<?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 Derby on Medium]]></title>
        <description><![CDATA[Stories by Derby on Medium]]></description>
        <link>https://medium.com/@derbyps?source=rss-94ccfc5c719b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*TOYUNEMf1hPNGtYg.jpg</url>
            <title>Stories by Derby on Medium</title>
            <link>https://medium.com/@derbyps?source=rss-94ccfc5c719b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 08:09:41 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@derbyps/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[The Vibe Coding Illusion: Is It Your Prompt, or Just a Cheap Model?]]></title>
            <link>https://derbyps.medium.com/the-vibe-coding-illusion-is-it-your-prompt-or-just-a-cheap-model-cbe142aa7d9c?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbe142aa7d9c</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[learning]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Fri, 24 Jul 2026 10:50:38 GMT</pubDate>
            <atom:updated>2026-07-24T10:50:38.331Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>“If you automate a mess, you get an automated mess.”<br> — <strong>Rod Michael</strong>, IT Executive</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U3r70STv-9ct-VL1paemQg.png" /><figcaption>Credit goes to Gemini for generating this image.</figcaption></figure><p>For a long time, I was a big skeptic of “vibe coding” — the practice of steering an AI to write your software through natural language.</p><p>Sure, I have used AI autocomplete tools like Copilot to help finish my lines, but the idea of relying on an AI to generate entire features from scratch just didn’t appeal to me. The tools felt pricey, and the generated code rarely hit the mark on the first try. Whenever I complained about the disappointing results, the common advice in the community was always the same: <em>“You have to write clean, highly structured prompts,”</em> or, <em>“Treat the AI like a junior developer; give it exact, step-by-step instructions.”</em></p><p>While that advice made sense, it felt entirely counterproductive. If I had to spend twenty minutes writing a detailed prompt to explain the exact logic, architecture, and edge cases, I could have just written the code myself in half the time. If they couldn’t handle abstract instructions, were they really saving me time, or just <strong>shifting my cognitive load from syntax to semantics?</strong></p><p>I was also worried about security. Giving a third-party tool deep access to a proprietary codebase felt risky. Because of that, I had the perfect excuse to avoid dedicated AI coding environments. Instead, I was perfectly content keeping a browser tab open with a free, web-based AI. By manually sanitizing and copy-pasting small, isolated snippets, I stayed in control. It felt safe, my usage limits were never capped, and it was undeniably faster and cheaper than paying for a premium coding agent.</p><p>But our industry moves fast, and AI has grown up. Recently, my company allocated a budget specifically for integrating AI tools directly into our development workflow. With the financial barrier removed, I decided to drop my biases and give it a real shot.</p><p>What I discovered completely shifted my perspective. With the current state of AI, the output is finally <em>good enough</em> — provided you do the heavy lifting at the beginning. If you spend the time upfront setting up comprehensive documentation and feeding the AI your codebase conventions, the process becomes incredibly smooth and reusable.</p><p>During this deep dive, I also realized something important about how we judge these tools. We always hear that if the AI fails, it is because our prompt was bad. But I tried many different models, and the reality is that the more expensive the model, the better the code. When I used a cheaper model, the process was slow; I had to write extra prompts just to fix its constant mistakes. But when I pasted the exact same prompt into a top-tier model, the result was excellent.</p><p>So, is the problem really my prompt, or is it the model? I learned that finding the “sweet spot” — balancing the subscription cost against the quality of the result and the time saved fixing errors — is a crucial skill for modern development.</p><p>I am still not 100% vibe coding. Right now, my workflow is a hybrid. Sometimes I write the code myself first and ask the AI to review and fix it. Other times, the AI writes the base, and I jump in to refine it. Maybe my prompting ability still has room to grow, or maybe I just need to wait for even better models.</p><p>But regardless of who writes the first draft, my biggest takeaway is this: the more I use AI, the less my daily work resembles traditional software engineering, and the more it looks like project management.</p><p>To get high-quality output, I have to deeply study the product requirements. I spend my time writing robust instructions, defining clear acceptance criteria, and translating business needs into technical parameters. It is highly ironic because it directly contradicts my earlier complaint about hating the process of writing detailed requirements!</p><p>Yet, I now believe this is a good future. As AI handles more of the heavy lifting, the most critical skills for an engineer will be understanding the business and managing the project. The engineers who will thrive are the ones who can sit with a business team, gather abstract requirements, and translate them into a flawless set of instructions that an AI can seamlessly execute.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbe142aa7d9c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[On Meetings]]></title>
            <link>https://derbyps.medium.com/on-meetings-714ab22841ab?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/714ab22841ab</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[time-management]]></category>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[leadership]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Mon, 22 Jun 2026 03:57:13 GMT</pubDate>
            <atom:updated>2026-06-22T03:57:13.159Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>The best meetings don’t spend time getting everyone on the same page — they start with everyone already there</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r67_xU6wMhXf7h2OZio3mg.png" /><figcaption>Dreamt by human, drawn by Gemini.</figcaption></figure><p>I’ve been thinking about meetings lately. Just noticing how much time they take, and how differently teams seem to approach them. Some rely heavily on real-time discussion and walking through documents together, while others seem to prefer digesting things independently first. The contrast made me curious about what actually works.</p><p>Especially when you have a great culture of shared ownership, but everyone also needs large blocks of quiet time to actually build the product. It’s a tough balance: how do we keep everyone’s voice in the mix without spending more time talking about the work than doing it?</p><p>I started reading up on how highly effective engineering teams tackle this. It’s easy to assume that optimizing meetings is only necessary for massive companies, but the more I read, the more it seemed like it’s actually about building solid habits early on. If a communication style protects focus time and work beautifully for a tight-knit team, it sets a great foundation.</p><p>While I’m not a management expert, as an engineer learning to improve my own time management, a few concepts really stood out to me:</p><h3>The “Silent Start”</h3><p>One interesting approach I read about (popularized by <strong>Amazon</strong>) is replacing presentations with reading time. When a new feature or project is proposed, someone writes a structured document (like a PRD). But instead of the author walking everyone through it, the first 20 minutes of the meeting are spent in total silence while everyone reads it independently. It sounds a bit unusual at first, but the logic is incredibly solid. It ensures everyone gets the exact same deep context at their own pace. The remaining meeting time is spent strictly debating the nuances, solving problems, and making decisions, rather than just catching everyone up.</p><h3>Drafting First, Reviewing Later</h3><p>Another pattern I noticed (championed by engineering cultures at companies like <strong>Stripe</strong> and <strong>Basecamp</strong>) is how these teams separate the creation phase from the review phase through &quot;distributed drafting&quot;. Instead of using meeting time to build project outlines or figure out tech specs from scratch, the persons/teams owning that specific piece of work — not necessarily the team lead — drafts a proposal beforehand. What’s really interesting is that this doesn’t make the rest of the team passive; it actually deepens shared ownership because everyone gets a turn to hold the pen. Then, the actual meeting becomes a rapid, high-energy review: “Does this flow make sense? Are there any missing edge cases?” As a bonus? reviewing documents beforehand gives a quieter team members an equal voice, since leaving a thoughtful comment is often easier than trying to jump into a fast-paced room discussion.</p><h3>Async First</h3><p>​This ties into the &quot;async-first&quot; mindset (a core philosophy at fully distributed companies like <strong>Gitlab</strong>). The golden rule here is that meetings are reserved for making decisions, not information sharing. If a discussion is just about updating the team on a new flow or sharing requirements, they don’t book a room. The approach is straightforward: if the goal can be accomplished with an email, a quick message thread, or a few comments on a document, it doesn’t get a calendar invite. They let everyone review the document on their own time between tasks. The only gather synchronously if there’s a debate that needs settling or a roadblock that requires a whiteboard. It completely respects everyone’s flow state.</p><h3>The Takeaway</h3><p>Looking at all these ideas, a natural question comes to my mind: Doesn’t this actually slow us down? I know a lot of people prefer just grabbing a room to discuss things because talking feels so much faster than waiting around for someone to reply to a comment thread. It’s a valid point. But it made me realize that there is a difference between &quot;collaboration&quot; and &quot;synchronization&quot;. Collaboration is when we are actively solving a complex puzzle together—and that absolutely deserves a real-time conversation. Synchronization, on the other hand, is just making sure we are all on the same page. Even when a team works closely together at the office, synchronization doesn’t always require pulling everyone away from their desks at exact same time.</p><p>It’s just an interesting shift in perspective. Sometimes the best way to foster a sense of ownership isn’t to walk through a document together, but to give the team that great document and the uninterrupted time to process it beforehand.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=714ab22841ab" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your Simple Holiday Itinerary Is a Problem That Kept Mathematicians Awake for a Century]]></title>
            <link>https://derbyps.medium.com/your-simple-holiday-itinerary-is-a-problem-that-kept-mathematicians-awake-for-a-century-4d76097c7899?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/4d76097c7899</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[optimization]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[computer-science]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Tue, 26 May 2026 07:46:03 GMT</pubDate>
            <atom:updated>2026-05-26T07:46:03.065Z</atom:updated>
            <content:encoded><![CDATA[<blockquote><em>“</em>The book of nature is written in the language of mathematics<em>”<br> — </em>Galileo Galilei</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5jm1wMCtuiJWQYXNCe0OKg.png" /><figcaption>Sparked by human, visualized by Gemini</figcaption></figure><p>Did you know that if you are planning a holiday and want to visit just 15 destinations, there are over <strong>1.3 trillion</strong> possible routes to choose from? Even if you wrote a script and your computer checked a million routes every single second, it would still take over <strong>15 days</strong> of non-stop processing just to guarantee the absolute perfect itinerary.</p><p>Welcome to the Traveling Salesman Problem (TSP).</p><h3>The Deceptive Setup</h3><p>It starts innocently enough. You’re plotting out a long weekend — maybe a road trip around Bali to hit up a mix of beaches, temples, and cafes. You open a map, drop a few pins, and your brain instantly traces a decent route. It feels intuitive. You just start at your hotel, go to the closest spot, then the next closest spot, and eventually loop back home, right?</p><p>Not quite.</p><p>This seemingly flawless human intuition is called the “Nearest Neighbor” approach, and it is a trap. By strictly choosing the immediate shortest path at every stop, you frequently paint yourself into a geographical corner, forcing a massive, cross-map drive at the end of the day that ruins your schedule.</p><p>It sounds like a simple scheduling task. Yet, this exact scenario has kept mathematicians and engineers awake for nearly a century.</p><h3>The Combinatorial Explosion</h3><p>To a software engineer looking at this problem for the first time, it looks like a standard array waiting to be iterated. The brute-force backend solution is obvious: just calculate every possible route, compare the total distances, and pick the smallest one.</p><p>But then we hit the mathematical reality of factorials (<em>n</em>!).</p><p>The number of possible routes multiplies rapidly with every single stop you add to your list. It is a phenomenon known as a combinatorial explosion, and the scale is staggering:</p><ul><li><strong>5 stops</strong> = 120 possible routes. Your laptop solves this in milliseconds.</li><li><strong>10 stops</strong> = 3.6 million routes. A slight delay, but perfectly fine.</li><li><strong>15 stops</strong> = 1.3 trillion routes. You’ll be waiting a couple of weeks.</li><li><strong>20 stops</strong> = 2.4 quintillion routes.</li></ul><p>If your server checked a million routes per second, it would take over <strong>77,000 years</strong> to find the perfect path for just 20 stops.</p><p>This math doesn’t exist here just to confuse us; it exists to prove our limits. This is the precise moment you realize you cannot just throw more compute power at the problem. Brute force is fundamentally broken here.</p><h3>Translating the Map into a Matrix</h3><p>So, how do we actually model this problem for a machine? Computers don’t see scenic coastal roads or terrible afternoon traffic. They see a “Graph”.</p><p>Whether you are mapping out tourist spots, or writing backend logic to route pharmaceutical deliveries to local clinics, the architecture is the same:</p><ul><li><strong>Nodes:</strong> These are your points of interest (the cafes, the beaches, or the clinics).</li><li><strong>Edges:</strong> These are the lines connecting every node to every other node.</li><li><strong>Weights:</strong> Every line has a “cost” attached to it. This could be distance in kilometers, time stuck in traffic, or literal fuel costs.</li></ul><p>The objective function isn’t a massive, incomprehensible string of calculus. Translated into plain English, the rule is strict but simple:</p><blockquote>“Minimize the total cost of the lines, subject to the rule that you must visit every point exactly once, and end where you started.”</blockquote><p>That is the entire mathematical model. The difficulty isn’t in understanding the rules; it’s in finding the answer before the sun burns out.</p><h3>The Engineer’s Compromise</h3><p>Since finding the <em>absolute perfect</em> route for a decent-sized trip takes 77,000 years, we have to cheat.</p><p>As a software engineer, part of the job is knowing when to accept defeat to win the war. We abandon the pursuit of absolute perfection and turn to something called <strong>Heuristics</strong>.</p><p>But diving into exactly <em>what</em> a heuristic is, and how algorithms like Flower Pollination, Simulated Annealing, Genetic Algorithms, or Ant Colony Optimization actually mimic nature to find these paths? That gets very technical, very fast. Explaining the mechanics of these solutions is a massive topic on its own, and honestly, it is another story for another time. We will tackle the actual logic and solutions in a follow-up piece!</p><p>These algorithms don’t promise absolute mathematical perfection, they <em>do</em> promise a 95% optimal route in about two seconds.</p><p>The beauty of the Traveling Salesman Problem isn’t in finding the perfect answer. It’s in the engineering mindset — realizing when perfection is mathematically impossible and designing an elegant compromise that keeps our servers running and our holidays on track.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4d76097c7899" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Math Meets Messy Data: A Vector Search Story]]></title>
            <link>https://derbyps.medium.com/math-meets-messy-data-a-vector-search-story-cbc2672f4607?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbc2672f4607</guid>
            <category><![CDATA[vector-database]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Mon, 27 Apr 2026 03:29:54 GMT</pubDate>
            <atom:updated>2026-04-27T03:29:54.919Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>“Mathematics is the art of giving the same name to different things”<br> — Henri Poincaré</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZtPSCZ99dWgGoTkuHkkK2g.png" /><figcaption>Dreamt by human, drawn by Gemini.</figcaption></figure><p>I have degree in mathematics, but back in college I had absolutely no idea what half of it was actually used for. I knew exactly how to do the math, but I always wondered: “<em>Where is anyone actually implementing this in the real world?</em>”</p><p>It took becoming a developer working with search and vector databases for it to finally click. Suddenly, those abstract vectors from college aren’t just equations on a whiteboard anymore — they are the literal representation of data. Today, we use vectors as the core representation of meaning and context in data.</p><p>Recently, while doing some trial-and-error to build a search function, I found myself relying on those old college fundamentals to debug why my search results were acting weird. It reminded me that knowing the underlying math is the difference between blindly guessing and actually engineering a solution.</p><p>Here is what happened, and why understanding the difference between “distance” and “direction” might save your next search implementation.</p><h3>The Problem: Matching Messy Catalog Data</h3><p>I was building a search feature to match catalog products. In the real world, users make typos, skip words, or search for partial names.</p><p>Let’s say a user searches for the keyword: <strong>“panadol merah”</strong>.</p><p>When I first queried the database, I was honestly surprised by the sheer volume of variations for a single SKU. There is practically no way to anticipate every naming convention, typo, or packaging detail that might end up in the system.</p><p>At one point, I actually stopped and asked myself: <em>Wait, why am I reinventing the wheel here? Why not just plug in Elasticsearch and call it a day?</em> But the more I thought about it, the less sense it made. First, spinning up a massive, heavy search cluster just for this felt like total architectural overkill. I wanted a lightweight solution. Second, and more importantly, traditional fuzzy search just looks at spelling differences. I didn’t just need character matching; I needed to capture the actual <em>meaning</em> of the words.</p><p>To handle this, I turned to vector search.</p><h3>Attempt 1: The Distance Trap (Nearest Neighbor)</h3><p>Initially, I visualized vector search the way most of us do: vectors are just dots in a multi-dimensional space, and each dot represents a catalog name. When I run a query, the system turns my query into a dot and looks for the closest dots around it.</p><p>So, I used <strong>Euclidean Distance</strong> (Nearest Neighbor), which measures the straight-line physical distance between two points:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/314/1*SKyK_3jZsEn2S8KrkiXmMA.png" /></figure><p>When I ran the search for <strong>“panadol merah”</strong>, I expected the exact match first, and then other red (<em>merah</em>) variants. Instead, here were the top three results:</p><ol><li><strong>panadol merah</strong></li><li><strong>panadol kuning (</strong>this is different SKU)</li><li><strong>panadol merah extra strip</strong></li></ol><p>Wait, why did <strong>panadol kuning</strong> rank higher than <strong>panadol merah extra strip</strong>? The word &quot;kuning&quot; is noisy and changes the context completely. The &quot;extra strip&quot; version was still the red (<em>merah</em>) variant I actually wanted.</p><h3>Attempt 2: The Direction Shift (Cosine Similarity)</h3><p>I swapped my algorithm to <strong>Cosine Similarity</strong>. Instead of measuring the physical distance between two dots, Cosine measures the angle between them from the initial origin point:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/213/1*KEaC547dsCy8yw-boC66vw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*-mkkI2qwo1mYTM4EEYZOYg.png" /></figure><p>By dividing the dot product by the magnitudes, this formula normalizes the vectors. It essentially says: <em>I don’t care how long or short the text is, I only care which way it’s pointing.</em></p><p>The result:</p><ol><li><strong>panadol merah</strong></li><li><strong>panadol merah extra strip</strong></li><li><strong>panadol kuning</strong></li></ol><p>This was exactly what I wanted.</p><h3>The Core Lesson</h3><p>Through this trial-and-error, the abstract math finally became a practical engineering rule: <strong>the direction of a vector represents its context or meaning.</strong>* The <strong>panadol merah extra strip</strong> points in the exact same semantic direction as <strong>panadol merah</strong>. It just sits further down that same line because it has more words. Because the angle is tight, Cosine Similarity knows they mean the same thing.</p><p>*<strong>panadol kuning</strong> changes the semantic meaning entirely. Therefore, its vector points in a slightly different direction. Cosine correctly identifies that the angle is wider, meaning it&#39;s a less relevant match.</p><h3>Stop the Trial-and-Error Loop</h3><p>It’s incredibly common for developers to treat algorithms like a black box. We try Algorithm A; if it doesn’t satisfy the requirements, we switch to B, C, or D. If Algorithm B happens to yield the closest ranking, we stick with it for no valid reason other than “it worked”.</p><p>But if we actually understand what the vector math is doing — that distance can be skewed by text length, while direction preserves meaning — we stop guessing. We can confidently tweak, tune, or even combine algorithms to get exactly what we need.</p><p>Sometimes, those abstract college fundamentals really do pay off.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cbc2672f4607" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Python’s Two Dialects: Devs vs. Data Scientists]]></title>
            <link>https://derbyps.medium.com/pythons-two-dialects-devs-vs-data-scientists-ea18cb3b91bc?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/ea18cb3b91bc</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Mon, 23 Mar 2026 02:09:13 GMT</pubDate>
            <atom:updated>2026-03-23T02:09:13.353Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>“A language that doesn’t affect the way you think about programming is not worth knowing.” — Alan Perlis</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GQMndFtZ4JFJvgG-osgDjQ.png" /><figcaption>Sparked by human, visualized by Gemini</figcaption></figure><p>Python is a famous for its readability and strict “Zen of Python” philosophy. However, seeing how each role uses the language is like watching two people use the same multi-tool — one is building a house, and the other is perform a delicate surgery.</p><p>Developers build systems that need to run forever; Data Scientists build experiments that need to find an answer now. Because their end goals are so different, their approaches to Python are wildly contrasting.</p><h3>The Execution Environment</h3><p>Notebooks vs. Scripts. The most obvious divide is where the code actually lives. Devs typically write scripts. When they run command like <em>python app.py</em>, the code executes from top to bottom. It’s stateless, reproducible, and behaves exactly the same way every time it runs.</p><p>Data Scientist, however, live in the world of Jupyter Notebooks. Because data ingestion and model training can take hours, they can’t afford to restart the script just to tweak a visualization. The Notebook allows them to keep the “state” of data alive in memory. They can train a model in Cell 1 and spend the rest of the day running predictions in Cell 2 without ever reloading the data. It’s an iterative, non-linear way of coding that prioritizes immediate feedback over structured execution.</p><h3>The Wild West of Typing and Syntax</h3><p>Devs would look at a function and demand type hints and clear conditional flows to ensure the code doesn’t break six months from now. If a dev needs to add a calculation based on condition, they will write a clear if/else statement or a ternary operator, prioritizing maintainability above all else.</p><p>But in the fast-paced world of Data Science, you often see “math-first” logic. Code isn’t the final product; it’s just the tool to get the answer. This leads to some incredibly clever — if highly unconventional — syntax choices. You might see a function writing like this:</p><pre>def calculate_rent_cost(is_has_basement, room_count):<br>    return is_has_basement * 1000 + room_count * 2000</pre><p>To developer, multiplying a boolean by an integer is a crime. But to a Data Scientist, it makes perfect mathematical sense. Because Python evaluates True as 1 and False as 0, the math simply add 1000 if there’s a basement, and 0 if there isn’t. It eliminates the need for an if statement entirely — a habit born out of a vectorizing massive datasets where math operations run significantly faster than a logical loops.</p><h3>Code as a Product vs. Code as a Tool</h3><p>Ultimately, the contrast comes down to the final deliverable. A Dev writes Python to run app that handles unexpected edge cases gracefully. A Data Scientist writes Python to answer a question, and once the insight is found, the code has served its purpose. It might lack type hints and use strange mathematical shortcuts, but it gets the job done at breakneck speed.</p><p>This flows really well and sets up the contrast beautifully.</p><h3>Two Dialects, One Language</h3><p>As an engineer who is always looking to understand different ways of solving problems, seeing these two worlds operate side-by-side has been eye-opening.</p><p>It is easy for dev to look at a Jupyter Notebook, see a missing type hint or a boolean multiplied by an integer, and immediately want to refactor it. But once you understand the context — that they are performing surgery to extract insights rather than building hospital itself — there is an undeniable elegance in how Data Scientist bend Python to their will.</p><p>At the end of the day, whether you are building a robust system design to run forever, or running a messy experiment to find a critical answer right now, we are all relying on the exact same multi-tool. It’s just a matter of knowing which attachment to use.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ea18cb3b91bc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Serverless? Kenapa Tidak?]]></title>
            <link>https://medium.com/farmacare/serverless-kenapa-tidak-d0840fe394c3?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0840fe394c3</guid>
            <category><![CDATA[serverless]]></category>
            <category><![CDATA[lambda]]></category>
            <category><![CDATA[backend]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Tue, 06 Dec 2022 00:45:21 GMT</pubDate>
            <atom:updated>2022-12-06T00:45:21.178Z</atom:updated>
            <content:encoded><![CDATA[<p>Pengembangan sebuah <em>apps</em> tak lepas dari pemilihan arsitekturnya. Semuanya dimulai dari <em>monolith</em> atau yang secara singkat kita membuat satu <em>apps</em> besar yang menangani semuanya. Karena satu <em>apps </em>tersebut melayani banyaknya <em>user </em>sekaligus, bisa saja <em>apps </em>tersebut akan terasa lambat. Kemudian muncullah arsitektur <em>microservice </em>atau yang bisa kita sebut membagi satu <em>apps </em>besar ke beberapa <em>apps </em>kecil yang independen antar satu sama lain. Pemecahan ini bertujuan untuk membagi <em>load </em>dari <em>user</em>-nya.</p><p>Namun, meskipun menggunakan arsitektur <em>monolith </em>ataupun <em>microservice </em>kita masih harus memikirkan biaya server yang pastinya itu tidak murah. Akhirnya muncullah arsitektur <em>serverless </em>dimana kita bisa membuat <em>apps </em>tanpa harus memiliki server yang artinya biaya dan segala macam konfigurasi server untuk menjalankan <em>apps </em>kita akan berkurang cukup signifikan.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0B38OzjHl5x8evEKCeUONw.png" /></figure><p>Serverless bekerja dengan cara kita menitipkan kode kita di dalam komputer atau server yang dimiliki oleh penyedia layanan seperti <em>AWS </em>atau <em>GCP</em>. Sehingga dengan arsitektur <em>serverless</em> kita tidak perlu memikirkan biaya dan konfigurasi <em>server</em>, <em>developer </em>hanya perlu fokus untuk menulis sebuah kode. Untuk urusan <em>server</em> sudah diatur oleh pihak dimana kita menaruh fungsi kita sehingga kita hanya ditagih biaya ketika fungsi kita dijalankan saja. Selain biaya yang sangat murah, kelebihan lainnya adalah <em>high</em> <em>availability </em>yang artinya <em>apps</em> kita tidak akan <em>down</em> atau mati walaupun ada <em>error</em>.</p><p>Dibalik kelebihan serverless, pasti juga ada kekurangannya. Misalnya serverless hanya mendukung beberapa bahasa pemrograman saja tergantung dari penyedia layanan. selain itu paradigma pemrograman kita akan sedikit berbeda dengan arsitektur yang lainnya. Fungsi kita juga akan masuk kedalam kondisi hibernasi jika dalam beberapa waktu tidak dieksekusi, yang mengakibatkan memberikan respon yang sedikit lebih lambat untuk pertama kalinya karena harus “membangunkan” terlebih dahulu fungsi yang terhibernasi tersebut.</p><p>Setiap arsitektur memiliki kelebihan dan kekurangan masing-masing. Untuk memilih sebuah arsitektur juga bergantung pada jenis <em>apps </em>yang akan kita bangun diatasnya. Jika kita ingin membangun suatu <em>high availability apps </em>dengan biaya yang murah, proses development yang relatif cepat, dan tidak ingin dihantui oleh biaya dan konfigurasi server maka <em>serverless </em>adalah pilihan yang tepat.</p><p><a href="https://aws.amazon.com/id/lambda/">referensi</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d0840fe394c3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/farmacare/serverless-kenapa-tidak-d0840fe394c3">Serverless? Kenapa Tidak?</a> was originally published in <a href="https://medium.com/farmacare">Farmacare Crew</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Komunikasi Tim 101]]></title>
            <link>https://medium.com/farmacare/komunikasi-tim-101-2e15f47309fd?source=rss-94ccfc5c719b------2</link>
            <guid isPermaLink="false">https://medium.com/p/2e15f47309fd</guid>
            <category><![CDATA[komunikasi]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[silikon-bali]]></category>
            <dc:creator><![CDATA[Derby]]></dc:creator>
            <pubDate>Thu, 21 Jan 2021 01:30:39 GMT</pubDate>
            <atom:updated>2021-01-21T01:30:39.208Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qZSuTZO4k4SzhE6B" /><figcaption>Photo by <a href="https://unsplash.com/@krakenimages?utm_source=medium&amp;utm_medium=referral">krakenimages</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><blockquote>coming together is a beginning; keeping together is progress; working together is success<br> — Henry Ford</blockquote><p>Banyak orang yang memiliki ide di dalam kepala mereka namun sulit untuk mengatakan ide mereka. Tak jarang mereka hanya mampu mengeluarkan sebagian ide mereka. Hal inilah yang kadang identik dengan diri seorang <em>developer, </em>hanya mampu menuangkan ide mereka ke dalam baris kode tanpa bisa membagi ide mereka kepada sesama manusia. Jika bekerja sendiri mungkin tidak akan timbul masalah, namun apa jadinya jika bekerja dalam tim?</p><p>Idealnya, sebuah tim terdiri dari beberapa orang. Untuk mencapai <em>goal</em>, tentu kerja sama yang baik antar anggota tim adalah syarat utamanya. Lantas apa yang menjadi indikator bahwa kerja sama tim sudah dijalankan dengan baik? Tidak ada patokan pasti, namun komunikasi efektif adalah hal yang paling jelas yang bisa menjadi indikatornya.</p><p>Bagaimana cara agar komunikasi dengan anggota tim atau rekan kerja kita berjalan dengan lancar? Sebelumnya, mari kita bahas tentang tipe rekan kerja berdasakan <em>“Social Styles” model </em>menurut <strong>Robert and Dorothy Bolton</strong>.</p><ol><li><strong><em>Driver</em></strong>; adalah tipe pemimpin. Tegas dan berani mengambil risiko yang ada. Hampir semua <em>driver </em>memiliki wibawa yang tinggi dan disegani. — <em>“Let’s go! Let’s get the job done right now!”</em></li><li><strong><em>Analytical; </em></strong>adalah tipe orang yang berfikir secara logis dan beracuan terhadap data. Apa yang mereka dapatkan haruslah berdasarkan fakta. — <em>“Let’s get the job done, but let’s get it done right!”</em></li><li><strong><em>Amiable; </em></strong>adalah tipe yang suka mengalah, tidak menginginkan konflik dan memiliki empati yang besar terhadap sesama. — <em>“Let’s get the job done, but let’s focus on the people!”</em></li><li><strong><em>Expressive; </em></strong>adalah tipe periang, optimis dan suka terhadap tantangan. tidak ada hari yang membosankan bagi mereka. —<em> “Let’s get this done! We can do it!”</em></li></ol><p>Setelah mengetahui tipe manakah rekan kerja kita, bagaimana cara berkomunikasi dengan mereka? Tiap tipe memiliki “cara” tersendiri agar komunikasi yang dihasilkan efektif. Secara garis besar, ada beberapa “cara” yang bisa kita lakukan sebagai rekan kerja untuk meningkatkan keefektifan dalam berkomunikasi terhadap suatu tipe.</p><ol><li><strong><em>Driver</em></strong>; Jangan pernah berbasa-basi. Langsung katakan tujuan utama pembicaraan. Mereka tidak memiliki waktu untuk mendengarkan basa-basimu.</li><li><strong><em>Analytical; </em></strong>Jangan pernah berkata tanpa ada data/fakta yang pasti. Sulit bagi mereka untuk menerima hal berdasarkan opini. Yakinkan mereka dengan alasan yang logis.</li><li><strong><em>Amiable; </em></strong>Jangan pernah berbicara dengan nada yang tinggi. Pilihlah kata-kata yang tidak menyinggung mereka. Jalin komunikasi pribadi dan bangun kepercayaan mereka terhadap kita.</li><li><strong><em>Expressive; </em></strong>Jangan pernah tunjukkan ketidak tertarikanmu. Tunjukkanlah antusiasmu dalam sebuah topik pembicaraan, biarkan mereka mengeluarkan ide mereka dahulu, dan jangan lupa untuk tersenyum.</li></ol><p>Untuk mengetahui tipe manakah kalian, dapat melakukan <em>quiz </em><a href="https://www.thevantagepoint.ca/sites/default/files/High%20Performing%20Teams%20-%20Self%20assessment%20handout.pdf"><strong>disini</strong></a>.</p><p>Dalam pengembangan suatu <em>apps </em>yang besar, akan sangat memakan waktu dan tenaga jika hanya dikerjakan oleh satu orang <em>developer, </em>idealnya diperlukan beberapa <em>developer </em>dengan latar belakang yang berbeda-beda yang bekerja sama untuk mengembangkannya. Selain masalah komunikasi, masalah lain yang sering dihadapi adalah bagaimana cara untuk memanajemen suatu proyek.</p><p>Bagaimana cara agar memastikan dua orang <em>developer </em>tidak mengerjakan satu hal yang sama? Tentunya kita membutuhkan suatu <em>framework </em>untuk mengatur <em>tasks </em>yang akan dikerjakan dan mengomunikasikan apa yang akan dikerjakan sehingga <em>developer </em>lain tidak mengerjakan hal yang sama.</p><p>Banyak <em>project management framework </em>untuk membuat suatu <em>apps, </em>salah satu contohnya adalah <strong><em>scrum </em></strong>yang mengedepankan proses <em>feedback </em>iteratif yang sangat cepat sehingga kita dapat melakukan dua kali pekerjaan dengan setengah kali waktu. Tentunya walau pun sudah memakai <em>tools </em>apa pun tanpa diiringi dengan komunikasi antar <em>developer </em>yang baik, tidak akan mendapatkan hasil yang maksimal.</p><p>Jika kita sudah menemukan cara yang efektif bagaimana berkomunikasi dengan rekan kerja kita, akan terbentuk kerja sama tim yang solid dan semakin sering kita berkomunikasi dengan rekan kerja kita, akan terbentuk ikatan antar anggota tim yang kuat. Akibatnya dalam sisi pengembangan suatu <em>apps </em>yang melibatkan banyak <em>developer, developer </em>tersebut akan terbiasa dalam berkomunikasi dan akan mengubah stigma “<em>developer </em>susah untuk bersosialisasi”</p><p>Penting sekali bagi seorang <em>developer </em>untuk memiliki kemampuan berkomunikasi yang baik dengan rekan timnya sehingga akan terbentuk tim yang solid dan memudahkan dalam membangun suatu aplikasi.</p><h4><strong>Referensi</strong></h4><p><a href="https://thehiredguns.com/four-types-of-co-worker/">https://thehiredguns.com/four-types-of-co-worker/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2e15f47309fd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/farmacare/komunikasi-tim-101-2e15f47309fd">Komunikasi Tim 101</a> was originally published in <a href="https://medium.com/farmacare">Farmacare Crew</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>