I'd been neglecting my GitHub blog for a while. I kept telling myself "I'll clean it up someday," but then I realized I actually needed a proper space to document and share what I was learning. So I decided to rebuild the blog from scratch. Why I Chose Next.js Svelte was gaining momentum at the time, but I figured Next.js was the better choice if I wanted to get the blog done quickly. The learning curve was much shorter sticking with just Next.js rather than learning Svelte from scratch, picking up SvelteKit, and hunting for the right libraries. My old blog was on Next.js 12, and when I tried upgrading via npx, it weirdly wouldn't work properly. So I ended up creating a new project and applying shadcn from there. Now I'm running a static blog with the Next.js 14 + Contentlayer combo. Setting Up the Dev Environment Since the iMac I was using at the time didn't have Node installed, I used Next.js's Docker example. I chose pnpm as the package manager — installing it via Homebrew is dead simple. pnpm create next-app --example with-docker nextjs-docker This single command creates a Docker-based Next.js project. Features a Blog Needs I put together a list of features the blog would need. Most of them can be built by combining Next.js's built-in features with external libraries. Routing — Page navigation handled by Next.js App Router Image rendering — External images in markdown optimized to data URIs Comments — A GitHub-based comment system Blog Layout Here's how the blog is currently structured: Page Role Home Latest posts, search, post list Books Book list and detail views Tags Browse posts by tag About About me Home handles both the post list and search, so there's no need for a separate Posts page. On the Books page, I decided to include the most memorable quotes from each book at the bottom. Converting Markdown to HTML To display blog posts, markdown files need to be converted to HTML. Currently Contentlayer handles this, but I originally built the conversion pipeline manually. The approach was to grab all posts from the posts folder in generateStaticParams, read each one, convert markdown to HTML with unified, and extract metadata from the frontmatter to create Post objects. Here's a comparison of popular tools for markdown conversion: Library Characteristics gray-matter Specialized for frontmatter parsing marked Lots of GitHub stars, battle-tested for 10+ years unified Rich plugin ecosystem with excellent extensibility I went with the unified family, prioritizing extensibility. Even now, I'm combining remark/rehype plugins inside Contentlayer for syntax highlighting, math rendering, external link handling, and more. This article (Japanese) was really helpful when working with the unified package. Visualizing the Learning Journey with a Mind Map While building the blog, another idea came to mind. I was inspired by the notion that attaching meaning to learning helps with retention. What if I created a page where each tag could be viewed as a mind map? Take the things I've studied and posted about, turn them into dots, and connect those dots with lines. For example, if I studied Docker, I could visually track the process: how I came to understand what Docker is, what perspective I developed, and what I'll study next. Being able to see these connections visually is motivating in itself. Looking Back Rebuilding the blog wasn't just about switching up the tech stack. It was about creating a system for organizing what I've learned and sharing that process with others. Rather than waiting for the perfect blog, just building it and improving one thing at a time is the best way to start. No man can succeed in a line of endeavor which he does not like.— Napoleon Hill

These days, posting a short video on social media often gets more engagement than writing a full article. But when you actually try to edit a video, it can feel overwhelming knowing where to start. CapCut is free yet surprisingly powerful, making it a perfect app for beginners. In this post, I'll walk you through the process of editing a video with CapCut and posting it on social media. Getting Started: Installation and Project Setup CapCut is free to download on the App Store and Google Play. Open the app, tap "New Project," and you can pick videos from your gallery and drop them straight onto the timeline. Here's how to think about the basic editing flow: Trim the video — Tap the desired position on the timeline and hit the "Split" button. Done. Add subtitles — Type them manually with "Add Text," or use the auto-caption feature (supports multiple languages) for convenience. Add music — Browse CapCut's built-in music library and pick background music that fits the mood. Effects and transitions — Adding transition effects between scenes makes the video feel much smoother. Before exporting, check that the TikTok watermark is off. Disabling it in settings keeps things clean when uploading to other platforms. Things to Keep in Mind While Shooting Editing matters, but putting in a little effort during the shooting stage makes post-production way easier. Shoot in vertical 9:16 ratio. Both Reels and Shorts default to vertical video. Use natural light — you can get bright, clean footage without any extra lighting. Write a simple script beforehand, and you can finish editing in under 30 minutes. For the script, think in terms of these three parts: Hook (first 3 seconds) — Start with an eye-catching scene or question. Highlight — The core content you want to deliver. CTA — Call to action like "Subscribe," "Like," or "More in the next video." The first 3 seconds are really critical. Place an impactful scene at the very beginning to stop people from scrolling. Sound Mixing: Balancing the Audio Audio balance in a video matters more than you might think. If the music is too loud, voices get buried. Too quiet, and the video feels flat. The basic principle is lowering other sounds relative to the voice. First, set the voice peak to -12 to -6 dB, then adjust everything else proportionally. Element Relative to Voice Absolute dB Example Tip Voice (dialogue) 0 dB (reference) -12 to -6 dB Normalize first Key sound effects -3 to -6 dB -12 to -15 dB Emphasize but slightly lower Background SFX -6 to -10 dB -15 to -18 dB Natural background level Background music -10 to -18 dB -18 to -27 dB Use ducking for auto-adjust You can fine-tune with the volume slider or keyframes on the timeline. The Pro version's audio ducking feature automatically lowers background music when voice is present, which is super convenient. Posting to Social Media Instagram Reels In the Instagram app, tap "+" button then "Reels" and import the video you made in CapCut. You can also add music or stickers within Instagram. 3-5 hashtags related to your content is a good number. If the CapCut logo is still on the video, you can crop it out in Instagram's editing screen. TikTok CapCut is made by ByteDance — the same company behind TikTok — so the integration is the smoothest here. From the CapCut editing screen, tap "Share to TikTok" to upload directly without a separate export. On TikTok, leveraging trending sounds helps with reach. Another solid approach is finishing your base edit in CapCut, then swapping in a trending sound within the TikTok app. Try mixing trending tags with content-related tags, around 5-8 total. YouTube Shorts In the YouTube app, tap "+" button then "Create a Short" to upload your video. Vertical videos are automatically recognized as Shorts. Aim for titles around 15-30 characters, and it's important to use copyright-safe music. Most of CapCut's built-in tracks are cleared for commercial use, but double-check before uploading. Consistency Is Key It's totally fine if your first one takes a while to make. If you post consistently 2-3 times a week, your editing speed will naturally improve and subscribers will gradually build up. Rather than waiting for the perfect video, the best start is making a Reel from whatever footage you have right now. Argue for your limitations, and sure enough they're yours.— Richard Bach

A lot of services are migrating from Java to Kotlin these days. The first thing I noticed about Kotlin was its conciseness. Being able to express the same logic in less than half the code compared to Java was pretty refreshing. In this post, I'll cover the Kotlin syntax I use most in practice. From Null Safety to coroutines — just the essentials. Variables and Type Inference Kotlin has two ways to declare variables. val is immutable, var is mutable. val name: String = "John" val age = 30 // Type inferred as Int var count: Int = 0 count = 1 const val MAX_SIZE = 100 // Compile-time constant Unlike Java where you had to keep adding final, Kotlin makes val the default so you naturally lean toward immutability. String handling is also much cleaner. With string templates and raw strings, you almost never need to reach for StringBuilder. val json = """ { "name": "$name", "age": $age } """.trimIndent() Null Safety This is one of the biggest reasons to use Kotlin. Since all types are non-null by default, you can catch NullPointerException at compile time. var nickname: String? = null // ? allows null nickname?.length // Safe call: returns null if null nickname!!.length // Non-null assertion: crashes if null val len = nickname?.length ?: 0 // Elvis operator: default value when null In practice, the most common pattern is combining let with the Elvis operator. nickname?.let { println("Nickname: $it") } fun process(name: String?) { val n = name ?: return // Early return if null println(n.uppercase()) } Avoid using the !! operator outside of test code. It's basically reintroducing NPEs. Functions and Extension Functions Default values and named arguments mean problems that required overloading in Java can be solved with a single function. fun greet(name: String = "Guest", age: Int = 0) = "Hello, $name (age $age)" greet(age = 30, name = "John") // Order doesn't matter Extension functions let you add methods to existing classes without modifying them. Especially useful for utility functions. fun String.toSlug() = this.lowercase().replace(" ", "-").replace(Regex("[^a-z0-9-]"), "") "Hello World".toSlug() // "hello-world" Classes: Data Class and Sealed Class Data Class When creating DTOs or model objects, it auto-generates equals(), hashCode(), toString(), and copy(). You can completely eliminate Java boilerplate. data class User( val id: Long, val name: String, val email: String ) val user2 = user1.copy(name = "Jane") // Copy with only some fields changed Sealed Class A class specialized for state management. When used with when expressions, the compiler checks all cases so you don't need an else branch. sealed class Result&x3C;out T> { data class Success&x3C;T>(val data: T) : Result&x3C;T>() data class Error(val message: String) : Result&x3C;Nothing>() object Loading : Result&x3C;Nothing>() } fun handleResult(result: Result&x3C;String>) = when (result) { is Result.Success -> "Success: ${result.data}" is Result.Error -> "Error: ${result.message}" Result.Loading -> "Loading..." } When you add a new state, you'll get compile errors everywhere when is used, so nothing gets missed. Control Flow as Expressions In Kotlin, both if and when are expressions that return values. val max = if (a > b) a else b val message = when (status) { "PENDING" -> "Pending" "CONFIRMED" -> "Confirmed" in listOf("CANCELLED", "FAILED") -> "Failed" else -> "Unknown" } There's no ternary operator (? :) in Java's sense — if itself fills that role. when is far more flexible than switch, supporting range checks and type checks. Functional Collection Processing Kotlin's collection API is more concise than Java Streams. val numbers = listOf(1, 2, 3, 4, 5) numbers.filter { it > 2 } // [3, 4, 5] numbers.map { it * 2 } // [2, 4, 6, 8, 10] numbers.groupBy { it % 2 == 0 } // {false=[1,3,5], true=[2,4]} numbers.chunked(2) // [[1,2], [3,4], [5]] No more .stream().collect(Collectors.toList()). Scope Functions The five scope functions can be confusing at first, but the key is the combination of reference style (it vs this) and return value (lambda result vs the object itself). Function Reference Returns Use Case let it Lambda result Null check, transform run this Lambda result Configure then compute with this Lambda result Multiple method calls apply this Object itself Object initialization also it Object itself Logging, side effects The most common combo in practice is apply + also. val user = User().apply { name = "John" age = 30 }.also { log("User created: $it") } Use apply to initialize the object, then also to handle side effects (like logging). Coroutines Kotlin's async handling is powered by coroutines. The learning curve is lower than RxJava, and the code reads like synchronous code. Basic API Call viewModelScope.launch { _loading.value = true _user.value = repository.getUser(id) _loading.value = false } Parallel Calls A task that takes 3 seconds sequentially can be cut to 1 second with async for parallel processing. coroutineScope { val user = async { fetchUser() } val posts = async { fetchPosts() } val comments = async { fetchComments() } Triple(user.await(), posts.await(), comments.await()) } Thread Switching with Dispatchers viewModelScope.launch { showLoading() // Main thread val user = withContext(Dispatchers.IO) { api.getUser(id) // IO thread } showUser(user) // Back to Main thread } Even with withContext switching threads, you can read top to bottom without callback hell. Java Interop If you need compatibility with existing Java code, use annotations like @JvmStatic and @JvmField. class Utils { companion object { @JvmStatic fun helper() { } } } // Callable from Java as Utils.helper() Wrap-Up Kotlin is ultimately a language that aims for safe and concise code. Among everything covered here, the three things that made the biggest difference in practice were: Null Safety — Elevates runtime crashes to compile-time errors Sealed Class + when — Branch handling with no missed states Coroutines — Write async code without callbacks If you want to go deeper, try practicing with Kotlin Koans. We learn what we have said from those who listen to our speaking.— Kenneth Patton