I'm giving you a week of product team updates from Slack, Jira, and meeting notes. Sort everything into these categories: SHIPPED: Things that went live to users BLOCKED: Problems that need attention METRICS: Numbers that show progress DECISIONS: Things leadership needs to approve NEXT: Priorities for next week” For each item, write one sentence explaining why it matters. Prompt 2 - Write the update: “Create a 7-section executive update using the organized information: Week Summary (one sentence + three key numbers) What Shipped (max 3 items) What's Blocked (problems + solutions) Decisions Needed (specific asks with deadlines) Key Metrics (actual numbers) Next Week Priorities (max 3 items) Strategic Impact (connection to quarterly goals)” Keep each section to 3 bullets. Use specific dates and numbers. Prompt 3 - Make it executive-ready: “Rewrite this for senior leadership. Remove uncertainty language. Make problems and solutions clear. Use numbers instead of vague terms like "significant" or "substantial." The 15-Minute Review AI gives you 90% of what you need. Now, you can apply the human judgment: Accuracy check: Are the numbers right? Any missing context? Strategic connections: How does this week connect to bigger objectives? Risk assessment: What should executives be worried about? Now, hit send. Real before and after impact you will notice Before (your old style): "The homepage experience encountered some technical challenges this week. The engineering team is working to resolve integration complexities with our payment provider. We expect more clarity on timeline next week." After (AI + your review): "Homepage flow delayed 3 days due to Stripe API rate limits affecting 2% of transactions. Engineering implementing retry logic. New ship date: March 15. Risk: Could impact Q1 revenue by $50K if unresolved." The second version tells executives exactly what's happening and what they need to know for their own planning. How you can get started on Monday · Today: Create the Slack channels and start using them · Monday: Add the JIRA labels to relevant tickets · Wednesday: Test the prompts with this week's data · Friday: Use the system for your actual update Most PMs see immediate improvement in their first AI-generated update. The consistency alone makes you look more organized. This is critical because better executive communication isn't just about saving time on Friday afternoons. When your updates are consistently clear and comprehensive, executives start including you in strategic conversations. The PM who communicates clearly is the PM who gets promoted. While other PMs spend hours wrestling with updates, you are using AI to handle the routine work and focusing on the strategic thinking that moves your career forward. Next week, I will show you how to build an AI agent that tracks everything your competitors are shipping, and struggling with. While everyone else checks competitor websites manually once a month, you will have real-time intelligence.

  • DEV CommunityThe Giantsread at source

    The AI Shame Epidemic

    Your director just complimented your “thoughtful” PRD in yesterday’s leadership meeting. What she doesn’t know: Claude wrote 80% of it in 10 minutes. You feel like a fraud. But you’re not going to tell anyone. You are not alone. Studies show that over 50% of executives are hiding their AI use at work. The interesting observation here is the people using AI the most are the ones most ashamed to admit it. This secret is costing PMs their careers. The Confession No One’s Making A PM at a Series B startup admitted to me: “I used Claude to analyze 47 user interviews in 20 minutes. Found three patterns our team missed in weeks of manual analysis. Presented it as if I did the entire thing myself. Took full credit.” Another PM at a Fortune 500: “I let AI draft all my user stories now. My velocity looks insane. My manager thinks I’m suddenly a rockstar. I’m terrified someone will find out.” The data reinforces this. According to WalkMe’s 2025 survey of over 1,000 workers: 48.8% of employees hide their AI use to avoid judgment 53.4% of C-suite executives hide it even more 62.6% of Gen-Z have used AI for work but pretended it was all their effort The people at the top, with the most power and influence, are the most afraid to admit what they’re doing. Think about it. Why This Is Happening A senior PM told me about an executive meeting where someone asked how their team produced the quarterly roadmap so fast. The PMs looked at each other. Neither said “we used AI to analyze market data and generate scenarios.” Instead, they talked about “improved processes” and “better collaboration.” Here is why they lied: Fear of looking incompetent ; If AI did the work, does that mean I’m not actually good at my job? If anyone can use ChatGPT to write user stories, what’s my value? Unclear company policies ; Most companies have zero guidance on AI use. You are operating in a gray area. Better to stay quiet than risk being the test case. Comparison to others ; Everyone else seems to be grinding it out manually. If you are using AI, are you cheating? Taking shortcuts? Job security fears ; In a market where PMs are getting laid off, the last thing you want is leadership thinking “wait, AI can do this role?” The shame runs deep. And I believe it is (silently) killing careers. The Real Cost of Staying Silent The fact is this - while you are hiding your AI use, here is what is happening: You are ultimately not getting better at it. This is because by default, you can’t improve what you can’t discuss. You can’t share prompts with your team. Can’t learn from others’ techniques. You are stuck at amateur level while pretending to be an expert at manual work. You are missing strategic opportunities. When you frame AI-generated insights as “I spent 40 hours analyzing this,” leadership allocates resources accordingly. They think you need more time and people. The opportunity to restructure how your team operates never comes up. Your competition is pulling ahead. Some PMs have companies with clear AI policies. They are experimenting openly. Building AI workflows. Getting exponentially better each week. You are still pretending it’s 2023. Make no mistake, you are reinforcing the problem. Every time you hide AI use, you signal to junior PMs that it’s shameful. You are creating the next generation of secret AI users instead of skilled practitioners. Teams that talk openly about AI use are moving 3x faster than teams that don’t. The gap widens every sprint. What Top PMs Are Actually Using AI For Since most folks are hiding it, let me just tell you what’s happening behind closed doors. User research synthesis : Remember my article two weeks ago about analyzing interviews? Every senior PM I know is doing this now. The ones who admit it get better at it. The ones hiding it stay mediocre. Competitive intelligence : AI tracks competitor releases, analyzes their job postings, monitors their customer reviews. Some PMs spend 20 hours a month doing this manually. Others spend 2 hours with AI and get better insights. Executive updates : You read about this last week. The PMs crushing it are not spending 4 hours writing updates. They are using AI and spending those 4 hours on actual strategy. PRD first drafts : It is true that no PM worth their weight, is shipping AI-written PRDs without tweaking them. But everyone is using AI to generate structure, catch edge cases, and draft acceptance criteria. Then adding human judgment and context. Roadmap scenario planning : “If we delay feature X by 2 weeks, what’s the downstream impact?” AI can model this in minutes. You can spend days doing it manually, or 15 minutes with AI and the rest of your time deciding which scenario to choose. Data analysis: Pulling insights from analytics, identifying patterns in user behavior, connecting metrics across tools. AI does in seconds what takes hours with spreadsheets. Can you see the pattern I am trying to piece together? AI handles the pattern recognition and initial processing. Humans add judgment, context, and strategic thinking. Now, here is how to break free without career risk Don’t get me wrong, you don’t need to walk into Monday’s standup announcing “I’ve been using ChatGPT for everything.” This will drown your perception fast. Your messaging and approach should be as follows: Start with low-stakes sharing ; Next time someone asks how you analyzed something quickly, say “I used Claude to process the initial data, then I reviewed it for accuracy and added context.” Simple. Honest. No drama. Position it as a skill, not a shortcut; Don’t say “AI did it for me.” Say “I’m developing my ability to work with AI tools to deliver faster insights.” Same outcome, different framing. Share what you verified; When presenting AI-generated insights, emphasize what you checked, what you changed, what you added. This demonstrates your judgment, not just AI’s output. Credit AI like you would credit a junior analyst; If a junior PM helped you analyze data, you would say “Sarah pulled these insights, and here is what they mean.” Do the same with AI. “I used Claude to process these interviews, and here are the patterns I validated.” Lead with outcomes, not methods; Focus on “we identified three critical user pain points” not “I used AI to analyze transcripts.” If someone asks how, be honest. But lead with value. The 48-Hour AI Transparency Test If you want to know if your company culture can handle AI transparency? Try this: Monday morning: In your next team meeting, mention one AI tool you used. Keep it casual. “I tried using Claude to draft these acceptance criteria. Saved about an hour. Happy to share the prompt if anyone is interested.” Watch what happens. Option 1: People lean in. Ask questions. Want to learn. Your company is ready for AI-augmented work. Option 2: Awkward silence. Leadership frowns. Someone makes a comment about “taking shortcuts.” Option 3: Mixed reactions. Some interested, some resistant. Normal organizational change. Based on the reaction, you will know how open you can be. What This Looks Like in Practice A PM at a major tech company shared their experience: They presented a competitive analysis to executives. Usually takes 3 weeks of manual research, analyst reports, website monitoring. This time, they used AI to monitor competitors, analyze their releases, track their hiring patterns, and synthesize customer sentiment. Took 4 hours. In the presentation, they said: “I used AI to process competitor data across multiple sources in real-time. Here’s what I validated and what it means for our strategy.” The VP of Product asked: “Can you share that workflow? Sounds useful for the whole team.” No one cared that they used AI. They cared that they delivered better insights faster. The Window Is Closing Right now, you’re in an interesting moment. Most PMs are (still and should not be)using AI secretly. A small group is using it openly and getting exponentially better. By mid-2026, companies will split into two camps: Organizations with clear AI policies where skilled AI use is a requirement Organizations where everyone’s still hiding it and collectively falling behind The PMs who get comfortable using AI openly now will (most likely and in addition to other competencies) dominate the promotion cycle next year. The ones still hiding it will be competing with people who’ve had 18 months of open practice and improvement. Your choice: keep pretending you’re grinding everything out manually or start building actual AI skills that make you irreplaceable. Listen, you are already using AI. Everyone knows this. Leadership knows. Your peers know. You are not fooling anyone. The shame isn’t protecting you. It’s limiting you. The PMs leading the charge right now aren’t the ones with the best AI prompts. They are the ones comfortable saying “I used AI for X, applied my judgment on Y, and here is why the outcome matters.” Stop hiding. Start improving. The career you save might be your own.

  • DEV CommunityThe Giantsread at source

    My review subagents said "waiting for the notification" and handed back. 11 of 21 did it, and the parent sat idle 12 minutes

    Short version: when a subagent puts its wait in the background and ends its turn with "waiting for the notification," the parent sees that as a finished task. On November 21 I dispatched 23 UI cards in a data platform repo, one after another. For 21 of them I can line up the subagent's transcript with the receipt timestamp. In 11 of those, the review subagent handed back before Codex had written its receipt. The parent got a message that said "still waiting," and then it stopped too. The longest stall was 12 minutes, until I typed "how's it going?" The fix was one commit: the wait now runs in the foreground in 170-second slices and returns a special exit code when it isn't done yet, so the caller just runs it again. Some background. Claude Code splits work into cards and reviews it; Codex writes the code. For each card, the parent session starts a background review subagent. The subagent opens a Codex terminal, waits for the receipt, reruns tests and browser checks against a checklist, and writes a verdict. The parent only reads the verdict and decides whether to merge. What happened It was a fast morning: 23 cards dispatched between 10:52 and 14:13 UTC. The trouble was in the "wait for the receipt" step. The review brief at the time said: "Run wait in the background with Bash, wait for the completion notification, don't poll." The subagents did exactly that. They put the wait in the background, printed something like "Dispatch done, waiting for Codex's round 1 receipt, will continue reviewing when the background notification arrives," and ended their turn. From the subagent's side that's fine. When the background task finishes, it gets woken up and carries on. But from the parent's side, a subagent that ends its turn has handed back. The parent gets a "task completed" notification, and what's inside is a half-finished report. Two concrete cases: UX-3a : the subagent handed back at 11:36:53 saying it was waiting for the receipt. A second later the parent replied "round 1 is still running, no result yet" and stopped. The receipt landed at 11:41:35. The parent only moved again because I sent it an unrelated instruction at 11:40. UX-10a : the subagent handed back at 13:59:05, and the notification summary said it was "waiting on its own background task." This time the parent didn't even reply. It just stopped. The receipt was written at 14:08:54, and the parent didn't move. At 14:11:13 I asked "how's it going?" Only then did it check the evidence folder and find the receipt had been there for a while. From the early hand-back to my question: 12 minutes. Early the next morning, in a different session, I wrote: "I keep seeing the parent agent sitting idle. Either it never dispatched a subagent, or the subagent already finished." Why it happened Reading through the transcripts, there were three layers. Layer one: the brief told the subagent to go to the background. When I wrote it I was thinking "no polling, save tokens." I didn't think about what a subagent ending its turn means to the parent. Layer two: the foreground can't wait forever either. Why not have the subagent wait two hours in the foreground? Because Claude Code moves any foreground command that runs past 180 seconds into the background, whatever timeout you set. So a two-hour foreground wait was never an option, and the subagent would end up at "background wait, end turn" anyway. Layer three: notifications can arrive early. A different child exposed this one. On November 21 a child agent that doesn't report its status to my terminal orchestration tool was announced as "turn finished" right after it launched. The parent trusted the note, went to read the result, saw the child was still busy, and stopped, with nothing left to wake it. I had fixed a bug in the same family on November 19 (a child messaging the parent mid-task made the wait return early). This was the same problem coming in through a different door. Stack the three together and the result is always the same: the parent thinks someone will wake it, and nobody does. The fix: one commit The November 22 fix (6 files, +96 / −23) did four things: The wait now runs in the foreground in slices. Each call waits at most 170 seconds, so it returns before the 180-second cutoff. If the round isn't done, it exits with code 4, meaning "Codex is still working, run the same command again." No receipt file is written until the round ends, so a half result can't be mistaken for a full one. Other messages no longer cut the wait short. When the parent's terminal receives a message from another child, the underlying wait returns early. Now it just keeps waiting until the slice is used up. "Turn finished" needs evidence. The dispatch script records the child's turn counter before sending the task. It only notifies the parent once the counter has moved and the status is no longer "working." For a child that never reports status, it tells the parent after 10 minutes, honestly: "can't tell, don't sit waiting for it, go look yourself." One line in the global rules. Before the parent ends its turn "waiting for a result," it has to confirm something will wake it. If a subagent hands back "still waiting" instead of a result, the parent checks the evidence folder right away: if it's done, carry on; if not, send the subagent back to wait in the foreground until the end. Where the rules live The dispatch script's wait subcommand : the default slice is 170 seconds, a named constant with a comment about the 180-second limit. While Codex is still working it exits 4 and writes no receipt. The review brief : step 2 of phase two now says "run wait in the foreground; on exit code 4, run it again unchanged; don't background it, don't replace it with a sleep loop, don't end your turn 'waiting for the notification'," along with the reason: once a subagent hands back, the parent only has half a report. Tests : three new ones. One checks that a message from someone else doesn't end the wait, one checks that a still-working round exits 4 and writes no receipt, and one asserts the default slice is under 180 seconds. All 9 tests in that file pass, and CI on the fix commit is green. Still open The fix is less than a day old. I don't have post-fix dispatch data yet, so I can't say the stalls are gone. After the next batch of cards I'll count early hand-backs the same way. The brief is text, not a hook. Slicing the wait removes the reason to background it, but if a subagent does it anyway, nothing stops it. The parent's "check before you stop" rule is the backstop. What you can take from this Know what "ending a turn" means one level up. A subagent's "background wait, wait for notification" looks to the parent like a finished task. Anything you need to wait for, wait for before handing back, not after. If your tool backgrounds long commands, slice the wait. Here's a generic version: wait for a receipt file, return 4 when the slice runs out, and the caller runs it again on 4. I tried both cases locally: with no file it returned 4 after 6 seconds; with the file showing up at 3 seconds it returned 0 right away. #!/bin/sh # usage: slice_wait.sh [seconds]; 0 when the receipt appears, 4 when the slice runs out (run again) receipt = $1 ; slice = ${2:-170} end = $( ( $( date +%s ) + slice ) ) while [ "$(date +%s)" -lt "$end" ] ; do [ -s "$receipt" ] && { echo "done: $receipt" ; exit 0 ; } sleep 5 done echo "still working: run the same command again" ; exit 4 Count early hand-backs. For each card, compare two times: when the subagent first handed back, and the receipt file's modification time. If the first is earlier, it handed back early. For me it was 11 of 21 that day. At that rate, a human watching won't keep up. Base "done" notifications on state changes, not on time. Record the turn count before sending the task, and only call the turn over once it has moved. If a child can't report status, say so plainly. This is part 10 of the series. Earlier parts were about putting rules where the agent can't route around them. This one is the other side: the wait has to go somewhere it can't be skipped too. A wait that happens after the hand-back is a wait nobody is doing. About the numbers 23 cards, dispatched 10:52–14:13: the 23 UI card folders in that repo's evidence directory for 2026-11-21; times are the first timestamp in each review subagent's transcript (UTC). 21 matched, 11 early: I read each review subagent's transcript from that day, took the card name from its wait command, and compared it against the modification time of the last receipt file in the evidence folder. If the parent got its first hand-back or completion notice from that subagent before the receipt time, it counts as early. 2 more cards are excluded because the card name in the command was quoted and didn't parse. UX-3a's 11:36:53, 11:36:54, 11:41:35, and 11:40, and UX-10a's 13:59:05, 14:08:54, and 14:11:13: timestamps from the parent session transcript and receipt file modification times (UTC). "12 minutes" is 13:59:05 to 14:11:13. The brief line "run wait in the background with Bash, wait for the completion notification, don't poll": the line removed from the brief in the fix commit (arguments omitted). 180 and 170 seconds: the wait slice constant and its comment in the fix commit. The child announced as finished at launch (11-21) and mid-task messages ending the wait early (fixed 11-19): from a code comment in the fix commit and the message of the 11-19 commit, respectively. 6 files, +96 / −23, 3 new tests, 10 minutes: from the diff of the 2026-11-22 fix commit. 9 tests passing and green CI: a local run of that test file on 2026-11-22 and the CI result for that commit. My own words: from my message in the fix session on 2026-11-22.

  • Hacker News Front PageThe Giantsread at source

    Show HN: I built a game where you are an LLM's subagent

  • Hacker News Front PageThe Giantsread at source

    Muru – AI-assisted Rust server for Python apps by mod_WSGI's creator

  • Hacker News Front PageThe Giantsread at source

    87% of the Anthropic power will come from gas and 11% from coal in 2028

  • Hacker News Front PageThe Giantsread at source

    Notes on the discrete Fourier transform (DFT)

  • Hacker News Front PageThe Giantsread at source

    Show HN: The Riemann Proof Handbook

Friday, 20 November 2026

  • CSS-TricksThe Giantsread at source

    Andy Clarke answers six questions about Standout Designs for the Web

    Web designer, author, and Stuff & Nonsense founder Andy Clarke has released Standout Designs for the Web , a new e-book that encourages designers to move beyond familiar templates and conventional layouts to create more personality-driven websites. The book draws … Andy Clarke answers six questions about Standout Designs for the Web originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • SitePointThe Giantsread at source

    How to Add AI Facial Expression Editing to Your Video Pipeline (2026)

    null Continue reading How to Add AI Facial Expression Editing to Your Video Pipeline (2026) on SitePoint .

  • SitePointThe Giantsread at source

    What to get right before you ship webhooks

    Before you ship webhooks: an event log, signed payloads, retries with backoff, deduplication and delivery logs, with working Node.js code and Standard Webhooks. Continue reading What to get right before you ship webhooks on SitePoint .

Thursday, 19 November 2026

  • CSS-TricksThe Giantsread at source

    scroll-state()

    The CSS scroll-state() query, surprisingly, has nothing to do with whether you’re doom-scrolling in a state of existential dread. It does, however, allow you to change the CSS used based on scrolling behavior. scroll-state() originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • SitePointThe Giantsread at source

    What Is BIO Tagging and How to Use It for Document AI Training Data

    null Continue reading What Is BIO Tagging and How to Use It for Document AI Training Data on SitePoint .

  • SitePointThe Giantsread at source

    What to Check Before Adding a Tripo Model to a Website

    null Continue reading What to Check Before Adding a Tripo Model to a Website on SitePoint .

  • SitePointThe Giantsread at source

    Your AI Pipeline Works. But Can You Trust It?

    null Continue reading Your AI Pipeline Works. But Can You Trust It? on SitePoint .

Wednesday, 18 November 2026

  • CSS-TricksThe Giantsread at source

    When Everything Has Changed, What Do We Do Next?

    It’s been a… year, huh? I think the most common emoji I’ve seen this year is “Face with Bags Under Eyes” (🫩) and that says a lot. One thing is for sure: the industry we make a living in is … When Everything Has Changed, What Do We Do Next? originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • SitePointThe Giantsread at source

    From Script to Screen: A Developer's Workflow for Long-Form AI Video

    null Continue reading From Script to Screen: A Developer's Workflow for Long-Form AI Video on SitePoint .

  • SitePointThe Giantsread at source

    Sizing GPU Memory for a RAG Service: KV Cache, Context and Concurrency

    null Continue reading Sizing GPU Memory for a RAG Service: KV Cache, Context and Concurrency on SitePoint .

  • SitePointThe Giantsread at source

    5 AI Tools for Creating Audio-Led Music Videos: A Practical Workflow

    null Continue reading 5 AI Tools for Creating Audio-Led Music Videos: A Practical Workflow on SitePoint .

  • SitePointThe Giantsread at source

    Retail Merchandising Principles for Web Designers and Developers: Display Design, Ranking Rules and Measurement

    null Continue reading Retail Merchandising Principles for Web Designers and Developers: Display Design, Ranking Rules and Measurement on SitePoint .

  • SitePointThe Giantsread at source

    5 Things Most Developers Get Wrong When Turning Images into HTML

    null Continue reading 5 Things Most Developers Get Wrong When Turning Images into HTML on SitePoint .

Tuesday, 17 November 2026

  • SitePointThe Giantsread at source

    Serving a Local LLM to Several Users: Ollama, LM Studio, llama.cpp and vLLM Benchmarked

    null Continue reading Serving a Local LLM to Several Users: Ollama, LM Studio, llama.cpp and vLLM Benchmarked on SitePoint .

  • Smashing MagazineThe Giants10 min read

    Meet AI 2027 Conference (Dec 1)

    Roll up your sleeves! Meet shiny new [AI 2027 Conference](https://smashingconf.com/ai-2027-conference), a friendly online conference on AI-ready design systems, AI + Figma and AI design patterns. What to expect in 2027 and where teams are getting AI right. [Jump to tickets.](https://smashingconf.com/ai-2027-conference/registration)

  • Smashing MagazineThe Giants8 min read

    AI Sped Up The Build: How Can Accessibility Testing Keep Up?

    AI has sped up how teams design and build, but quality assurance (QA) testing hasn’t kept pace, and the web may be getting less accessible. Drawing on new research, Andrew Chung explains what separates the teams whose accessibility efforts are keeping up, and how to embed human judgment into AI-enabled workflows without slowing delivery.

  • SitePointThe Giantsread at source

    How Consumer AI Companions Handle Context Windows, Persistent Memory, and Voice Latency

    See how consumer AI companions manage context windows, persistent memory and streaming voice, with practical architecture lessons for stateful LLM apps. Continue reading How Consumer AI Companions Handle Context Windows, Persistent Memory, and Voice Latency on SitePoint .

Monday, 16 November 2026

  • SitePointThe Giantsread at source

    Is Kling 4.0 the Right AI Video Generator for Your Project?

    null Continue reading Is Kling 4.0 the Right AI Video Generator for Your Project? on SitePoint .

  • SitePointThe Giantsread at source

    Hostinger VPS in 2026: Hands-On With the AI-Managed VPS

    Hands-on with Hostinger’s KVM 1 VPS: deploy a Node.js app in Docker and let Hostinger Agent handle HTTPS, the firewall and a reboot. Plans, specs and pricing. Continue reading Hostinger VPS in 2026: Hands-On With the AI-Managed VPS on SitePoint .

  • CSS-TricksThe Giantsread at source

    Staggering Scroll-Driven Animations with animation-range-start

    Imagine a CSS karaoke machine that displays a ball bouncing on text that incrementally highlights, based on progress through via speech recognition and running that kind of animation as a single predictable, holistic effect driven by one progress value. Staggering Scroll-Driven Animations with animation-range-start originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • SitePointThe Giantsread at source

    How to Test WCAG 2.2 in CI: What axe-core Misses

    null Continue reading How to Test WCAG 2.2 in CI: What axe-core Misses on SitePoint .

  • SitePointThe Giantsread at source

    Building Email Integrations: APIs, Webhooks, Automation, and Deliverability

    null Continue reading Building Email Integrations: APIs, Webhooks, Automation, and Deliverability on SitePoint .

  • SitePointThe Giantsread at source

    Build Live Video Chat With WebRTC: From First Call to Production Debugging

    null Continue reading Build Live Video Chat With WebRTC: From First Call to Production Debugging on SitePoint .

  • SitePointThe Giantsread at source

    What Is the Best UI design Tool? 8 Tools Compared by the Job They Do (2026)

    null Continue reading What Is the Best UI design Tool? 8 Tools Compared by the Job They Do (2026) on SitePoint .

  • SitePointThe Giantsread at source

    Markdown to PDF in 2026: No-Install Methods and the Rendering Problems Nobody Warns You About

    null Continue reading Markdown to PDF in 2026: No-Install Methods and the Rendering Problems Nobody Warns You About on SitePoint .

Sunday, 15 November 2026

  • SitePointThe Giantsread at source

    How AI Image-to-Image and Image-to-Video Tools Work, and How to Build a Reliable Workflow

    null Continue reading How AI Image-to-Image and Image-to-Video Tools Work, and How to Build a Reliable Workflow on SitePoint .

Friday, 13 November 2026

  • CSS-TricksThe Giantsread at source

    10 Big Ol’ 2027 Black Friday Deals

    A big ol' roundup of Black Friday deals on front-end courses from the best independent publishers and educators in the business. 10 Big Ol’ 2027 Black Friday Deals originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Thursday, 12 November 2026

  • CSS-TricksThe Giantsread at source

    Beyond 12 Columns: Designing Layouts With Personality

    Too many websites look alike. They use familiar layouts, repeat the same patterns, and often leave little impression once you’ve moved on. Andy Clarke explores how understanding a company’s story can lead to more distinctive designs, using modular and compound grids to create layouts with character. Beyond 12 Columns: Designing Layouts With Personality originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • Smashing MagazineThe Giants15 min read

    Should You Actually Add AI To That? A Four-Question Audit For Product Teams

    An AI feature can save users time on a task, only for them to spend it checking the results. This practical audit helps product teams decide where AI earns its place and what users need to verify its output.

Tuesday, 10 November 2026

  • CSS-TricksThe Giantsread at source

    CSS Optimization: Is Using Declarations Just Once (UDJO) Still Effective Today?

    Is the technique inspired by the DRY principle by identifying and reducing (if not eliminating) CSS declaration repetition still a relevant optimization strategy? CSS Optimization: Is Using Declarations Just Once (UDJO) Still Effective Today? originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Monday, 9 November 2026

  • CSS-TricksThe Giantsread at source

    CSS-Tricks, Version 20

    Things look a little different around here? I hope so because we just pushed a site redesign! Here's everything new with the 20th(!!) version of CSS-Tricks. CSS-Tricks, Version 20 originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Friday, 6 November 2026

  • CSS-TricksThe Giantsread at source

    RSS Club: CSS-Tricks Redesign Preview

    Hey you wonderful RSS subscribers! We just soft-pushed a site redesign and wanted to give ya'll the first heads-up. A full write-up is coming but here are the high-level details. RSS Club: CSS-Tricks Redesign Preview originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • Smashing MagazineThe Giants4 min read

    A New Community For In-House UX People

    We’ve been working on something new! We’ve teamed up for a **new small community** for people doing UX-inside organizations — with weekly calls and a shared library of the business cases, templates, and research plans that have actually worked for other folks. [Jump to details](https://uxinsiders.boagworld.com/).

Thursday, 5 November 2026

  • CSS-TricksThe Giantsread at source

    Taming Typography: Building a Modern CSS Reset

    Stu created tidy-typography.css: a style reset for taming typography with modern CSS features for everything from fluid typography and vertical rhythm, to hyphens and variable fonts. Taming Typography: Building a Modern CSS Reset originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • Smashing MagazineThe Giants20 min read

    Making Immersive LLM Chat Experiences More Accessible

    As LLM chat patterns become more widespread, making them accessible is essential to ensuring everyone can access the information, tools, and opportunities they offer. Drawing on his work improving these interfaces, Eric Bailey shares practical ways to make chat-based experiences on the web more accessible and usable for people who rely on assistive technology.

Wednesday, 4 November 2026

  • CSS-TricksThe Giantsread at source

    Smashing AI Conference

    The AI 2027 Conference is a live online day for designers and engineers on where AI stands today, what to expect in 2027, and where teams are getting it right. Smashing AI Conference originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Tuesday, 3 November 2026

  • CSS-TricksThe Giantsread at source

    Re-Implementing :nth-child() With Modern CSS

    Combining :nth-child with style queries and sibling-index() to conditionally select child elements. Re-Implementing :nth-child() With Modern CSS originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

  • Smashing MagazineThe Giants12 min read

    Modern CSS Might Make Your Front-End More Complex

    Replacing JavaScript with a modern CSS feature can simplify a piece of code while making the wider application harder to reason about. Gabriel Shoyombo explores what happens when logic is split between CSS and JavaScript, sharing practical guidance to help you keep complexity in check during migration.

Monday, 2 November 2026

  • CSS-TricksThe Giantsread at source

    What’s !important #20: GTA 6 Scroll Effects, Scrollspies, random()

    This is the final issue of What’s !important on CSS-Tricks.com. The good news is that it’s coming back in an expanded form as the CSS-Tricks newsletter. I had ya there for a second, didn’t I? Sorry, I couldn’t decide whether to trick or treat, so I went for both. Happy Halloween What’s !important #20: GTA 6 Scroll Effects, Scrollspies, random() originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Saturday, 31 October 2026

  • Smashing MagazineThe Giants11 min read

    A Spark Of Color In November’s Gray (2026 Wallpapers Edition)

    Let’s kick of November with a new collection of desktop wallpapers! Designed by creatives from across the globe, the wallpapers come in a variety of screen resolutions and can be downloaded for free. Enjoy!

Thursday, 29 October 2026

  • CSS-TricksThe Giantsread at source

    The Frankenstein Table

    A super scary way to go about tables using other layout techniques, like CSS Grid. The Frankenstein Table originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Tuesday, 27 October 2026

  • CSS-TricksThe Giantsread at source

    What’s !important #19: CSS Refraction Effects, Scroll State Queries, @custom-media, and More

    We’re back! Don't worry, this isn’t a trick (despite what our name suggests!). However, due to the hiatus, we're also a little behind. That’s why this issue of What’s !important is absolutely stacked. All killer, no filler. Here's what you missed. What’s !important #19: CSS Refraction Effects, Scroll State Queries, @custom-media, and More originally handwritten and published with love on CSS-Tricks . You should really get the newsletter as well.

Articles belong to their publishers. This site only collects what their feeds provide.

Updated 22 November 2026 at 15:10