https://nick-liu.com/categories/development-environment/ Recent content in Development Environment on Nick Liu - Software Engineer Hugo -- gohugo.io en nickboy@users.noreply.github.com (Nick Liu) nickboy@users.noreply.github.com (Nick Liu) 2026 Nick Liu Fri, 10 Jul 2026 20:05:23 -0700 https://nick-liu.com/posts/cloudflare-web-analytics-hugo/ Fri, 10 Jul 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/cloudflare-web-analytics-hugo/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> For six months I assumed nobody could tell whether anyone read this blog, because I had never added analytics. Wiring up Cloudflare Web Analytics by hand taught me two things: the obvious place to paste the snippet would have shipped my Playwright suite's page views into the dashboard, and the dashboard had been quietly counting my visitors for two months anyway. </div> <div class="tested-with" role="note"> <span class="tested-with-icon" aria-hidden="true">🧪</span> <span>Tested with <strong>Hugo</strong> <code>0.163.3</code> · Blowfish 2.104</span> </div> <h2 class="relative group">Publishing into the void <div id="publishing-into-the-void" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#publishing-into-the-void" aria-label="Anchor">#</a> </span> </h2> <p>The site&rsquo;s <code>hugo.toml</code> had a <code>googleAnalytics</code> line commented out since roughly the first commit. I never uncommented it. GA4 wants a cookie disclosure, ships a chunky client, and ad blockers eat it anyway, which felt like a lot of ceremony for a personal blog whose one open question was &ldquo;does anybody visit.&rdquo;</p> https://nick-liu.com/posts/hugo-baseurl-dual-domain-canonical/ Fri, 10 Jul 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/hugo-baseurl-dual-domain-canonical/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Paste one of my post links into a social preview and the card comes up with no image. The site itself renders fine, every page, every browser. The culprit was one character in `hugo.toml`: `baseURL = "/"`, which quietly turns every absolute URL the site emits into a relative one that only a browser can love. </div> <div class="tested-with" role="note"> <span class="tested-with-icon" aria-hidden="true">🧪</span> <span>Tested with <strong>Hugo</strong> <code>0.163.3</code> · Blowfish 2.104</span> </div> <h2 class="relative group">The symptom <div id="the-symptom" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#the-symptom" aria-label="Anchor">#</a> </span> </h2> <p>Share cards without images, that was the visible part. View source on any page and the metadata told the fuller story:</p> https://nick-liu.com/posts/content-pr-ci-path-filter/ Sun, 21 Jun 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/content-pr-ci-path-filter/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> I changed one sentence in a blog post, opened a pull request, and watched CI spend about 85 seconds installing a headless Chromium to confirm my prose still turned into HTML. The obvious fix, telling the workflow to ignore content paths, would have quietly broken every merge instead. </div> <h2 class="relative group">Symptom: a browser for a one-line edit <div id="symptom-a-browser-for-a-one-line-edit" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#symptom-a-browser-for-a-one-line-edit" aria-label="Anchor">#</a> </span> </h2> <p>Every pull request on this blog runs two jobs: a <code>lint</code> job, and a <code>build-and-test</code> job. The second one builds the site with Hugo, link-checks the output with htmltest, then installs a headless Chromium and runs a Playwright suite against a live <code>hugo server</code>. End to end, roughly 85 seconds, and almost all of it is Playwright.</p> https://nick-liu.com/posts/lru-session-eviction/ Sun, 10 May 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/lru-session-eviction/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> A Stream Deck has five session keys. I usually have six or seven Claude Code sessions running. When a new one shows up, the muscle memory test isn't "does the right session get evicted", it is "do the four survivors stay on the keys they were already on." </div> <nav class="series-progress" aria-label="Series progress"> <span class="series-progress-label">Building ClaudeDeck · Part 3 of 10</span> <ol class="series-progress-dots"><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/claude-code-hooks-reality/" title="The Claude Code hooks docs are wrong. Here&#39;s what&#39;s actually on the wire.">1</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/streamdeck-sdk-quirks/" title="Two Stream Deck SDK quirks that cost me a weekend">2</a> </li><li class="series-progress-dot is-current"> <a href="https://nick-liu.com/posts/lru-session-eviction/" title="Five Stream Deck keys, N Claude sessions: LRU that keeps the order I see" aria-current="step">3</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/statusline-side-channel/" title="The Claude Code statusline is a per-turn telemetry side channel">4</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/plan-usage-statusline-pivot/" title="I polled an undocumented endpoint for 18 hours. The data was on stdin.">5</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/permission-round-trip/" title="Holding HTTP open for 590 seconds so a Stream Deck key can approve a tool call">6</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/pty-wrap-migration/" title="I split my daemon in two so a Node subprocess could own the PTY">7</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" title="TCC pins your Accessibility grant to a cdhash. Every rebuild breaks it.">8</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/tahoe-hotkey-dead-end/" title="What replaced CGEventPost in my Stream Deck daemon">9</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/launchd-bootstrap-debugging/" title="launchctl unload returned 0. The daemon was still running. KeepAlive raced.">10</a> </li></ol> </nav> <div class="tested-with" role="note"> <span class="tested-with-icon" aria-hidden="true">🧪</span> <span>Tested with <strong>Claude Code</strong> <code>2.1.x</code> · macOS</span> </div> <p>(Two bits of context for anyone new to the stack: Stream Deck is Elgato&rsquo;s USB grid of programmable LCD keys, and a &ldquo;session&rdquo; here is a single Claude Code conversation: <code>claude</code> running in one terminal tab, with its own working directory, its own context window, its own history. LRU stands for &ldquo;least-recently used,&rdquo; the standard cache-eviction policy: when you need to make room, drop the entry nobody has touched in the longest time.)</p> https://nick-liu.com/posts/launchd-bootstrap-debugging/ Sun, 10 May 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/launchd-bootstrap-debugging/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> `launchctl unload ~/Library/LaunchAgents/com.nickboy.claudedeck.plist` exited 0. Then `pgrep -f claudedeck-daemon` printed a fresh PID. Three seconds after the "unload succeeded" line. Spoiler: KeepAlive is a polling supervisor, not an event-driven one, and when you tell launchd to tear a job down, there is a window where the supervisor has already noticed the previous PID is gone and started a replacement. </div> <nav class="series-progress" aria-label="Series progress"> <span class="series-progress-label">Building ClaudeDeck · Part 10 of 10</span> <ol class="series-progress-dots"><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/claude-code-hooks-reality/" title="The Claude Code hooks docs are wrong. Here&#39;s what&#39;s actually on the wire.">1</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/streamdeck-sdk-quirks/" title="Two Stream Deck SDK quirks that cost me a weekend">2</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/lru-session-eviction/" title="Five Stream Deck keys, N Claude sessions: LRU that keeps the order I see">3</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/statusline-side-channel/" title="The Claude Code statusline is a per-turn telemetry side channel">4</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/plan-usage-statusline-pivot/" title="I polled an undocumented endpoint for 18 hours. The data was on stdin.">5</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/permission-round-trip/" title="Holding HTTP open for 590 seconds so a Stream Deck key can approve a tool call">6</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/pty-wrap-migration/" title="I split my daemon in two so a Node subprocess could own the PTY">7</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" title="TCC pins your Accessibility grant to a cdhash. Every rebuild breaks it.">8</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/tahoe-hotkey-dead-end/" title="What replaced CGEventPost in my Stream Deck daemon">9</a> </li><li class="series-progress-dot is-current"> <a href="https://nick-liu.com/posts/launchd-bootstrap-debugging/" title="launchctl unload returned 0. The daemon was still running. KeepAlive raced." aria-current="step">10</a> </li></ol> </nav> <p>(One-paragraph grounding if launchd isn&rsquo;t your daily driver: <em>launchd</em> is macOS&rsquo;s init system, the equivalent of <code>systemd</code> on Linux or Windows Services on Windows. It boots PID 1, brings up daemons, restarts them when they crash. A <em>LaunchAgent</em> is a per-user launchd job, defined by an XML <em>plist</em> (property list) at <code>~/Library/LaunchAgents/&lt;name&gt;.plist</code>. <em>KeepAlive</em> is one of the plist keys; set it to <code>true</code> and launchd will respawn the job whenever it exits. <em><code>launchctl</code></em> is the CLI you use to load, unload, and inspect those jobs. The Linux mental model: think <code>systemctl</code> driving <code>systemd</code> unit files. The Stream Deck plugin and its daemon are described in <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" >the TCC cdhash trap post</a> if you want the project context.)</p> https://nick-liu.com/posts/tcc-cdhash-trap/ Sun, 10 May 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/tcc-cdhash-trap/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> My daemon's preflight log said `osascript is not allowed assistive access. (-1719)`. System Settings disagreed: the entry was right there, toggled on. Spoiler: ad-hoc codesigning pins TCC's designated requirement to the binary's cdhash, and `bun build --compile` produces a different cdhash on every rebuild. </div> <nav class="series-progress" aria-label="Series progress"> <span class="series-progress-label">Building ClaudeDeck · Part 8 of 10</span> <ol class="series-progress-dots"><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/claude-code-hooks-reality/" title="The Claude Code hooks docs are wrong. Here&#39;s what&#39;s actually on the wire.">1</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/streamdeck-sdk-quirks/" title="Two Stream Deck SDK quirks that cost me a weekend">2</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/lru-session-eviction/" title="Five Stream Deck keys, N Claude sessions: LRU that keeps the order I see">3</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/statusline-side-channel/" title="The Claude Code statusline is a per-turn telemetry side channel">4</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/plan-usage-statusline-pivot/" title="I polled an undocumented endpoint for 18 hours. The data was on stdin.">5</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/permission-round-trip/" title="Holding HTTP open for 590 seconds so a Stream Deck key can approve a tool call">6</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/pty-wrap-migration/" title="I split my daemon in two so a Node subprocess could own the PTY">7</a> </li><li class="series-progress-dot is-current"> <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" title="TCC pins your Accessibility grant to a cdhash. Every rebuild breaks it." aria-current="step">8</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/tahoe-hotkey-dead-end/" title="What replaced CGEventPost in my Stream Deck daemon">9</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/launchd-bootstrap-debugging/" title="launchctl unload returned 0. The daemon was still running. KeepAlive raced.">10</a> </li></ol> </nav> <p>I&rsquo;m building a Stream Deck plugin called <a href="https://github.com/nickboy/claudedeck" target="_blank" rel="noreferrer">ClaudeDeck</a> (Stream Deck is Elgato&rsquo;s little USB grid of programmable keys with LCD displays under each one). The plugin talks to a background daemon (a long-running process that starts at login and waits for events), and that daemon needs to call System Events via AppleScript to switch Ghostty tabs (Ghostty is my terminal emulator) whenever I press a Stream Deck key. macOS gates that capability, automating other apps, through <strong>System Settings → Privacy &amp; Security → Accessibility</strong>, the pane you&rsquo;ve probably toggled for tools like Rectangle or BetterTouchTool. On first install I added the daemon, toggled it on, and got back to work.</p> https://nick-liu.com/posts/streamdeck-sdk-quirks/ Sun, 10 May 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/streamdeck-sdk-quirks/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Two undocumented behaviours in the Elgato Stream Deck SDK ate most of a weekend: a per-key title-alignment cache that silently ignores manifest updates, and a `willAppear` event that doesn't always re-fire after a plugin restart. The fixes are short. Finding them was not. </div> <nav class="series-progress" aria-label="Series progress"> <span class="series-progress-label">Building ClaudeDeck · Part 2 of 10</span> <ol class="series-progress-dots"><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/claude-code-hooks-reality/" title="The Claude Code hooks docs are wrong. Here&#39;s what&#39;s actually on the wire.">1</a> </li><li class="series-progress-dot is-current"> <a href="https://nick-liu.com/posts/streamdeck-sdk-quirks/" title="Two Stream Deck SDK quirks that cost me a weekend" aria-current="step">2</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/lru-session-eviction/" title="Five Stream Deck keys, N Claude sessions: LRU that keeps the order I see">3</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/statusline-side-channel/" title="The Claude Code statusline is a per-turn telemetry side channel">4</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/plan-usage-statusline-pivot/" title="I polled an undocumented endpoint for 18 hours. The data was on stdin.">5</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/permission-round-trip/" title="Holding HTTP open for 590 seconds so a Stream Deck key can approve a tool call">6</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/pty-wrap-migration/" title="I split my daemon in two so a Node subprocess could own the PTY">7</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" title="TCC pins your Accessibility grant to a cdhash. Every rebuild breaks it.">8</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/tahoe-hotkey-dead-end/" title="What replaced CGEventPost in my Stream Deck daemon">9</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/launchd-bootstrap-debugging/" title="launchctl unload returned 0. The daemon was still running. KeepAlive raced.">10</a> </li></ol> </nav> <p>(Skip this paragraph if you&rsquo;ve shipped a Stream Deck plugin before. The Stream Deck is Elgato&rsquo;s USB grid of programmable LCD keys, common on streamer desks for scene switching. A &ldquo;plugin&rdquo; is a small program (TypeScript, in my case) that runs as a child process of Elgato&rsquo;s Stream Deck app, registers one or more <em>actions</em> the user can drag onto keys, and reacts to events like &ldquo;key pressed&rdquo; or &ldquo;key visible.&rdquo; The SDK is <code>@elgato/streamdeck</code> from npm. A <em>manifest</em> is a <code>manifest.json</code> next to the plugin that declares its actions, supported devices, default icons, and per-state defaults like title alignment.)</p> https://nick-liu.com/posts/tahoe-hotkey-dead-end/ Sun, 10 May 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/tahoe-hotkey-dead-end/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> I press the Stream Deck key. The daemon logs the press, synthesizes `Cmd+Opt+;` through CoreGraphics, and exits cleanly. Wispr Flow does nothing. Three Apple subsystems and one decompiled Electron bundle later, the working trigger turned out to be a one-line URL. </div> <nav class="series-progress" aria-label="Series progress"> <span class="series-progress-label">Building ClaudeDeck · Part 9 of 10</span> <ol class="series-progress-dots"><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/claude-code-hooks-reality/" title="The Claude Code hooks docs are wrong. Here&#39;s what&#39;s actually on the wire.">1</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/streamdeck-sdk-quirks/" title="Two Stream Deck SDK quirks that cost me a weekend">2</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/lru-session-eviction/" title="Five Stream Deck keys, N Claude sessions: LRU that keeps the order I see">3</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/statusline-side-channel/" title="The Claude Code statusline is a per-turn telemetry side channel">4</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/plan-usage-statusline-pivot/" title="I polled an undocumented endpoint for 18 hours. The data was on stdin.">5</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/permission-round-trip/" title="Holding HTTP open for 590 seconds so a Stream Deck key can approve a tool call">6</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/pty-wrap-migration/" title="I split my daemon in two so a Node subprocess could own the PTY">7</a> </li><li class="series-progress-dot is-done"> <a href="https://nick-liu.com/posts/tcc-cdhash-trap/" title="TCC pins your Accessibility grant to a cdhash. Every rebuild breaks it.">8</a> </li><li class="series-progress-dot is-current"> <a href="https://nick-liu.com/posts/tahoe-hotkey-dead-end/" title="What replaced CGEventPost in my Stream Deck daemon" aria-current="step">9</a> </li><li class="series-progress-dot"> <a href="https://nick-liu.com/posts/launchd-bootstrap-debugging/" title="launchctl unload returned 0. The daemon was still running. KeepAlive raced.">10</a> </li></ol> </nav> <p>The plan was the boring kind: Stream Deck key (the physical button on Elgato&rsquo;s programmable USB grid) → WebSocket message → my daemon (long-running background process) → synthesized global hotkey → Wispr Flow&rsquo;s hands-free dictation starts (Wispr Flow is the voice-to-text Mac app that types your speech into the focused window) → I talk → words show up in my editor. I&rsquo;d done variants of this with <code>osascript</code> (macOS&rsquo;s command-line AppleScript runner) years ago. Should have taken an afternoon.</p> https://nick-liu.com/posts/obsidian-claude-code-knowledge-base/ Sat, 21 Mar 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/obsidian-claude-code-knowledge-base/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Obsidian + Claude Code is everywhere right now. But pointing an AI at a folder of markdown files and hoping for the best doesn't work. What matters is how you structure the knowledge base. Get that right, and Claude becomes genuinely useful. Get it wrong, and you get confident garbage. </div> <p>There&rsquo;s been a wave of posts about this combo lately: <a href="https://x.com/jameesy/status/2026628809424781787" target="_blank" rel="noreferrer">James Bedford&rsquo;s full walkthrough</a>, <a href="https://x.com/gregisenberg/status/2026036464287412412" target="_blank" rel="noreferrer">Greg Isenberg&rsquo;s &ldquo;personal OS&rdquo; approach</a>, <a href="https://x.com/kepano/status/2008578873903206895" target="_blank" rel="noreferrer">kepano (Obsidian&rsquo;s CEO) sharing Claude Skills</a>. They&rsquo;re all worth reading.</p> https://nick-liu.com/posts/boris-claude-code-tips/ Sat, 21 Mar 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/boris-claude-code-tips/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Boris Cherny created Claude Code. When he shared how he actually uses it day-to-day, the setup was surprisingly simple. I went through every tip, tried most of them, and have opinions about all of them. </div> <p>The original thread is on <a href="https://x.com/bcherny/status/2007179832300581177" target="_blank" rel="noreferrer">Boris&rsquo;s X account</a>. A good companion site is <a href="https://howborisusesclaudecode.com" target="_blank" rel="noreferrer">howborisusesclaudecode.com</a> which compiles everything in one place.</p> https://nick-liu.com/posts/case-study-aws-billing-auditor/ Sun, 22 Feb 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/case-study-aws-billing-auditor/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> I spent five years on the AWS Billing team. The hardest problem I tackled was detecting when customers used AWS services but weren't charged correctly. This post walks through how I designed a system that reduced charge discrepancies by **300x** and eliminated **230 million** monthly false positives. </div> <h2 class="relative group">The Problem <div id="the-problem" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#the-problem" aria-label="Anchor">#</a> </span> </h2> <p>AWS billing is trickier than it looks. When a customer launches an EC2 instance, writes to S3, or queries DynamoDB, each action generates a usage record. These records flow through a pipeline that calculates charges based on the customer&rsquo;s pricing plan, region, and service tier.</p> https://nick-liu.com/posts/case-study-twitter-fleet-automation/ Sun, 22 Feb 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/case-study-twitter-fleet-automation/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> At Twitter, I was responsible for kernel updates across **5,000+ production servers**. Updating a kernel is risky on one machine. Doing it across a fleet, without downtime, without data loss, and without breaking the services that millions of people depend on, is a different problem entirely. </div> <h2 class="relative group">The Problem <div id="the-problem" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#the-problem" aria-label="Anchor">#</a> </span> </h2> <p>Twitter&rsquo;s production infrastructure ran on thousands of bare-metal servers across multiple data centers. Each server ran a Linux kernel that needed regular updates for security patches, performance improvements, and hardware compatibility.</p> https://nick-liu.com/posts/catppuccin-mocha-theming/ Sun, 22 Feb 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/catppuccin-mocha-theming/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Your development environment should feel like **one cohesive tool**, not a collection of unrelated windows with clashing colors. I theme everything with the same palette: <span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> Catppuccin Mocha </span> </span> . The result is a workspace where context-switching between tools is effortless. </div> <h2 class="relative group">Why One Palette Everywhere? <div id="why-one-palette-everywhere" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#why-one-palette-everywhere" aria-label="Anchor">#</a> </span> </h2> <p>Most developers pick a theme for their editor and call it a day. Their terminal is one color, their editor another, their tmux status bar a third, and their Git diffs something else entirely. Every time they switch contexts, their brain spends a fraction of a second recalibrating.</p> https://nick-liu.com/posts/my-terminal-setup-2026/ Sun, 15 Feb 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/my-terminal-setup-2026/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> After years of refining my terminal workflow, I've landed on a stack I genuinely enjoy using every day: **Ghostty** as the terminal emulator, **tmux** with **sesh** for session management, and **Neovim** with **LazyVim** for editing. </div> <p>Everything runs on macOS (Apple Silicon) with a consistent <span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> Catppuccin Mocha </span> </span> theme across all tools.</p> https://nick-liu.com/posts/modern-cli-tools-replacing-unix-classics/ Sun, 01 Feb 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/modern-cli-tools-replacing-unix-classics/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> I've been gradually replacing classic Unix tools with modern alternatives, mostly written in <span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> Rust </span> </span> . After a year of daily use, these aren't experiments anymore. They're muscle memory. </div> <h2 class="relative group">The Replacements <div id="the-replacements" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#the-replacements" aria-label="Anchor">#</a> </span> </h2> <table> <thead> <tr> <th>Classic</th> <th>Modern</th> <th>Why</th> </tr> </thead> <tbody> <tr> <td><code>cat</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> bat </span> </span> </td> <td>Syntax highlighting, line numbers, git integration</td> </tr> <tr> <td><code>ls</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> eza </span> </span> </td> <td>Icons, git status, tree view, color-coded</td> </tr> <tr> <td><code>grep</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> ripgrep </span> </span> </td> <td>10x faster, respects <code>.gitignore</code>, smart case</td> </tr> <tr> <td><code>find</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> fd </span> </span> </td> <td>Simpler syntax, respects <code>.gitignore</code>, colored output</td> </tr> <tr> <td><code>cd</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> zoxide </span> </span> </td> <td>Learns your habits, fuzzy matching</td> </tr> <tr> <td><code>sed</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> sd </span> </span> </td> <td>Intuitive regex syntax, no escaping nightmare</td> </tr> <tr> <td><code>du</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> dust </span> </span> </td> <td>Visual directory size with a tree view</td> </tr> <tr> <td><code>df</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> duf </span> </span> </td> <td>Colorful, filterable disk usage</td> </tr> <tr> <td><code>top</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> btop </span> </span> </td> <td>Beautiful TUI with mouse support, per-core graphs</td> </tr> <tr> <td><code>ps</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> procs </span> </span> </td> <td>Colorized, searchable, tree view</td> </tr> <tr> <td><code>history</code></td> <td><span class="flex cursor-pointer"> <span class="rounded-md border border-primary-400 px-1 py-[1px] text-xs font-normal text-primary-700 dark:border-primary-600 dark:text-primary-400"> atuin </span> </span> </td> <td>Encrypted sync, full-text search, workspace filtering</td> </tr> </tbody> </table> <h3 class="relative group">Setting Up Aliases <div id="setting-up-aliases" class="anchor"></div> <span class="absolute top-0 w-6 transition-opacity opacity-0 -start-6 not-prose group-hover:opacity-100 select-none"> <a class="text-primary-300 dark:text-neutral-700 !no-underline" href="#setting-up-aliases" aria-label="Anchor">#</a> </span> </h3> <p>In my <code>.zshrc</code>, I alias the classics to their replacements so the transition is invisible:</p> https://nick-liu.com/posts/managing-dotfiles-with-yadm/ Thu, 15 Jan 2026 00:00:00 +0000 nickboy@users.noreply.github.com (Nick Liu) https://nick-liu.com/posts/managing-dotfiles-with-yadm/ <div class="lead text-neutral-500 dark:text-neutral-400 !mb-9 text-xl"> Every developer eventually reaches the point where their configs become too valuable to lose. Here's how I use **yadm** to manage my macOS dotfiles with automated testing, daily maintenance, and a pre-commit workflow that keeps everything in check. </div> <p>For me, the turning point was spending a weekend setting up a new MacBook and realizing I couldn&rsquo;t reproduce my environment reliably. That&rsquo;s when I started managing my dotfiles properly.</p>