<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Mayowa Fasugba on Medium]]></title>
        <description><![CDATA[Stories by Mayowa Fasugba on Medium]]></description>
        <link>https://medium.com/@mayowafasugba?source=rss-e92172a5665b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ghPVRjZ9Klvxp9weh5GMGg.jpeg</url>
            <title>Stories by Mayowa Fasugba on Medium</title>
            <link>https://medium.com/@mayowafasugba?source=rss-e92172a5665b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 13:51:58 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mayowafasugba/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Why Code Can’t Chill a Box]]></title>
            <link>https://journal.messyfounder.com/why-code-cant-chill-a-box-7825cef06aa9?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/7825cef06aa9</guid>
            <category><![CDATA[nigeria]]></category>
            <category><![CDATA[founder-stories]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[cold-chain]]></category>
            <category><![CDATA[founders]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 00:23:35 GMT</pubDate>
            <atom:updated>2026-11-13T00:23:35.051Z</atom:updated>
            <content:encoded><![CDATA[<p>What I learned going from dashboard to hardware</p><figure><img alt="ColdStack first logo" src="https://cdn-images-1.medium.com/max/512/1*0zS16u5PaRMUqoMPTPPOkg.png" /></figure><p>I had a live map, a clean dashboard, and alerts that fired when a shipment drifted out of range. None of it was watching anything real, because there was no box to watch.</p><h3>Starting with what I know</h3><p>I’m a software developer, so when I started looking at Nigeria’s cold chain (fish, poultry, produce, vaccine, anything that spoils), I started where I was comfortable. I built it with Next.js on the front end and Supabase for the database, sign-in, and real-time updates.</p><p>The first version came together in this order:</p><ul><li><strong>Accounts and roles.</strong> Admin, carrier, shipper and consignee, each seeing only their own world.</li><li><strong>Trips and live updates.</strong> A trip created in one place had to show up for everyone involved without a refresh.</li><li><strong>A live map.</strong> Trip coordinates stored in PostGIS, drawn on a map you could watch move.</li><li><strong>Alerts.</strong> Every trip got its own target temperature, because frozen chicken and fresh fish don’t share one safe range, and the system flagged anything outside it.</li><li><strong>A Proof of Condition report.</strong> When a trip ends, the app generates a report you can hand to someone who wasn’t on the road.</li><li><strong>A phone-friendly version.</strong> Drivers needed something that worked on a phone.</li></ul><p>It wasn’t smooth. I chased screens that wouldn’t update until you refreshed them, and I fought with how Next.js rendered pages. Next.js got me a working product fast, but I later moved the app to React Native. That’s a story for a later post.</p><p>To test all of this without hardware, I wrote a telemetry simulator: a fake box that sent fake readings. The location pings are real now, but the temperature is still simulated.</p><figure><img alt="ColdStack Dashboard" src="https://cdn-images-1.medium.com/proxy/1*hmuZOIdw7XKD9vHZpwiT5Q.png" /><figcaption>ColdStack Dashboard Map View</figcaption></figure><h3>The first reality check</h3><p>The simulator did its job, and then it showed me the problem. I was staring at a perfect temperature curve for a box that didn’t exist.</p><p>A dashboard can show you spoilage in high definition, but it can’t stop it. What keeps cargo cold is physical: insulation, power, sealing, and a cooling system that works in heat and survives bad roads.</p><p>So I started researching hardware, and quickly learned I couldn’t just send a design to a factory. Traditional manufacturers generally want large minimum orders and expensive tooling before they’ll make a single box. And I had neither the tools nor the skills to build one myself.</p><h3>Finding my co-founder</h3><p>I needed someone who thinks in materials, power, and physical components the way I think in code. That person turned out to be David, now ColdStack’s co-founder and Head of Hardware.</p><p>David and I have been friends for over 4 years, and we studied at the same university. I thought of him because he is an electrical and electronics engineer who has built multiple physical products. While I was doing my hardware research, I realised that it’s either I take a deep dive into engineering, or I ask for help from someone who already has that domain knowledge and experience. That was when I thought of David, and he immediately loved the concept and jumped into doing his own research on what a stack should be.</p><p>The first thing that changed after he joined: we stopped calling ColdStack a dashboard and started calling it what it is, a cold stack with tracking software.</p><h3>What’s next</h3><p>We’re now planning Run Zero, our first real test. We’re outsourcing the build; we don’t even have a 3D model yet, and we’ll hand over a performance spec instead: what the box must do and how we’ll test it.</p><p>In the next post, I’ll walk through how we specced the first Stack, the trade-offs we made, and what we expect to learn once we hit the road.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7825cef06aa9" width="1" height="1" alt=""><hr><p><a href="https://journal.messyfounder.com/why-code-cant-chill-a-box-7825cef06aa9">Why Code Can’t Chill a Box</a> was originally published in <a href="https://journal.messyfounder.com/">Messy Founder</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The ₦3.5 Trillion Rot: Why Nigeria’s Cold Chain Is Broken, and What We’re Building to Fix It]]></title>
            <link>https://journal.messyfounder.com/the-3-5-trillion-rot-why-nigerias-cold-chain-is-broken-and-what-we-re-building-to-fix-it-e88784f1d429?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/e88784f1d429</guid>
            <category><![CDATA[enterpreneurship]]></category>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[founders]]></category>
            <category><![CDATA[cold-chain-logistics]]></category>
            <category><![CDATA[cold-chain]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Wed, 04 Nov 2026 00:34:40 GMT</pubDate>
            <atom:updated>2026-11-04T00:34:40.456Z</atom:updated>
            <content:encoded><![CDATA[<p>Right now, somewhere on a Lagos road, a dispatch rider is sitting in dead-stopped traffic. Strapped to his bike is a plastic cooler packed with ice and blood samples headed for a diagnostic lab. The ice is melting, and the temperature inside is climbing. The clinic manager waiting on that delivery has no way to know.</p><p>That’s compliance roulette.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*mi6OUhfO4u1urpWVL63RfQ.jpeg" /><figcaption>make-shift cold chain solution</figcaption></figure><p>The scale of the problem is hard to overstate. The industry body <a href="https://www.thecable.ng/expert-nigeria-records-n3-5trn-post-harvest-losses-annually-reliable-cold-chain-system-needed/">OTACCWA estimates that Nigeria loses over ₦3.5 trillion worth of food after harvest every year</a>, about 40% of production. That’s about $2.4 billion, and other estimates run higher, from $4.5 billion to over $9 billion. The infrastructure gap is just as stark: a US International Trade Administration report counts <a href="https://www.freshplaza.com/europe/article/9727693/nigeria-s-cold-chain-gap-threatens-food-supply/">fewer than 1,000 cold trucks in Nigeria against roughly 25,000 needed</a>. I don’t have a clean number for vaccines and lab samples, but the failure is the same, and harder to see.</p><p>Today, operators are stuck between two bad options:</p><ol><li><strong>The CapEx nightmare.</strong> A refrigerated truck costs tens of millions of naira. For a mid-sized lab or independent distributor, that’s a bet the business can’t afford, on a vehicle that spends much of its life stuck in gridlock.</li><li><strong>Ice and prayer.</strong> Pack sensitive goods in a cooler with ice packs and hope the rider beats the traffic. There’s no telemetry, no record, and no data to show NAFDAC that Good Distribution Practice (GDP) conditions held.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*fwwU-39_lbOgehMawfCi6Q.jpeg" /><figcaption>post-harvest losses</figcaption></figure><p>Software alone can’t fix this. You can build the best logistics dashboard in the world, but software can’t chill a box. You have to get your hands dirty with hardware that survives vibration, heat, and bad roads.</p><p>That’s why we are building ColdStack.</p><p>ColdStack pairs modular, insulated, battery-powered cold units, which we call Stacks, with a detachable IoT logger and a real-time software platform. Any dry van, hatchback, or motorbike can become a monitored cold-chain vehicle, with no refrigerated truck required. The thinking behind it:</p><ul><li><strong>Cold that doesn’t depend on the driver.</strong> The Stack runs on its own battery, so it doesn’t matter if the engine is off or the driver takes a break.</li><li><strong>Data that survives dead zones.</strong> The logger stores readings when the network drops and syncs when it returns.</li><li><strong>Proof, not promises.</strong> Every trip streams temperature and location to the carrier, shipper, and receiver. If a shipment leaves its target range, the right people are alerted before the inventory is ruined, and a Proof of Condition report is generated when the trip ends.</li></ul><p>We’re not building coolers. We’re building the data trail that cold-chain compliance depends on.</p><p><strong>Where we are today (the messy part).</strong> The MVP app is built, with live tracking, deviation alerts, and Proof of Condition reports. The temperature data in it is still simulated. The first Stack hasn’t been built yet. We’re about to outsource it, without even a 3D model, and <strong>Run Zero</strong> will tell us what survives contact with a real road.</p><p>In the next post, I’ll go back to the start: how the Next.js and Supabase backend came together, the first reality check on hardware, and how I got my co-founder and Head of Hardware.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e88784f1d429" width="1" height="1" alt=""><hr><p><a href="https://journal.messyfounder.com/the-3-5-trillion-rot-why-nigerias-cold-chain-is-broken-and-what-we-re-building-to-fix-it-e88784f1d429">The ₦3.5 Trillion Rot: Why Nigeria’s Cold Chain Is Broken, and What We’re Building to Fix It</a> was originally published in <a href="https://journal.messyfounder.com/">Messy Founder</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Built A Lyrics Translation and Romanization App]]></title>
            <link>https://blog.stackademic.com/i-built-a-lyrics-translation-and-romanization-app-a50aa5e7cb67?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/a50aa5e7cb67</guid>
            <category><![CDATA[android-app-development]]></category>
            <category><![CDATA[backend-development]]></category>
            <category><![CDATA[language-translation]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[indiedev]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Mon, 26 Oct 2026 00:01:06 GMT</pubDate>
            <atom:updated>2026-10-26T00:01:17.640Z</atom:updated>
            <content:encoded><![CDATA[<p>From hacky Spotify interceptions to zero-latency edge engineering: The Evolution of PolyTune.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/222/1*imyGHeC-_7RIY5ZQk9h_lg.png" /><figcaption>PolyTune Logo</figcaption></figure><p>Still in the spirit of “solving my own problems”, I discovered one that not just me, but a lot of people all around the world face — Google searching for lyrics. As normal as it sounds, many people go online to search for song lyrics, translations, or romanizations so they can sing along. But that friction -playing a song and then searching for the lyrics- can be inconvenient, especially when you do it over and over for different songs(or even the same song).</p><blockquote>So I thought to myself, “<strong>Why don’t I just build an app that does the searching for me?</strong>” That was the first thought that planted the seed in my head.</blockquote><p>That was when I began researching where I can get lyrics from, and how I can build an auto-scroll mechanic that flows with the lyrics and song.</p><p>I decided not to use web scrapers to get the lyrics from lyrics sites because the implementation of that will come with a lot of drama. Websites now have guards against random scrapers, and building that will come with a lot of engineering overhead.</p><p>I came to the conclusion of building a react native (Expo) application, with edge functions in Supabase.</p><p>Over the weeks of building, multiple changes have taken place. What started as <strong>LyricsMine</strong>, a tool to help me grab lyrics from the internet, translate, and Romanize my favorite polyglot tracks, eventually transformed into <strong>PolyTune</strong>: a standalone, hyper-scalable mobile application powered by Supabase Edge Functions, local JS transpilers, and Google Gemini.</p><p>Here is the full engineering breakdown of how PolyTune evolved through four architectural generations, the edge runtime battles I fought along the way, and how I built a zero-latency lyric engine designed to scale to millions of users.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/607/1*Lb6I17PF8jvwIB201ePKoQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/606/1*I4mBxmVZ1aKlfX-sfOvFGw.jpeg" /><figcaption>PolyTune screenshot</figcaption></figure><h3>Generation 1: The Spotify Era &amp; The Brute-Force LLM Pipeline (LyricsMine)</h3><p>Before PolyTune became a standalone React Native app, the project was called <strong>LyricsMine. </strong>The initial goal was simple: bridge the gap when Spotify displayed <em>“Lyrics not available”</em> or served just the non-Latin scripts of a song.</p><h4>1. The Original Ingestion Setup</h4><p>LyricsMine was using the Spotify API to get the metadata of the currently playing song for the logged-in user, and then the app passed that metadata to the backend, which did the fetching, romanization, and translation, and then saved the result into the database so the next time a song plays, the backend checks if it was saved in the database and then passes it immediately to the application. The backend was a single Supabase Edge Function that ingests .lrc timestamped lyrics from a single source: <strong>LRCLIB</strong>. If LRCLIB missed, the app was dead in the water.</p><h4>2. The Brute-Force AI Translation</h4><p>To produce Romanized pronunciation and English translations, I originally fired the entire raw LRC payload into a heavy LLM pipeline. I ran <strong>Groq (Llama 3.3–70b)</strong> as the primary streaming model, with <strong>Google Gemini</strong> acting as a backup. I forced the model to return text using custom string delimiters:</p><pre>[00:15.22] ~~~ 夢のよう ~~~ yume no you ~~~ like a dream<br>[00:18.45] ~~~ Hello ~~~ Hello ~~~ Hello</pre><h4>Why Generation 1 Failed</h4><ul><li><strong>The Alignment Bug:</strong> LLMs are probabilistic, not deterministic. When streaming 80+ lines of lyrics, Llama and Gemini would occasionally drop a line, merge two lines, or hallucinate extra commentary. A single missing line shifted the entire array index by one, completely corrupting the timestamp sync for the rest of the song.</li><li>LyricsMine could only work with Spotify because it was using the Spotify API. I wanted something that would be a universal tool, not tied to one music application.</li><li><strong>The August 2026 Model Shift:</strong> Groq was going to decommission the Llama model that I was using in LyricsMine in August, so naturally I had to remove it and start considering other options.</li><li><strong>High Latency &amp; Costs:</strong> Asking an LLM to romanize standard scripts like Hangul or Kanji on every single cache miss meant paying high token costs and waiting 3 to 5 seconds for response streams.</li></ul><h3>Generation 2: Including lyrics.ovh, listening to the Android OS, and swapping AI models</h3><p>To fix the fragility of Generation 1, the pipeline underwent a major refactoring focused on catalog availability and OS integration.</p><h4>1. Listening to the Android OS Media Session</h4><p>Instead of relying on rigid, manual user inputs, the client app was upgraded to hook directly into the Android OS Media Session notification listeners. This allowed the app to automatically detect whatever song was playing anywhere on the device (Spotify, YouTube Music, local media) and immediately trigger the lyric lookup pipeline in the background.</p><h4>2. Expanding Ingestion with lyrics.ovh</h4><p>To prevent total failures when timestamped .lrc files were unavailable, <strong>lyrics.ovh</strong> was integrated as an unsynced plain-text fallback layer. While it lacked timecodes, it ensured users at least received the original text, translations, and romanizations when synced sources were missing.</p><h3>Generation 3: The Spotify-Karaoke Discovery, The Local JS Shift &amp; Edge Runtime Hell</h3><p>This was also when I realized that I can’t have my app doing what it’s doing and carrying “Lyrics” in the name without attracting unwanted attention. That’s when I changed the name to <strong>PolyTune. </strong>In my search for zero-latency romanization, I had a breakthrough when I discovered the open-source <a href="https://github.com/haroldalan/spotify-karaoke"><strong>spotify-karaoke</strong></a> repository by <a href="https://github.com/haroldalan">Harold Alan</a>. Studying its inner mechanics sparked a complete paradigm shift.</p><p>Instead of paying an LLM for the phonetic pronunciations, deterministic, zero-latency local JavaScript packages could be executed right inside the edge worker based on Unicode range detection.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/751/1*2KsXzX9o2sExMFX51uc0lw.png" /></figure><p>A full suite of specialized libraries was integrated:</p><ul><li><strong>Chinese (Hanzi):</strong> pinyin-pro</li><li><strong>Korean (Hangul):</strong> @romanize/korean</li><li><strong>Cyrillic (Russian, Ukrainian):</strong> cyrillic-to-translit-js</li><li><strong>Indic (Devanagari, Hindi):</strong> @indic-transliteration/sanscript</li><li><strong>Tamil:</strong> tamil-romanizer</li><li><strong>Thai:</strong> @dehoist/romanize-thai</li><li><strong>Universal Fallback:</strong> transliteration</li></ul><h4>Surviving Edge Runtime Hell (Deno, CJS, and Dict Files)</h4><p>Shifting 7+ npm libraries into a serverless <strong>Deno Edge Runtime</strong> (Supabase Edge Functions) triggered a wave of worker boot crashes. Modern edge environments do not run standard Node.js, and debugging them required deep runtime engineering.</p><h4>1. The CJS Export Crash (@romanize/korean &amp; tamil-romanizer)</h4><p>Deno’s static analysis requires named imports (import { romanize } from ...) to be explicitly provable at worker boot time. When CJS packages were bundled, Deno threw fatal syntax errors: worker boot error: Uncaught SyntaxError: The requested module &#39;@romanize/korean&#39; does not provide an export named &#39;romanize&#39;</p><p><strong>The Fix:</strong> All library imports were rewritten into <strong>Namespace Imports</strong> (import * as Module), paired with a custom runtime resolver (resolveExport) and lazy memoization (lazy()). By checking the export shape at runtime inside warm isolates rather than boot time, a package error could no longer crash the entire edge function.</p><pre>function resolveExport&lt;T = any&gt;(ns: any, ...names: string[]): T {<br>  const candidates = [ns?.default, ...names.map((n) =&gt; ns?.[n]), ...names.map((n) =&gt; ns?.default?.[n])];<br>  for (const c of candidates) {<br>    if (typeof c === &quot;function&quot;) return c as T;<br>  }<br>  throw new Error(&quot;Could not resolve usable export.&quot;);<br>}</pre><h4>2. The Kuroshiro / Kuromoji Binary Dictionary Wall</h4><p>For Japanese Kanji-to-Romaji conversion, kuroshiro relies on kuromoji, which attempts to load 15MB of binary lexicons (dict/*.dat.gz) off the local Node filesystem (fs). In Deno&#39;s ephemeral, read-only edge sandbox, those filesystem paths failed, causing worker boot loops.</p><p><strong>The Fix:</strong> I routed Japanese through the high-speed universal transliterator on the local layer, delegating complex Kanji context resolution directly to Gemini via structured JSON prompts.</p><h3>Generation 4: The Current Masterpiece — Reverse Engineering &amp; Structured Output.</h3><p>Today, PolyTune runs on a resilient, production-grade edge function built for extreme scale, multi-tier ingestion, and zero line-misalignment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/1*czs33OWS7noDxqmzGRYGCQ.png" /></figure><p>To eliminate missing lyrics for niche and underground tracks, I combined exact match (/get) and fuzzy search (/search) in the LRCLIB layer to maximize lookup success on the primary source before falling through to backup providers, and I reverse-engineered the network layer of Musixmatch’s official Windows desktop client (since I couldn’t afford to pay for it).</p><h4>1. Spoofing the Desktop Handshake</h4><p>The Edge Function mimics the desktop app by requesting a session token from [https://apic-desktop.musixmatch.com/ws/1.1/token.get](https://apic-desktop.musixmatch.com/ws/1.1/token.get) using a fake Windows Chrome User-Agent and app_id=&quot;web-desktop-app-v1.0&quot;.</p><ul><li><strong>Token Lifecycle:</strong> The acquired user_token is cached at the edge module level. If Musixmatch throws a 401 Unauthorized, the function instantly invalidates the promise and fetches a fresh token on the next tick.</li></ul><pre>let musixmatchTokenPromise: Promise&lt;string&gt; | null = null;<br><br>function getMusixmatchToken(): Promise&lt;string&gt; {<br>  if (!musixmatchTokenPromise) {<br>    musixmatchTokenPromise = fetchMusixmatchToken().catch((e) =&gt; {<br>      musixmatchTokenPromise = null; // Reset cache on fetch failure<br>      throw e;<br>    });<br>  }<br>  return musixmatchTokenPromise;<br>}<br><br>async function fetchMusixmatchLyrics(trackTitle: string, artistName: string): Promise&lt;string&gt; {<br>  const token = await getMusixmatchToken();<br>  const res = await fetch(`https://apic-desktop.musixmatch.com/ws/1.1/macro.subtitles.get?...`);<br>  const data = await res.json();<br><br>  // Self-healing check: if Musixmatch rejects the token, invalidate the cache<br>  const matcherStatus = data?.message?.body?.macro_calls?.[&quot;matcher.track.get&quot;]?.message?.header?.status_code;<br>  if (matcherStatus === 401) {<br>    musixmatchTokenPromise = null; // Force the next request to fetch a fresh token<br>    throw new Error(&quot;Musixmatch token rejected (401)&quot;);<br>  }<br><br>  return data?.message?.body?.macro_calls?.[&quot;track.subtitles.get&quot;]?.message?.body?.subtitle_list?.[0]?.subtitle?.subtitle_body;<br>}</pre><ul><li><strong>The 4-Tier Waterfall:</strong></li></ul><p>a. LRCLIB Exact Match (/api/get)</p><p>b. LRCLIB Fuzzy Search (/api/search)</p><p>c. Musixmatch Desktop API (macro.subtitles.get returning synced .lrc)</p><p>d. lyrics.ovh (Unsynced plain text fallback)</p><h4>2. Eliminating AI Misalignment with Structured JSON Schemas</h4><p>To ensure Gemini never scrambles line orders or drops timestamps during translation or fallback romanization, I strictly enforce Native JSON Schema outputs at the API level:</p><pre>const aiResponse = await ai.models.generateContent({<br>  model: &quot;gemini-2.5-flash&quot;,<br>  contents: prompt,<br>  config: {<br>    temperature: 0,<br>    responseMimeType: &quot;application/json&quot;,<br>    responseSchema: { type: &quot;ARRAY&quot;, items: { type: &quot;STRING&quot; } },<br>  },<br>});</pre><p>By enforcing responseSchema, Gemini returns a strict string array matching the exact line length of the input. If the returned array length does not match the input line count 1:1, the output is discarded instantly, protecting the app from desynchronized lyrics.</p><h4>3. The Scalability Shield (Supabase Postgres Cache)</h4><p>Every successfully processed track is saved directly to Supabase (songs and lyrics_translations tables). When a user requests a song that has been played anywhere in the world, PolyTune bypasses all APIs and AI models entirely—streaming the pre-calculated payload directly from Postgres over SSE for $0.00.</p><h3>Key Takeaways for Edge Engineering</h3><ul><li><strong>Don’t use AI where deterministic code works:</strong> Offloading romanization from LLMs to local JS packages dropped the Edge processing times from seconds to milliseconds for standard scripts.</li><li><strong>Defend against CJS in Deno:</strong> Always use namespace imports (import * as Mod) and lazy runtime resolution when importing legacy npm packages into edge runtimes.</li><li><strong>Enforce JSON Schemas on LLMs:</strong> Never rely on raw string splitting or custom delimiters for synced data. Force the model to output a strictly typed array and validate its length before serving it to the client.</li><li><strong>Zero-Hallucination Prompt Engineering:</strong> By constraining models with a native JSON Schema array (responseSchema) and strict role boundaries, the translation pipeline forces deterministic, 1:1 mapped outputs that eliminate temporal misalignment and AI commentary.</li></ul><blockquote><strong>Try it Yourself👇</strong></blockquote><p><a href="https://github.com/Fash-Mayor/polytune_app">GitHub - Fash-Mayor/polytune_app: PolyTune is a native Android application built with React Native and Expo that provides real-time, synced, and translated lyrics for whatever is currently playing on your device.</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/607/1*3tutjPVYUlBrYpcM6KBy2A.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/609/1*ThQMs5qALLYE2fz3PMK-XA.jpeg" /><figcaption>PolyTune screenshot</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a50aa5e7cb67" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/i-built-a-lyrics-translation-and-romanization-app-a50aa5e7cb67">I Built A Lyrics Translation and Romanization App</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 Part VI]]></title>
            <link>https://medium.com/@mayowafasugba/6in6-part-vi-95a3819b6fe5?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/95a3819b6fe5</guid>
            <category><![CDATA[application-development]]></category>
            <category><![CDATA[movies]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[self-improvement]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Tue, 04 Aug 2026 10:38:17 GMT</pubDate>
            <atom:updated>2026-08-04T10:38:17.020Z</atom:updated>
            <content:encoded><![CDATA[<p>Building WatchVault: How I Conquered Native Modules, MMKV, and Dynamic UI for My 6in6 Finale 🚀</p><p>If you’ve ever spent thirty minutes scrolling through Netflix, Prime Video, and a dozen other apps just trying to decide what to watch — only to give up and watch <em>The Office</em> for the fourth time — you know the pain. That exact “endless streaming scroll” was the itch I needed to scratch for the grand finale of my 6in6 challenge.</p><p>Enter <strong>WatchVault</strong>: a lightning-fast, local-first media discovery and playlist companion built from scratch using React Native. It cures decision paralysis by letting users browse, search, and curate movies, TV shows, anime, and K-dramas, complete with regional streaming availability for Nigeria right out of the box.</p><p>Building a production-ready mobile app designed to feel buttery-smooth wasn’t a walk in the park. Here is a breakdown of the technical hurdles I faced, the architecture choices I made, and how I brought WatchVault from my mind to my phone.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/1*cc3TjOMMtahlDtbLPp7Kug.jpeg" /><figcaption>WatchVault</figcaption></figure><h3>🛠️ The Tech Stack &amp; Architecture</h3><p>To keep WatchVault hyper-fast and scalable, I skipped heavy cloud backends and built a <strong>local-first architecture</strong>:</p><ul><li><strong>Framework:</strong> React Native (Expo)</li><li><strong>Styling:</strong> NativeWind v4 (Tailwind CSS for React Native)</li><li><strong>State &amp; Storage:</strong> Zustand paired with react-native-mmkv</li><li><strong>Data Sources:</strong> TMDB API + JustWatch regional streaming integration (defaulting to Nigeria NG)</li><li><strong>Native Magic:</strong> react-native-image-colours for native thread colour extraction</li></ul><h3>🧗‍♂️ The Major Hurdles &amp; How I Solved Them</h3><h3>1. Stepping Beyond Expo Go: The Native Module Wall</h3><p>Early on, I hit the classic React Native roadblock: standard Expo Go couldn’t handle the raw performance and custom native code required for our storage and color-extraction engines.</p><ul><li><strong>The Fix:</strong> I migrated to a <strong>Custom Dev Client</strong> workflow using EAS Build. This allowed us to leverage blindingly fast native modules while keeping the developer experience clean and rapid.</li></ul><h3>2. Blazing Speed with MMKV &amp; Zustand Hydration</h3><p>Traditional storage solutions like AsyncStorage are asynchronous and can cause micro-stutters or empty UI flashes on boot.</p><ul><li><strong>The Fix:</strong> I integrated react-native-mmkv, which executes synchronous C++ storage operations under the hood. To make it bulletproof in React, I built a custom Zustand storage bridge combined with an onRehydrateStorage hydration listener. The app knows the <em>exact</em> millisecond local playlists are loaded, showing a loading state before rendering instantly with zero layout shift.</li></ul><h3>3. The “Alive” Dynamic UI Challenge</h3><p>I didn’t want WatchVault to look like every other dark-mode clone out there. I wanted it to feel <em>alive</em>. But calculating a dominant color palette from a remote image without tanking the UI thread is tricky.</p><ul><li><strong>The Fix:</strong> I utilized react-native-image-colors inside a dedicated detail modal (AliveModal). When a user taps a movie poster, the native layer extracts the dominant and vibrant hex codes on the fly, instantly shifting the modal’s background gradient and aesthetic to match the movie&#39;s unique vibe—whether it’s the neon blues of an anime or the gritty crimsons of a thriller.</li></ul><h3>4. Smart Regional Streaming &amp; Infinite Pagination</h3><p>Hardcoding movies is easy; building a living, breathing discovery engine that never runs out of content is hard.</p><ul><li><strong>The Fix:</strong> I wired up TMDB’s multi-genre endpoints alongside its native JustWatch regional provider mapping. By appending dynamic country parameters (NG), WatchVault automatically overlays precise platform mini-badges (Netflix, Prime, Showmax) onto poster cards. Combined with optimized FlatList pagination hooks, users can scroll infinitely through trending hits, sci-fi epics, and top-tier anime without ever hitting a dead end.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/1*EkeMGmkGuNo-eZl4Auwm4g.jpeg" /><figcaption>WatchVault</figcaption></figure><h3>🏁 Crossing the Finish Line</h3><p>Building WatchVault pushed me to dive deep into native-side performance optimization, state hydration edge cases, and modern React Native architecture. Turning an idea into a fully responsive, beautiful, high-performance application right at the end of an intense development sprint proves what’s possible when you keep building and shipping.</p><p>This concludes my 6-problem-in-6-weeks challenge. Can’t wait to see what I do next!🚀</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=95a3819b6fe5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 Part V]]></title>
            <link>https://medium.com/@mayowafasugba/6in6-part-v-84ae9ce9760a?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/84ae9ce9760a</guid>
            <category><![CDATA[reactjs]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[devtools]]></category>
            <category><![CDATA[careers]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Mon, 27 Jul 2026 11:11:52 GMT</pubDate>
            <atom:updated>2026-07-27T11:11:52.478Z</atom:updated>
            <content:encoded><![CDATA[<p>Simulating 10,000 Users on Localhost with React &amp; Web Workers</p><p>I’ve officially hit <strong>Week 5 of my 6in6 challenge</strong> (building 6 projects in 6 weeks), and I wanted to tackle something that used to terrify me: <strong>high concurrency</strong>.</p><p>When you’re building apps to eventually scale to millions of users, you have to answer a scary question: <em>What happens when thousands of people click “Book Now” at the same millisecond?</em></p><p>To figure this out and level up my ReactJS skills, I built <strong>BIT (Booking Interface Terminal)</strong> — a dark-mode VIP slot booking dashboard equipped with a custom <strong>“Chaos Mode”</strong> that floods the app with simulated traffic in real time.</p><p>Here’s how I built it, the technical flexes I learned along the way, and what it taught me about modern React development.</p><h3>The Problem: How Do You Test Scale When You’re the Only User?</h3><p>It’s easy to build a clean UI that works when you’re the only person on localhost:3000. But real-world traffic is chaotic. I wanted to see how a frontend handles a massive flood of data without freezing the screen, lagging the animations, or crashing the browser.</p><p>Since I didn’t have an army of 10,000 real beta testers sitting in my room, I had to build a system that would attack my own frontend on command.</p><p>Enter <strong>Chaos Mode</strong>.</p><h3>The Cool Flexes (What I Learned &amp; Built)</h3><p>I’m not a senior cloud architect — I’m a developer using these weekly sprints to master React inside out. Pushing past standard useState hooks to build this opened up a whole new world of browser capabilities.</p><h3>1. Offloading Heavy Traffic to Web Workers</h3><p>If you run a loop generating thousands of fake server requests directly inside a standard React component, your UI will instantly freeze. The browser thread gets choked, and your buttery-smooth 60 FPS drops to zero.</p><p>To fix this, I used <strong>Web Workers</strong>. Web Workers let you run JavaScript on a completely separate background thread. When I flip the “Chaos Mode” switch, the worker thread spins up and slams the system with operations, leaving the main React thread 100% free to render UI and handle button clicks effortlessly.</p><p>Normally, Web Workers require you to maintain a separate .js file in your public/ directory. But to keep everything self-contained within our React component architecture, we can turn a raw string of JavaScript into an in-memory worker using <strong>Blob</strong> and <strong>URL.createObjectURL()</strong>:</p><pre>import { useEffect, useRef } from &#39;react&#39;;<br><br>export function useInMemWorker() {<br>  const workerRef = useRef(null);<br><br>  useEffect(() =&gt; {<br>    // 1. Define the worker logic directly as a multiline string<br>    const workerCode = `<br>      self.onmessage = (e) =&gt; {<br>        if (e.data.command === &#39;START&#39;) {<br>          // Run background CPU heavy task on a separate thread<br>          const rps = Math.floor(Math.random() * 5000) + 1000;<br>          self.postMessage({ rps });<br>        }<br>      };<br>    `;<br><br>    // 2. Convert string to a JS Blob and create a temporary browser URL<br>    const blob = new Blob([workerCode], { type: &#39;application/javascript&#39; });<br>    const workerUrl = URL.createObjectURL(blob);<br><br>    // 3. Instantiate the Web Worker directly in React<br>    const worker = new Worker(workerUrl);<br>    <br>    worker.onmessage = (e) =&gt; {<br>      console.log(&#39;Telemetry received from background thread:&#39;, e.data);<br>    };<br><br>    workerRef.current = worker;<br><br>    // 4. Clean up memory when component unmounts<br>    return () =&gt; {<br>      worker.terminate();<br>      URL.revokeObjectURL(workerUrl);<br>    };<br>  }, []);<br><br>  return workerRef;<br>}</pre><h3>2. Building a “Zero-Lag” IndexedDB Buffer Pipeline</h3><p>Sending thousands of messages per second from a background worker straight into React state can still overload the UI with re-renders.</p><p>Instead of updating React state on every single request, I built an <strong>IndexedDB pipeline</strong>. The Web Worker dumps high-speed traffic logs directly into the browser’s IndexedDB storage buffer. Then, a custom React hook periodically polls that buffer to update the charts. It felt like building a mini streaming pipeline (like Kafka or Redis) right inside the browser!</p><h3>3. Real-Time Telemetry with Tremor Charts</h3><p>To make the traffic spike visually thrilling, I integrated <strong>Tremor</strong> with Tailwind CSS. The moment Chaos Mode is activated, a real-time area chart spikes wildly, showing live requests per second (req/s) while a status badge flips from a calm green INITIALIZE CHAOS MODEto a glowing red DISABLE CHAOS MODE.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RYSPDcPsJCvbZnI0Gvsr0Q.png" /><figcaption>BIT interface (Active Chaos Mode)</figcaption></figure><h3>Key Lessons from Week 5</h3><ul><li><strong>React State Isn’t for Everything:</strong> Not every piece of rapid data belongs in a useState variable. Learning when to keep high-frequency math <em>out</em> of React’s render loop was a game-changer for UI performance.</li><li><strong>Asynchronous UX Matters:</strong> When simulating database row locks under heavy fire, giving the user instant visual feedback (like optimistic loading states and pulsing status badges) makes the app feel resilient and fast.</li><li><strong>Building Visual Dev Tools Is Fun:</strong> Adding “dev panels” or “chaos toggles” directly into your projects doesn’t just make for great portfolio demos — it actually helps you understand your own code’s breaking points.</li></ul><h3>One Week Left!</h3><p>BIT went from a blank canvas to a multi-threaded, high-concurrency simulation dashboard in just a few days. I’m now heading into the final stretch — <strong>Week 6</strong>!</p><p>Thanks for following along on this 6in6 journey. Finishing strong! 🚀</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=84ae9ce9760a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 part IV]]></title>
            <link>https://medium.com/@mayowafasugba/6in6-part-iv-d89a575b54c9?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/d89a575b54c9</guid>
            <category><![CDATA[code-snippet]]></category>
            <category><![CDATA[nextjs]]></category>
            <category><![CDATA[web-development]]></category>
            <category><![CDATA[ui]]></category>
            <category><![CDATA[developer]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Mon, 20 Jul 2026 09:43:47 GMT</pubDate>
            <atom:updated>2026-07-20T09:44:08.709Z</atom:updated>
            <content:encoded><![CDATA[<p>From Deadline Panic to Shipping SnapText in 2 Hours</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qFHA6AyuJ5DvST3YyUZzow.png" /><figcaption>SnapText</figcaption></figure><p>It wouldn’t be a true 6-week build challenge without at least one emergency late-night sprint.</p><p>I was supposed to drop Week 3 a few days ago, but the publication holding my draft moved too slow, and I waited until week 4 submission (which was a terrible plan). But right now, I’m pulling a Beyoncé. Welcome to a 6in6 Double Feature: <a href="https://medium.com/@mayowafasugba/6in6-part-iii-11dde1bc1d2b">SpotyDrop </a>AND SnapText dropping on the exact same day.</p><p>With the Week 4 submission deadline looming and no active project lined up, I had two options: break my <strong>6in6 streak</strong> or build and ship a complete, high-utility micro-app before the clock ran out.</p><p>Naturally, I chose the sprint.</p><p>Meet <strong>SnapText 📸</strong> — a lightweight, type-safe code and log visualizer that turns raw code snippets, terminal errors, and developer notes into crisp, high-resolution graphic cards ready for social media.</p><h3>The “Itch”: Why SnapText?</h3><p>Every developer loves sharing milestones, code snippets, or hilarious terminal logs on Twitter/X and LinkedIn. But raw IDE screenshots usually look clunky, unformatted, or poorly cropped. Existing visualizer tools often feel bloated or require unnecessary sign-ups just to export a simple image.</p><p>I wanted something dead simple, blazing fast, and entirely client-side:</p><ol><li>Paste text/code.</li><li>Tweak the layout, font, and background gradient.</li><li>Export a pristine PNG in one click.</li></ol><h3>The Tech Stack</h3><p>To keep execution fast and prevent deployment bottlenecks, I built SnapText entirely on the client side:</p><ul><li><strong>Framework:</strong> Next.js (App Router) + TypeScript (.tsx)</li><li><strong>Styling:</strong> Tailwind CSS (featuring custom radial gradients and backdrop blur controls)</li><li><strong>Canvas Renderer:</strong> html-to-image for DOM-to-PNG generation</li><li><strong>Typography &amp; UI:</strong> Custom font family bindings and dynamic padding states</li></ul><h3>What I Built (Features Packed into a 2-Hour Sprint)</h3><p>Even under extreme time pressure, I didn’t want to ship a barebones UI (like <a href="https://medium.com/stackademic/6in6-part-ii-3bf4d2dad0c7">ZoneDo</a>). I built SnapText with full type safety and modular styling controls:</p><ul><li>🎨 <strong>Vibrant Background Canvas:</strong> Pre-configured radial and linear gradient themes (Aurora, Midnight Ocean, Retro Sunset) that make code pop.</li><li>🔤 <strong>Type-Safe Font Switcher:</strong> On-the-fly toggling between <em>Default Mono</em>, <em>Modern Sans</em>, and <em>Classic Serif</em> typography.</li><li>📐 <strong>Dynamic Padding Customizer:</strong> Adjustable canvas padding (<em>Compact</em>, <em>Standard</em>, <em>Cinematic</em>) to perfectly balance aspect ratios for different platforms.</li><li>💻 <strong>macOS-Style Window Controls:</strong> A stylized code container complete with classic window buttons and dark mode contrast.</li><li>⚡ <strong>High-DPI PNG Export:</strong> Utilizes a 2x pixel ratio configuration during image generation so exports remain crisp on retina displays without layout shifting.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WfSK0sQMOwZFNkPsa2TI_w.png" /><figcaption>SnapText code snippet image</figcaption></figure><h3>Engineering Takeaways</h3><p>Building under pressure forces you to simplify your architecture and make smart trade-offs:</p><ol><li><strong>Client-Side Power Over Backend Bloat:</strong> Not every app needs a database or server-side routing. Running everything inside browser state meant zero network latency, zero API rate limits, and an instant Vercel deployment.</li><li><strong>Strict TypeScript Saves Time:</strong> When rushing, it’s easy to introduce scope bugs or undefined state errors. Enforcing strict interfaces (GradientOption, FontOption, PaddingOption) and typing the useRef&lt;HTMLDivElement&gt; target caught potential compile errors in the editor before they could break the Vercel build.</li><li><strong>Utility as Side-Project Marketing:</strong> Building small, focused tools that solve real developer headaches is one of the most effective ways to build a personal brand and drive organic traffic.</li></ol><h3>Week 4 Down, 2 to Go!</h3><p>SnapText went from a blank terminal to a deployed product in under two hours, keeping the 6in6 momentum alive and well.</p><p>I’m officially past the halfway mark! See you in Week 5! 🚀</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d89a575b54c9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 Part III]]></title>
            <link>https://medium.com/@mayowafasugba/6in6-part-iii-11dde1bc1d2b?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/11dde1bc1d2b</guid>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[system-architecture]]></category>
            <category><![CDATA[spotify]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Mon, 20 Jul 2026 09:30:27 GMT</pubDate>
            <atom:updated>2026-07-20T09:30:27.943Z</atom:updated>
            <content:encoded><![CDATA[<p>How I Turned My PC into a Cloud Audio Server for SpotyDrop</p><p>We are officially halfway through the 6in6 challenge! For Week 3, I wanted to build something highly personal that scratched my own itch. I had an old Windows batch script that use Python (spotdl) to download Spotify tracks to my laptop. It worked great, but I wanted those songs on my <em>phone</em>, and I didn&#39;t want to plug in a USB cable like it&#39;s 2015, and obviously mobile phones don’t speak .bat.</p><p>The goal for Week 3? Build <strong>SpotyDrop</strong> — a React Native mobile app that lets me paste a Spotify link, beam it to my laptop over the internet, have the laptop download and convert the audio, and then stream the MP3 straight back to my phone’s local storage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*3UDN3bkkKN6K5yH4idvlsQ.jpeg" /><figcaption>SpotyDrop interface.</figcaption></figure><p>Here is how I built it, and the absolute chaos of making a phone talk to a PC.</p><h3>The Tech Stack</h3><p>To bridge the mobile-to-desktop gap, I had to build a full-stack pipeline:</p><ul><li><strong>The Front-End (Mobile):</strong> React Native and Expo (with expo-file-system and expo-sharing).</li><li><strong>The Back-End (PC Brain):</strong> Node.js and Express to listen for mobile requests.</li><li><strong>The Engine:</strong> Python and spotdl running via hidden child processes on the PC.</li><li><strong>The Bridge:</strong> Ngrok to create a secure tunnel through my Wi-Fi router.</li></ul><h3>The Developer Reality: Hurdles &amp; Boss Fights</h3><p>Building the happy path is easy. Making it survive the real world is where the actual engineering happens. Here are the four major bosses I had to defeat this week:</p><p><strong>1. The Expo Red Screen of Death (Scope Issues):</strong> Right when I thought the UI was ready, my phone hit me with the dreaded React Native red screen: [runtime not ready]: ReferenceError: Property &#39;url&#39; doesn&#39;t exist.</p><ul><li><em>The Fix:</em> A classic JavaScript scope mistake! I was trying to dynamically append my React state (url) to the Ngrok backend string at the very top of my file. But because url was a React hook living <em>inside</em> my component, the global file crashed because url didn&#39;t exist yet. Moving the URL construction inside my download function fixed it instantly.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*kSYOpqe3imfENiMwhIr_Cw.jpeg" /><figcaption>ReferenceError screen.</figcaption></figure><p><strong>2. The Fragility of Web Scrapers:</strong> Right out of the gate, my underlying Python engine (spotdl) crashed with a massive KeyError: &#39;header&#39; traceback. It turns out YouTube had changed their website structure, completely breaking the API scraper.</p><ul><li><em>The Fix:</em> A harsh reminder that when relying on third-party scrapers, you are at the mercy of the host. Forcing a global pip upgrade on my PC pulled down the community patch and got the engine breathing again.</li></ul><p><strong>3. The Ngrok “Bad Gateway”:</strong> To bypass local Wi-Fi restrictions, I used Ngrok to expose my local Node server to the internet. But when my phone tried to connect, I got a 502 Bad Gateway and an ERR_NGROK_8012 error.</p><ul><li><em>The Fix:</em> Windows networking is tricky. Node was listening on an IPv4 address (0.0.0.0), but Ngrok was trying to route traffic to the IPv6 version of localhost ([::1]). Explicitly telling Ngrok to target 127.0.0.1 fixed the routing instantly.</li></ul><p><strong>4. The Final Boss: The HTTP Timeout (</strong><strong>ECONNABORTED):</strong> This was the hardest bug of the week. My app would successfully reach the PC, the PC would download the song, but right before the file was sent to the phone, the connection would randomly die. Why? Because audio conversion takes time (sometimes up to a minute). The HTTP connection between the phone and the PC was sitting completely silent, so the network assumed the connection was dead and severed the wire.</p><ul><li><em>The Fix:</em> I had to completely rewrite the architecture to use an <strong>Asynchronous Polling Handshake</strong>. Instead of waiting on one long connection, my phone now sends a “Trigger” request and gets a tracking ID. The phone then pings a “Status” route every 3 seconds (<em>“Are you done yet?”</em>). Once the PC says “Ready!”, the phone instantly grabs the file in one fast, uninterrupted stream.</li></ul><h3>What I Achieved</h3><p>As of today, SpotyDrop is fully operational. I can paste a link on my Android phone, watch my PC terminal light up in the background as it processes the audio, and a few seconds later, the native Android Share menu pops up on my phone with a fresh MP3 ready to be saved.</p><p>It’s a massive step up from Week 2. I moved from local frontend UI into actual systems architecture, managing cross-device networking, file streaming, and background process execution.</p><p>Three weeks down. Three to go. The momentum is crazy — see you in Week 4!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*LilwvdtxXfHdqgKLBV35Bg.png" /><figcaption>my PC terminal processing requests.</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=11dde1bc1d2b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 Part II]]></title>
            <link>https://blog.stackademic.com/6in6-part-ii-3bf4d2dad0c7?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/3bf4d2dad0c7</guid>
            <category><![CDATA[android-app-development]]></category>
            <category><![CDATA[react-native]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[ios-app-development]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Mon, 06 Jul 2026 09:28:40 GMT</pubDate>
            <atom:updated>2026-07-06T09:28:40.074Z</atom:updated>
            <content:encoded><![CDATA[<p>Shipping ZoneDo: Geo-Fenced To-Do Lists and the Reality of Background Location Tracking.</p><p>Welcome to Week 2 of my 6in6 challenge! Week 1 was about getting my feet wet, Week 2 was about diving deep into mobile development. My goal this week was to build an app that solves a very specific, annoying problem: forgetting to do things when you are actually in the right place to do them.</p><p>Enter <strong>ZoneDo</strong>.</p><h3>The Vision: A Geo-Fenced Brain</h3><p>Have you ever driven past a Market or a mall, gotten all the way home, and suddenly realized you forgot to buy that HDMI cord? ZoneDo is built to stop exactly that.</p><p>The premise is simple: You type in what you want to do, and you set the “Zone” where it needs to happen. ZoneDo runs quietly in the background, listening to your location. The second you enter that designated area, the app pings your phone with a push notification: <em>“Things You Want to do at Yaba Market.”</em></p><p>It is a location-aware extension of your brain.</p><h3>The Tech Stack</h3><p>To build this fast and reliably, I went with heavy hitters in the mobile space:</p><ul><li><strong>React Native &amp; Expo:</strong> For rapid development and cross-platform capabilities.</li><li><strong>TypeScript:</strong> Because catching errors before they compile is a lifesaver.</li><li><strong>NativeWind:</strong> To bring the speed of Tailwind CSS to mobile UI.</li><li><strong>Expo Location &amp; Task Manager:</strong> The core engines running the background geo-fencing.</li></ul><h3>The Saturday Struggle: The Developer Reality</h3><p>You see the app, but let me tell you about the messy in-betweens. Building background location tracking isn’t just a logic problem; it’s a configuration battlefield because operating systems (iOS and Android) aggressively kill background apps to save battery. I had to navigate these strict OS-level restrictions to hit my deadline. Hitting my MVP deadline required battling a series of notorious React Native “final bosses”:</p><ol><li><strong>The </strong><strong>DOMException Polyfill Trap</strong> When setting up background tasks, my app suddenly crashed, demanding web APIs that don&#39;t exist on a mobile phone! I had to explicitly write a global polyfill (global.DOMException = class DOMException extends Error {};) and trick TypeScript into accepting it just to get the location tracker breathing again.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*VkqYcK0sWlNaYDJmQoo_dw.jpeg" /><figcaption>DOMException Error</figcaption></figure><p><strong>2. The Windows </strong><strong>ERR_UNSUPPORTED_ESM_URL_SCHEME Nightmare</strong> Just as things were compiling, modern Node.js and Windows file paths went to war. Node expected web URLs (file://), but Windows gave it standard hard drive paths (C:\). This completely broke the Metro bundler. The fix? I had to move my entire project out of a cloud-synced directory to my local root directory (C:\zonedo\), rip out strict ES modules from my package.json, and rebuild the dependency tree from scratch.</p><h3>The Async Pivot: Knowing When to Pause</h3><p>As of writing this, the app is alive. It is running on my physical phone, and the background location pinger is fully functional — the core MVP is achieved!</p><p>But the styling? My NativeWind CSS is currently taking a nap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*UIvGawXv-6NH3BbZ7p5sbQ.jpeg" /><figcaption>Ugliest MVP you and I have ever seen.</figcaption></figure><p>Part of the 6in6 challenge is knowing when to keep fighting a configuration bug, and when to step back, work asynchronously, and document the win. The logic works. The tracking works. I’m shipping this update now, and I’ll tackle the visual polish with a fresh set of eyes tomorrow.</p><p>Shipping is about momentum, and right now, the momentum is pushing forward. See you in Week 3!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3bf4d2dad0c7" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/6in6-part-ii-3bf4d2dad0c7">6in6 Part II</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[6in6 Part I]]></title>
            <link>https://blog.stackademic.com/6in6-part-i-72199925c417?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/72199925c417</guid>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[personal-growth]]></category>
            <category><![CDATA[front-end-development]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[nextjs]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Sat, 27 Jun 2026 09:46:11 GMT</pubDate>
            <atom:updated>2026-06-27T09:46:11.026Z</atom:updated>
            <content:encoded><![CDATA[<p>Building a Local-First Challenge Tracker with Next.js and Dexie.js</p><h3>The Irony of Problem 1</h3><p>When I launched my <strong>6 Problems in 6 Weeks</strong> challenge, I needed a way to track my problems, solutions, progress, deadlines, store notes and to-dos, and keep my thoughts organised.</p><p>I could have used Notion. I could have used Trello. But why use an existing tool when I can build a custom one from scratch?</p><p>For Week 1, I built <strong>6in6 Personal Blog Tracker</strong> — a sleek, offline-first dashboard designed specifically to manage this 6-week challenge.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/487/1*nBoomN7Fb4iyR3IlRczJgQ.png" /><figcaption>Interface of 6in6 blog tracker</figcaption></figure><h3>What the Tracker Does</h3><p>The app is built using <strong>Next.js</strong> and handles everything I need to stay accountable:</p><ul><li><strong>Weekly Tabs:</strong> A dynamic navigation system to switch between weeks and view project parameters.</li><li><strong>Tracking:</strong> Real-time visibility into project status (e.g., “Ongoing”), start dates, and deadlines.</li><li><strong>Interactive Checklists:</strong> A custom To-Do section to track tasks for the current solution.</li><li><strong>Attachments &amp; Notes:</strong> The ability to store project notes and even directly embed images.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/467/1*NfIK4KkdHfhr9NAQGHYY2A.png" /><figcaption>Interface of 6in6 blog tracker</figcaption></figure><h3>Discovering Dexie.js</h3><p>Because this is a personal execution tool, spinning up an SQL database, managing cloud hosting, or dealing with network latency every time I checked off a task didn’t make sense. I wanted this app to be incredibly fast, private, and functional even without internet.</p><p>This led me to explore <strong>IndexedDB</strong> — the browser’s built-in database.</p><p>IndexedDB is low-level and stressful to configure. To bypass the boilerplate, I used <strong>Dexie.js</strong>, a minimalist wrapper that provides a clean, promise-based API for browser storage.</p><p>Here is why Dexie completely changed how I think about frontend state management for local tools:</p><h4>1. Zero-Friction Schema Definition</h4><p>Setting up the database takes minutes. You define your database instance and declare your stores and keys without complex migration files:</p><pre>import Dexie, { type Table } from &#39;dexie&#39;;<br><br>export interface Project {<br>  id?: number;<br>  title: string;<br>  stage: string;<br>  Deadline: string;<br>  createdAt: number;<br>  description: string;<br>  todos: string[];<br>  imageBlob: Blob | null;<br>  notes: string;<br>  mediumLink: string;<br>}<br><br>class ProjectDatabase extends Dexie {<br>  projects!: Table&lt;Project&gt;;<br><br>  constructor() {<br>    super(&#39;ProjectDatabase&#39;);<br>    this.version(1).stores({<br>      projects: &#39;++id, title, createdAt&#39;<br>    });<br>  }<br>}<br><br>export const db = new ProjectDatabase();</pre><h4>2. Real-Time React Reactivity</h4><p>Dexie integrates very well with React hooks via useLiveQuery. Whenever the database updates, the UI automatically re-renders without needing to manage complex global state arrays or manual refetching triggers. It feels just as snappy as local component state, but persists forever across browser refreshes.</p><h3>Frontend W’s &amp; Lessons Learned</h3><p>Beyond the database layer, this project forced me to focus heavily on UI scannability and structural layout:</p><ul><li><strong>Handling Local State vs. Persistent State:</strong> Figuring out when to write directly to Dexie versus holding inputs in temporary component state (like typing out a long description before saving) was a great exercise in performance optimisation.</li><li><strong>Component Architecture:</strong> Breaking down the dashboard into isolated, reusable blocks — the tab selector, the countdown metrics, the task runner, and the asset viewer.</li></ul><h3>What’s Next?</h3><p>Week 1 is in the books. The tracker is live on my machine, the data is persisting in IndexedDB, and the deadline for the next problem is looming.</p><p>Now that I have the tool to track my journey, it’s time to start building something for Week 2.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=72199925c417" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/6in6-part-i-72199925c417">6in6 Part I</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The “Planning Trap,” A Broken Chrome Extension, and My 6-Week Comeback Challenge]]></title>
            <link>https://blog.stackademic.com/the-planning-trap-a-broken-chrome-extension-and-my-6-week-comeback-challenge-bfdd5c0d241c?source=rss-e92172a5665b------2</link>
            <guid isPermaLink="false">https://medium.com/p/bfdd5c0d241c</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[personal-growth]]></category>
            <category><![CDATA[inspiration]]></category>
            <category><![CDATA[careers]]></category>
            <dc:creator><![CDATA[Mayowa Fasugba]]></dc:creator>
            <pubDate>Fri, 19 Jun 2026 10:54:15 GMT</pubDate>
            <atom:updated>2026-06-19T10:54:15.265Z</atom:updated>
            <content:encoded><![CDATA[<p>When planning becomes procrastination, the only cure is shipping imperfect code (and fighting unstable internet).</p><p>Between surviving NYSC orientation camp and trying to recalibrate my life, I had to press pause on a lot of things — including writing here on Medium. But my hiatus ends today.</p><p>Looking back at the last few months, I realized I fell into a classic developer trap of over-refining.</p><p>Instead of actually building, I’ll spend weeks tweaking my learning roadmaps, adjusting schedules, and trying to chart the “perfect” path to becoming the type of developer I want to be. Planning feels like productivity, but it’s sometimes just a form of procrastination.</p><p>So, I decided to stop planning and start breaking things again.</p><h3>The “Auto-Resume” Experiment</h3><p>To break the dry spell, I decided to scratch a personal itch.</p><p>If you know anything about internet stability in Nigeria, you know the pain of interrupted downloads on Chrome. A sudden network lag, a temporary server drop, or some mysterious cosmic error, and <em>boom</em> — your 2GB download is dead in the water.</p><p>I got tired of babysitting my browser, so I wrote a script/extension that monitors my active downloads. The moment a download gets interrupted, the script automatically triggers a resume command.</p><p>Does it work?<strong> Yes.</strong><br>Is it perfect? <strong>Absolutely not.</strong></p><p><strong>The Current Flaws:</strong></p><ol><li><strong>It’s a bit <em>too</em> enthusiastic:</strong> If I explicitly pause a download because I need the bandwidth, the script assumes I’m wrong and forcefully resumes it anyway.</li><li><strong>The UI icons look incredibly bad:</strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gA6E4fxIcYE12b65Mo1kkQ.png" /><figcaption>screenshot of the extention in my browser</figcaption></figure><p>It’s messy, but it solved a real problem, and more importantly, it reminded me how much better it feels to ship imperfect code than to stare at a perfect roadmap.</p><h3>The Next Chapter: 6 Problems, 6 Weeks</h3><p>Learning by doing is the only method that sticks. To force myself out of the planning phase and master React.js, Next.js, and React Native, I am launching a self-imposed challenge:</p><p>🔥 6 Problems in 6 Weeks.</p><p>Every week for the next month and a half, I will identify a personal or real-world problem, build a production-ready application to solve it, and document the entire architectural process right here. No tutorials, no hand-holding — just pure, problem-first engineering.</p><p>If you want to see the methodologies I use, the architectural bottlenecks I run into, and the inevitable lessons learned from breaking things, hit follow.</p><p>Week 1 starts now. Let’s build.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bfdd5c0d241c" width="1" height="1" alt=""><hr><p><a href="https://blog.stackademic.com/the-planning-trap-a-broken-chrome-extension-and-my-6-week-comeback-challenge-bfdd5c0d241c">The “Planning Trap,” A Broken Chrome Extension, and My 6-Week Comeback Challenge</a> was originally published in <a href="https://blog.stackademic.com/">Stackademic</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>