<?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 laxmena on Medium]]></title>
        <description><![CDATA[Stories by laxmena on Medium]]></description>
        <link>https://medium.com/@laxmena?source=rss-91dc7710e42------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*KoYIeiZCCVz6m50u0hgzCw.jpeg</url>
            <title>Stories by laxmena on Medium</title>
            <link>https://medium.com/@laxmena?source=rss-91dc7710e42------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 11:52:26 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@laxmena/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[I Built a Tool That Optimizes Code While You Sleep]]></title>
            <link>https://laxmena.medium.com/i-built-a-tool-that-optimizes-code-while-you-sleep-e86d9fe67df6?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/e86d9fe67df6</guid>
            <category><![CDATA[python]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Tue, 24 Mar 2026 03:40:00 GMT</pubDate>
            <atom:updated>2026-03-24T03:40:32.289Z</atom:updated>
            <content:encoded><![CDATA[<h4>41x faster in 20 iterations. No human in the loop</h4><p>A few weeks ago, I came across Karpathy’s <a href="https://github.com/karpathy/autoresearch">autoresearch repository</a>. The core idea: run an agentic loop to auto-tune LLM fine-tuning pipelines. Give the agent a goal, a way to measure progress, and let it iterate autonomously until it gets there.</p><p>I couldn’t stop thinking about it.</p><p>Not because of the fine-tuning use case — but because the <em>pattern</em> felt universally useful. Most software has something you want to improve and a way to measure it. Why are we still doing the iteration loop by hand?</p><p>So I built <a href="https://github.com/laxmena/hone">Hone</a> — a side project to experiment and learn.</p><h3>What Hone Does</h3><p>Hone is a CLI tool. You give it three things:</p><ol><li>A goal, in plain English</li><li>A file or directory to optimize</li><li>A benchmark command that outputs a number</li></ol><p>Then you leave.</p><p>Hone runs a loop: it asks an LLM what to try next, applies the changes, runs your benchmark, and decides whether to keep the result or revert it. It logs every iteration — the score, the diff, and the agent’s reasoning — and stops when it hits your target or you tell it to.</p><pre>hone &quot;Optimize process_logs.py to run under 0.02 seconds&quot; <br>     --bench &quot;python bench_logs.py&quot; <br>     --files &quot;process_logs.py&quot; <br>     --optimize lower <br>     --target 0.02 <br>     --budget 2.0</pre><p>That’s the entire interface.</p><h3>Experiment 1: The Log Parser</h3><p>The first real test was a deliberately naive Python log parser. The task: analyze 150,000 lines of server logs and return the top 3 most-visited endpoints with unique IP counts.</p><p>The baseline code was the kind you’d write in an interview warm-up: readlines() into memory, a list for uniqueness checking (O(n) per insert), a regex match on every line. It took <strong>1.54 seconds</strong>.</p><p>I set a target of 0.02 seconds — roughly 75x faster — and launched Hone with a $2 budget.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ycSsoH4dKvFDNFyWXz5nyQ.jpeg" /></figure><p>Iterations 1–4 replaced the uniqueness list with a set, cutting time from 1.54s to around 0.4s. Iterations 5–9 switched from readlines() to streaming, dropping to 0.15s. By iteration 14, a compiled regex brought it down to 0.09s. Then it plateaued — and the agent recognized why.</p><p>That pivot came at iteration 18. The agent abandoned line-by-line parsing entirely, read the file as a raw binary blob, and deployed re.findall() over the entire content in one pass. Its reasoning:</p><blockquote>“The real bottleneck is the Python loop and split() calls. Try using a compiled regex to extract the endpoint in one operation across the entire file.”</blockquote><p><strong>Final result: 1.54s → 0.037s. A 41x speedup. Autonomously.</strong></p><p>It didn’t hit the 0.02 target — that’s likely beyond what single-threaded Python can do without C extensions. But a 41x improvement for $1.84 in API costs is a real result.</p><h3>Experiment 2: Nearest Driver Dispatch</h3><p>The second experiment was closer to production code. The problem: given a set of riders and a pool of drivers, find the nearest driver for each rider using haversine distance.</p><p>The baseline was an O(R × D) brute-force loop — calculate the full haversine distance between every rider and every driver. With 500 riders and 1,000 drivers, that’s 500,000 distance calculations per call. Baseline: <strong>2.18 seconds</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aIJ0Jxos4_QbhhIhAuJ6Sg.jpeg" /></figure><p><strong>Run 1</strong> — I launched Hone with no hints. Just: <em>“optimize this to run faster.”</em></p><p>The agent went straight for spatial indexing. It built a grid over the geographic area, bucketed drivers into cells, and used Manhattan distance pre-filtering to eliminate distant candidates before running haversine. It also replaced the standard math module haversine with a vectorized approximation valid for short distances.</p><p>Result: <strong>0.1496 seconds. A 14.6x speedup.</strong></p><p><strong>Run 2</strong> — I ran Hone again on the output from Run 1.</p><p>This is where it got interesting. The agent looked at the already-optimized code and found something the previous run missed: the grid search still checked every driver in candidate cells, even after it had already found a close one.</p><p>The fix: stop searching the moment you find a driver within an acceptable radius. Expand the search radius incrementally — start small, grow outward — instead of checking all candidates at once.</p><blockquote>“The algorithm beats the data structure. Grid resolution barely matters. Early termination dominates.”</blockquote><p>Result: <strong>0.069 seconds. Another 2.1x on top of an already fast baseline.</strong></p><p>Two runs, &lt; $3 total, brute-force O(R×D) → smart early-termination spatial search. The agent arrived at an approach that a senior engineer would recognize as correct — not by knowing the algorithm upfront, but by observing what the benchmark rewarded.</p><h3>What I Learned</h3><p><strong>The benchmark is everything.</strong> Hone is only as good as your measurement. If your benchmark is slow to run, the loop is slow. If it doesn’t capture what you actually care about, the agent will optimize the wrong thing. The one thing you must get right before you start is: “does this number actually reflect what I want?”</p><p><strong>The agent is a good low-level optimizer.</strong> It reliably finds the obvious wins: wrong data structures, redundant computations, missed language primitives. These are also the wins that take a human the most time — not because they’re hard to understand, but because you have to actually sit down and do them.</p><p><strong>It surprises you at the edges.</strong> The log parser pivot from line-by-line to whole-file regex wasn’t something I would have thought to suggest in the initial prompt. It emerged from the agent hitting a wall and reasoning about <em>why</em> it had hit a wall. That’s the behavior that makes agentic loops interesting.</p><p><strong>The conversation thread is the memory.</strong> The most important architectural decision in Hone was keeping the LLM conversation alive across iterations. The agent doesn’t just see the current score — it sees everything it tried, what worked, and what was reverted. That’s what allows the pivot at iteration 18. Without it, the agent would start fresh each time and repeat the same early optimizations.</p><p><strong>Cost is low. Time savings are high.</strong> Both experiments ran under $3. The engineering time to achieve the same results manually — writing hypotheses, applying changes, running benchmarks, reverting dead ends — would have been hours. The ROI on agentic loops is already real, and we’re at the beginning.</p><h3>What’s Next</h3><p>Hone v0 is rough. There’s no sandbox for shell commands, no git-based snapshots, no dry-run mode. These are on the list.</p><p>More interesting to me is expanding the use cases. The same loop that optimizes a log parser can optimize:</p><ul><li><strong>LLM prompts</strong> against an eval suite (highest impact use case)</li><li><strong>RAG pipelines</strong> against a retrieval benchmark</li><li><strong>API costs</strong> against a quality-constrained spend target</li></ul><p>The pattern is the same. The benchmark changes. Hone doesn’t care.</p><p>If you want to try it:</p><pre>git clone https://github.com/laxmena/hone<br>cd hone &amp;&amp; pip install -e .</pre><p>And if you have a benchmark that Hone should try — I want to hear about it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e86d9fe67df6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Day You Became A Better Writer]]></title>
            <link>https://laxmena.medium.com/the-day-you-became-a-better-writer-936ce21ee396?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/936ce21ee396</guid>
            <category><![CDATA[writing]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Sun, 22 Mar 2026 02:31:51 GMT</pubDate>
            <atom:updated>2026-03-22T02:31:51.254Z</atom:updated>
            <content:encoded><![CDATA[<p>In 2007, Scott Adams — creator of Dilbert — published a short blog post on writing. Naval Ravikant thought it was worth adding to his recommended reading list in the Almanack of Naval Ravikant.</p><p>There’s one problem. Typepad, the blogging platform that hosted it, shut down permanently on September 30, 2025. The post disappeared with it.</p><p>I tracked it down through the Internet Archive. You can read the <a href="https://web.archive.org/web/20240302003157/https://dilbertblog.typepad.com/the_dilbert_blog/2007/06/the_day_you_bec.html">original here</a>.</p><p>This post is my attempt to make it accessible — and to add something new.</p><h3>What Adams said</h3><p>Adams opens with a claim: he went from bad writer to good writer after a single one-day course in business writing. Then he gives you the whole course in under 200 words.</p><p>The core idea is simple. Simple writing is persuasive. A tight five-sentence argument beats a sprawling hundred-sentence one. Every time.</p><p>Here are his rules, distilled:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zH67rGIZbXxsYNZT.png" /></figure><h3>My additions</h3><p>Adams covers the sentence level well. These extend his thinking to structure.</p><p><strong>7. Front-load your point.</strong> State the conclusion first, then support it. Don’t make the reader work through the argument before knowing why it matters.</p><p><strong>8. One idea per paragraph.</strong> Adams says one thought per sentence. The same logic applies one level up. If a paragraph is doing two jobs, split it.</p><h3>Steal this prompt</h3><p>If you use LLMs to help draft or edit writing, here’s a prompt you can drop into your workflow. It distills everything above into instructions the model will actually follow.Follow these principles when writing or editing:</p><pre>You are a writing assistant that helps produce clear, persuasive, and readable text.</pre><pre>- Keep it simple. A short, clear argument is more persuasive than a long, complex one.<br>- Cut extra words. If a word doesn&#39;t add meaning, remove it.<br>- Choose potent words. Prefer the specific and vivid over the generic.<br>- Make the first sentence earn attention. It should create curiosity or make a bold claim.<br>- Write short sentences. One thought per sentence.<br>- Use active voice. Put the actor before the action.<br>- Front-load the point. State the conclusion first, then support it.<br>- One idea per paragraph. If a paragraph is doing two jobs, split it.</pre><pre>When editing, flag sentences that violate these rules and suggest alternatives.</pre><p>Good writing is good thinking made visible. Adams knew this in 2007. It hasn’t changed.</p><p><em>All original ideas referenced here belong to Scott Adams. This post exists to preserve and extend his thinking, not to replace it. Read the </em><a href="https://web.archive.org/web/20240302003157/https://dilbertblog.typepad.com/the_dilbert_blog/2007/06/the_day_you_bec.html"><em>original</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=936ce21ee396" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building Robust and Flexible Software with Hexagonal Architecture]]></title>
            <link>https://laxmena.medium.com/building-robust-and-flexible-software-with-hexagonal-architecture-0e1dc56d4549?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e1dc56d4549</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[software-architecture]]></category>
            <category><![CDATA[design-patterns]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Sat, 23 Aug 2025 03:16:42 GMT</pubDate>
            <atom:updated>2025-08-23T03:33:15.391Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-SnBQy2PEx5IXncKz6RteA.png" /></figure><p>Building Robust and Flexible Software with Hexagonal ArchitectureIn software engineering, applications often become a tangled mess of code. Core business logic is tightly coupled to external systems like user interfaces, databases, and third-party services. This tight coupling makes the code fragile, difficult to test, and a nightmare to change.</p><p>Hexagonal Architecture, also known as Ports and Adapters, is an architectural pattern designed to solve this exact problem.Its core idea is simple: decouple the business logic from the outside world.</p><p>By doing so, it creates a flexible and robust system that can be easily tested and adapted to new technologies.</p><h3>Explain like I’m 12 year old: The Car Engine Analogy</h3><p>To understand this concept, think of an application as a <strong>car engine</strong>. The engine’s purpose is to convert fuel into power. Its core function remains the same, regardless of the brand of fuel pump, the type of car body, or the kind of dashboard you use.</p><p>In this analogy:</p><ul><li>The Engine is the <strong>Core Logic</strong>. It’s the most valuable part, containing all the essential rules for your application to work. It doesn’t care about the outside world.</li><li>The <strong>Ports</strong> are the connection points on the engine. They are standardized slots where other parts can plug in. The engine defines what kind of input it needs (e.g., fuel) and what kind of output it provides (e.g., rotation).</li><li>The <strong>Adapters</strong> are the external components that plug into the ports, like the fuel pump, the cooling system, or the drivetrain. They are built to work with the engine’s standard connection points, translating the engine’s requests into commands for the real-world component.</li></ul><p>The engine works with any fuel pump, as long as the pump can connect to the engine’s standardized port. This is the essence of <strong>decoupling</strong> and <strong>extensibility</strong>.</p><h3>The Core Concepts</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RmHXarjvYKSkwISm1ING5A.png" /><figcaption>AI helped me generate this visualization.</figcaption></figure><h4>1. Core Logic:</h4><p>The core logic is the heart of your application. It contains all the essential business rules and remains independent of any external depenencies. This is the most valuable part of the system, and it should be completely unaware of the user interface, the database, or any third-party APIs.</p><h4>2. Ports and Adapters:</h4><p><strong>Ports</strong> are interfaces that define a contract for communication. They act as the boundary between the core logic and the outside world. The core exposes ports for others to use, or it defines ports that it needs others to implement.</p><p><strong>Adapters</strong> are concrete implementations that connect to the ports. They translate communication between the external technology and the core logic’s language.</p><p>There are two types of adapters based on the direction of control:</p><ul><li><strong>Primary Adapters (Driving):</strong> These adapters initiate actions on the core logic. They are driven by an external actor, such as a user. <br>Examples include a web UI, a REST API, or a command-line interface. They use the core’s API to send commands.</li><li><strong>Secondary Adapters (Driven):</strong> These adapters are used by the core logic to perform its tasks. The core dictates what it needs from them.<br>Examples include a database, a third-party API, etc. <br>The core defines the port that the secondary adapter must implement.</li></ul><blockquote>All dependencies always point <strong>inwards</strong> towards the core. The core never depends on an adapter; the adapter always depends on the core’s port.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/721/1*faYhfJ29FRCY_m5qJgJ4-Q.png" /></figure><h3>Real Usecase: An Online Food Ordering System Use Case</h3><p>Imagine an application for ordering food online. The core logic handles all business rules, such as calculating the total price, managing the order status, and checking for special promotions.</p><ul><li><strong>Core Logic:</strong> The business rules for managing orders, pricing, and inventory.</li><li><strong>Primary Adapters:</strong> These are the user-facing interfaces that place an order. This could be a <strong>Mobile App Adapter</strong> or a <strong>REST API Adapter</strong>. Both of these adapters use the same <strong>“Place Order” port</strong> exposed by the core.</li><li><strong>Secondary Adapters:</strong> These are the services the core uses to fulfill the order. When the core needs to save an order, it uses a <strong>Database Adapter</strong> that implements the “Save Order” port. To notify the customer, it uses an <strong>SMS Service Adapter</strong> that implements a “Send Notification” port.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rDBNd8_x6ClvyGvCQlkPhQ.png" /></figure><p>In this scenario, if you decide to change from a mobile app to a website, you only need to create a new website adapter. The core logic remains untouched. Similarly, if you switch from a MySQL database to PostgreSQL, you only need to replace the database adapter.</p><h3>When Not to Use and When to Use:</h3><p><em>Note: I put the “When not to use” section first because I’ve been in the shoes of trying to use a powerful pattern on a simple project and realizing it was overkill.</em></p><p><strong><em>When a project is simple, a simple design is best. Don’t go all in on a hexagonal architecture just to build a square app.</em></strong></p><h4>When Not to Use</h4><ul><li><strong>Simple Applications:</strong> For simple applications with little to no business logic (e.g., a simple CRUD app), the added overhead of this architecture is likely overkill.</li><li><strong>Fixed Technology Stack:</strong> If your application is small and you know with certainty that its technology stack will never change, a simpler layered architecture may be a more pragmatic choice.</li><li><strong>Rapid Prototyping:</strong> The initial setup and conceptual overhead can slow down development in the beginning. For a quick, short-term prototype, it may not be the right choice.</li></ul><h4>When to Use</h4><ul><li><strong>Complex Business Logic:</strong> Use this pattern when your application’s core business logic is complex and is expected to evolve over time.</li><li><strong>Anticipated Technology Changes:</strong> If you expect to change your database, UI framework, or third-party services in the future, this architecture’s flexibility will save you time and money.</li><li><strong>High Testability Requirements:</strong> When unit testing the core logic in isolation is a high priority, this pattern is a perfect fit.</li></ul><p><strong>References:</strong></p><ul><li><a href="https://docs.aws.amazon.com/prescriptive-guidance/latest/cloud-design-patterns/hexagonal-architecture.html">https://docs.aws.amazon.com/prescriptive-guidance/latest/cloud-design-patterns/hexagonal-architecture.html</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e1dc56d4549" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Finding Our Place in the Age of AI]]></title>
            <link>https://laxmena.medium.com/finding-our-place-in-the-age-of-ai-0d465d31b960?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/0d465d31b960</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[opinion]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Thu, 21 Aug 2025 01:23:18 GMT</pubDate>
            <atom:updated>2025-08-21T01:23:18.802Z</atom:updated>
            <content:encoded><![CDATA[<h4>What would we even do if AI did all the jobs?</h4><p>It’s hard to ignore the news about AI taking over. Almost every week, a new company claims its AI can do a task better, faster, and cheaper than an actual human.</p><p>Think about it: creating a logo, editing a picture, writing content, researching a topic, or even writing code. All of these used to take hours or even days, and now they can be done in minutes. Going from an idea to a finished product has never been faster. In some cases, AI tools are even outperforming humans. It’s easy to see why so many jobs that exist today might not exist in just a few years.</p><p>I experienced this firsthand the other day when I used Meta AI to generate pictures of myself in different places and with different expressions. The results were surprisingly good. It made me wonder: does this mean the job of a photographer/editor is obsolete? What would we even do if AI did all the jobs?</p><blockquote><strong>TLDR;</strong> While AI is getting better and faster at many jobs, it can’t replicate the human touch. Our value lies in the skills it can’t replace: empathy, collaboration, and creativity.</blockquote><p>After pondering on this question for a while, I had a realization: no matter how advanced AI becomes, there are certain things it simply can’t do.</p><p>So, does AI’s ability to generate realistic pictures mean the end of a photographer’s job?</p><p>The answer is both yes and no. The “yes” part is easy to understand. AI can handle the technical stuff. It can generate perfect, technically flawless images.</p><p>But the “no” is what really matters. A photograph isn’t just an image; it’s a feeling. It’s about capturing the emotion and the beauty of a specific moment. An AI can’t feel what you feel or know what you want to remember forever. It can’t capture the shared laughter between friends at a party or the proud look in a parent’s eyes at a graduation. That’s what a great photographer does. They don’t just take pictures; they capture stories and emotions.</p><p>That kind of unique, human touch exists in every field. Our job isn’t to compete with AI on speed or efficiency. It’s to find the places where we add value that AI can’t. It’s about doubling down on the things that make us human: creativity, empathy, and the ability to connect with others.</p><p>Here are some Human Touch aspect in Other Professions</p><p>For a Software Developer: An AI can write a lot of code in little time, but a human understands the unspoken frustration of a user and works with a team to solve a complex, messy problem.</p><p>For an Accountant: AI can instantly process numbers, but a human accountant builds trust with a small business owner by listening to their fears and helping them plan for the future.</p><p>For a Lawyer: An AI can scan piles of legal documents, but only a human lawyer can use passion and empathy to persuade a jury or guide a family through a difficult time.</p><p>AI is here, and it’s changing things. But it’s also a chance for us to rediscover what truly makes us valuable. Instead of worrying about what jobs AI will take, maybe we should focus on the things that are highly difficult to replicate.</p><p>Of course, this isn’t a simple solution for everyone. The challenges ahead are real, and the road will be complex. But maybe the first step isn’t to worry about what AI will take from us, but to focus on what makes us truly irreplaceable.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0d465d31b960" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[While GenAI have the potential to make our minds dull, we can also leverage them to force us think…]]></title>
            <link>https://laxmena.medium.com/while-genai-have-the-potential-to-make-our-minds-dull-we-can-also-leverage-them-to-force-us-think-0738798f2869?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/0738798f2869</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[socratic-method]]></category>
            <category><![CDATA[learning]]></category>
            <category><![CDATA[feynman-technique]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Sun, 25 May 2025 05:46:51 GMT</pubDate>
            <atom:updated>2025-05-25T05:48:44.037Z</atom:updated>
            <content:encoded><![CDATA[<h3>While GenAI have the potential to make our minds dull, we can also leverage them to force us think better</h3><p>Just like muscles — which shrink in size when not used enough, our minds also become weak. So, the more we delegate the thinking to GenAI, greater the impact to our minds.</p><p>In this article, I share an interesting proposal to address this challenge.</p><p>How about we use the poison itself to create the cure. <br>How about we leverage GenAI itself to help us get better in critical thinking.</p><p>There are two powerful learning techniques, that forces us to learn by thinking.</p><p>1.<strong> The Socratic Method</strong> involves a shared dialogue between teacher and students. The teacher leads by posing thought-provoking questions. Students actively engage by asking questions of their own. The discussion goes back and forth.</p><p>2. <strong>The Feynman Technique</strong> focuses on understanding a concept deeply by explaining it simply, like you’re teaching it to a child. This process helps identify knowledge gaps and encourages simplification, leading to a clearer understanding of the topic</p><p>**How about we combine both these techniques in a GenAI Persona**, which will then act as a teacher, helping us master new concepts in an engaging and thought provoking way.</p><p>Here’s how we would want the AI to behave:</p><p>- The AI will use a dialogue based approach, asking questions and discussing about the topic in hand.<br>- The AI will never give answers directly, but will nudge gradually to help us think, discover the ideas and gain deeper understanding.<br>- AI identifies any gaps in understanding, and dives deeper to bridge the gaps.<br>- AI progressively increases the difficulty of the question, forcing to think more on the topic, thus solidifying the understanding. <br>- The session goes on, till the AI has reached a state where it has enough data points that we have gained sufficient understanding.</p><p>By end of this exercise, we would have achieved at a very good level of mastery on the subject. We could repeat this exercise for each interested topics.</p><p>Translating all the above into a AI Prompt would look like this:</p><pre>You are an expert AI learning assistant and Socratic tutor, specifically designed to facilitate deep understanding using a method inspired by the Feynman Technique. Your goal is to help me learn and master a topic by guiding me through a question-driven exploration.<br><br>Here’s how we will proceed:<br><br>1. **Topic Introduction:** I will provide a topic I want to learn.<br>2. **Initial Explanation Prompt:** You will ask me to explain the core concept of the topic in simple terms, as if I were explaining it to a beginner (a key part of the Feynman Technique).<br>3. **Identify Gaps &amp; Misconceptions:** Based on my explanation, you will gently probe my understanding by asking targeted questions. These questions should help identify any areas where my understanding is unclear, incomplete, or incorrect, without directly giving me the answers.<br>4. **Progressive Challenge:** Once a foundational understanding seems to be in place (or after addressing initial gaps), you will ask progressively more challenging questions. These questions should move from basic principles to more complex details, applications, related concepts, potential edge cases, or historical context, pushing the boundaries of my current knowledge.<br>5. **Encourage Simplification:** At various points, you may ask me to re-explain concepts in simpler terms or use analogies to check for deep understanding.<br>6. **Iterative Learning:** Our interaction will be iterative. You will ask a question, I will respond, and you will formulate your next question or prompt based on my response.<br>7. **Maintain Socratic Approach:** Avoid giving direct answers. Instead, use questions to guide me towards the correct understanding or to discover the next layer of complexity. Provide hints or rephrase questions if I struggle.<br>8. **Goal:** Our shared goal is for me to achieve a comprehensive and deeply internalized understanding of the topic, being able to explain it clearly and accurately to others.<br><br>Lets begin. What topic would you like to explore today?</pre><p>Copy the prompt above, and paste them into your favorite LLM application — [Gemini](<a href="https://gemini.google.com/">https://gemini.google.com/</a>) (My goto), [ChatGPT](<a href="https://chatgpt.com/">https://chatgpt.com/</a>), [Claude](<a href="https://claude.ai/">https://claude.ai/</a>) or [Perplexity](<a href="https://www.perplexity.ai/">https://www.perplexity.ai/</a>).</p><p>The Chatbot would ask you to provide the topic that you would like to start learning. Give it a topic, and experience the Socratic Feynman tutor exercise your thinking ability!</p><p>I have been using this technique for over a month, and found this super amazing! Try it out and share your experiences.</p><p>Do you have your own way of learning things? I would love to hear more!</p><p>Article originally published on: <a href="https://write.as/laxmena/while-genai-have-the-potential-to-make-our-minds-dull-we-can-also-leverage">https://write.as/laxmena/while-genai-have-the-potential-to-make-our-minds-dull-we-can-also-leverage</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0738798f2869" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Over-Reliance on GenAI impacts Critical thinking; Here’s how I think we can fix it]]></title>
            <link>https://laxmena.medium.com/over-reliance-on-genai-impacts-critical-thinking-heres-how-i-think-we-can-fix-it-9600a11c031c?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/9600a11c031c</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Sat, 24 May 2025 18:46:21 GMT</pubDate>
            <atom:updated>2025-05-24T18:46:21.833Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Note: The article addresses to Software Engineers, but the ideas apply to knowledge workers in every domain.</em></p><p>I’ve spent considerable amount of time using GenAI past year at work. Also, having spent time with many power users, I begin to see an interesting trend.</p><p>Engineers are increasingly using GenAI to accomplish wide range of tasks, from advanced software engineering problems, to drafting a simple slack message.</p><p>The AI having digested the whole of internet, generates very reasonable responses(most often) — which increases the user’s confidence, and makes them rely more on the tool. Thus helping software engineers accomplish more in less time.</p><p>But here’s something that we fail to acknowledge: In the pursuit of increased productivity, we delegate some or most of the critical thinking to the GenAI.</p><p>In April 2025, Microsoft Research Labs published a paper — <a href="https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/">The Impact of Generative AI on Critical Thinking</a>, where they studied the impacts of using GenAI by knowledge workers.</p><p>Here’s a quote from the research paper:</p><blockquote>while GenAI can improve worker efficiency, it can inhibit critical engagement with work and can potentially lead to long-term overreliance on the tool and diminished skill for independent problem-solving.</blockquote><p>I understand the concerns raised by the researchers in the study, and find them fairly reasonable.</p><p>The engineers are increasingly reviewing code written by Generative AI, than writing their own. This is a <strong><em>significant mindset shift from “Problem Solving and Execution” to “Task stewardship and verification”.</em></strong></p><p>I strongly believe this track could seriously impede critical thinking amongst engineers in long term. We could become Armchair critics (Offering judgements and opinions without having enough involvement or experience), thus questioning the quality and usefulness of the code review.</p><p>After thoughtfully contemplating on this problem for a while, I came up with an approach to address this as a consumer. And I recommend every engineer to consider.</p><p>Moving forward, AI is inevitable in workplaces, and I don’t see future of workplace without them. There will be a push for productivity from the management, and AI has shown potential to double, triple or 10x the speed of an average engineer.</p><p>The challenge is to be more productive and not lose skill. Here’s what I think an engineer using AI should do:</p><p>**<strong>Go on an AI Detox for a week every 2 months</strong>**.</p><p><strong>Rules for the AI Detox week:</strong></p><p>1. No AI tools should be used<br>2. Any tools that existed during the Pre-ChatGPT era can be used<br>3. Engineer benchmarks their performance on various tasks/metrics without using AI<br> — Time it takes to complete a task<br> — Critical thinking and Problem solving abilities<br> — Skills check (Writing, Coding, Design, Reading, Communication, etc)<br> — Domain understanding (Strengths and Weakness)</p><p>At the end of the Detox week, perform a careful analysis on the benchmarks and observations during the week. Compare it against how they were performing with AI.</p><p>Use this information to carefully tackle and resolve any weaknesses found and bridge the gaps. AI tools can themselves be used to help achieve this.</p><p>(I’m working on another article about: how I use Gen AI/LLM tools to learn and think critically).</p><p>This approach strikes a right balance between both the worlds, achieving the productivity while UpSkilling and improving critical thinking at the same time.</p><p>Do share your thoughts and comments.</p><p>Article originally published here: <a href="https://laxmena.com/over-reliance-on-genai-impacts-critical-thinking-heres-how-i-think-we-need-to">https://laxmena.com/over-reliance-on-genai-impacts-critical-thinking-heres-how-i-think-we-need-to</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9600a11c031c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Deep dive into JavaScript Primitive Data-types and Auto-boxing — Nuts and Bolts series]]></title>
            <link>https://laxmena.medium.com/deep-dive-into-javascript-primitive-data-types-and-auto-boxing-nuts-and-bolts-series-de2a27ae2de8?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/de2a27ae2de8</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Wed, 18 Oct 2023 06:40:39 GMT</pubDate>
            <atom:updated>2023-10-18T06:49:35.223Z</atom:updated>
            <content:encoded><![CDATA[<h3>Deep dive into JavaScript Primitive Data-types and Auto-boxing — Nuts and Bolts series</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7o7gENCMIQJOokv5" /><figcaption>Photo by <a href="https://unsplash.com/@blakeconnally?utm_source=medium&amp;utm_medium=referral">Blake Connally</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>JavaScript primitive is a data-type which is <strong>not an object</strong>, and <strong>has no methods or properties</strong>. Primitives are the lowest level of JS implementation, and they cannot be boiled down further.</p><p>Primitive data types available in JavaScript:</p><ol><li>string</li><li>number</li><li>bigint (Added in ES11)</li><li>boolean</li><li>symbol (Added in ES6)</li><li>undefined</li><li>null</li></ol><p><strong>Note: </strong>These primitives are not to be confused with variables. Variables are containers that hold data, they are mutable. Whereas, primitives are immutable.</p><pre>var fruit = &quot;apple&quot;</pre><p>Here fruit is a variable, which points to the primitive string value — “apple”.</p><pre>var fruit = &quot;orange&quot;</pre><p>Variable fruit can be made to point to a different value like “orange” in this case, so they are mutable. But for primitives, the value cannot be modified.</p><p>“apple” remains the same, JavaScript doesn’t provide any tools to modify the value of the primitives.</p><p>Getting back to the other point that I mentioned previously: Primitives don’t have any methods.</p><p>Contradicting to my claim the following expressions seems to work amazingly well:</p><p><em>Case 1:</em></p><pre>&gt; &quot;Hello world&quot;.substring(6)<br><br>Output: &#39;world&#39;</pre><p><em>Case 2:</em></p><pre>&gt; (1).toExponential()<br><br>Output: &#39;1e+0&#39;</pre><p>When JavaScript primitives are not supposed to have any methods associated with them, how does <em>substring()</em> and <em>toExponential() </em>work?</p><h4>How does it work?</h4><p>It all has to do with what happens behind the scenes in JavaScript.</p><p>When JavaScript sees an expression with primitives, JavaScript does something called <strong>Auto-boxing. </strong>When it realizes that a method is being called on a primitive type, it tries to wrap this primitive object using inbuilt JavaScript classes to make JavaScript wrapper objects.</p><p><strong>Primitives vs Wrapper objects examples:</strong></p><p>Primitive type: <strong>string </strong>| JavaScript wrapper object: <strong>String<br></strong>Primitive type: <strong>number</strong> | JavaScript wrapper object: <strong>Number</strong></p><p>This wrapper object contains the definition of the methods that were invoked in the examples.</p><p>JS executes those methods and returns the results.</p><p>I felt this is a really interesting behavior of JavaScript.</p><p>Have you encountered this scenario before? Or do you have any experiences from your work along these lines, would love to hear them!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=de2a27ae2de8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Curious case of Addition(+) operator in JavaScript — Nuts and Bolts Series]]></title>
            <link>https://laxmena.medium.com/the-curious-case-of-addition-operator-in-javascript-nuts-and-bolts-series-c6e98a2629e7?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/c6e98a2629e7</guid>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[software]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Tue, 17 Oct 2023 04:54:27 GMT</pubDate>
            <atom:updated>2023-10-17T04:54:27.604Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Curious case of Addition(+) operator in JavaScript — Nuts and Bolts Series</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qreLRG3JVx1R8S47" /><figcaption>Photo by <a href="https://unsplash.com/@gamell?utm_source=medium&amp;utm_medium=referral">Joan Gamell</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Here’s a quiz, what would be the result of the following expression:</p><pre>var result = null + true + “hello”;</pre><p>At first glance, any sane Software engineer would think compiler would throw an error.</p><p>But this is JavaScript, and JS has its own quirkiness. JS not only does not throw an error for this expression, it also successfully evaluates and returns a response.</p><p>This sparked my curiosity, and I started looking deep into it. As I call it — go till the Nuts and Bolts of things. Lets dive deep!</p><p><strong>Something to note:</strong></p><p>In JavaScript, The (+) operator is overloaded for two distinct operations:</p><ol><li>Adds two numbers</li><li>Concatenates two strings</li></ol><p><strong>How does JS evlautes expressions with Addition(+) operator:</strong></p><p>When JS engine recognizes an expression with (+) operator, it first boils down the operands to its primitive types. More on JS primitives in the next post :)</p><p><em>Then the following rules are used to evaluate the final result:</em></p><ol><li><em>If any one operand is a string, other operands are converted to strings. then string concatenation is performed.</em></li><li><em>BigInt value can only be added with another BigInt. When used with other types, TypeError is thrown.</em></li><li><em>Otherwise both operands are converted into numbers.</em></li></ol><p><strong>Walking through step by step evaluation of the expression:</strong></p><p>Heres the expression again:</p><pre>var result = null + true + “hello”;</pre><p><strong>Step 1:</strong> <strong>null + true</strong></p><ul><li>Both operands are already in their primitive type.</li><li>They are <strong>not string or BigInt</strong>.</li><li>So as per rule 3, <strong>operand’s are converted to numbers.</strong></li><li>Number(null) + Number(true) =&gt; 0 + 1 =&gt; <strong>1</strong></li></ul><p><em>Note: Number(null), Number(false) resolve to 0.</em></p><p><strong>Step 2: 1 + “hello”</strong></p><ul><li><strong>One of the operand is a string.</strong> <strong>So</strong> <strong>other operand is converted to string</strong> using the toString() method.</li><li><strong>String concatenation</strong> is performed on the two strings.</li><li>(1).toString() + “hello” =&gt; “1” + “hello” =&gt; <strong>“1hello”</strong></li></ul><p>So the final result of the expression is: <strong>1hello</strong></p><p>Interesting isn’t it? Let me know your thoughts, and if you have faced these scenarios before, but have skipped past this!</p><p>Just for fun, I would recommend you to try and resolve the following expression, and guess what the result would be!</p><pre>var another_result = “hello” + undefined + 1</pre><p>What would be the result of <strong>another_result </strong>?</p><h4>Reference</h4><ul><li><a href="https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Addition">https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Addition</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c6e98a2629e7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Story of a **Hacky tool** which is used by over 130 million machines worldwide today!]]></title>
            <link>https://laxmena.medium.com/story-of-a-hacky-tool-which-is-used-by-over-130-million-machines-worldwide-today-ffde784858d1?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/ffde784858d1</guid>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Tue, 17 Oct 2023 00:58:00 GMT</pubDate>
            <atom:updated>2023-10-17T00:58:27.498Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XMZl7H7WtuluXu7u" /><figcaption>Photo by Fotis Fotopoulos on Unsplash</figcaption></figure><p>This is the story of a Simple hacky tool, which was put together in very less time, but has impacted the development experience of people worldwide! It’s one of the most under appreciated but amazingly powerful tool in almost all of our dev machines!</p><p><strong>Ken Thompson</strong>&#39;s boss at Bell Labs was facing a problem: <em>he needed a tool to search for patterns in files.</em></p><p>Thompson realized that he already had some work done on a tool that could do just that, even though it was a bit hacky and inefficient.</p><p>Thompson decided to polish up his existing tool. He made it easier to use and more efficient. He showed the program to his boss, who immediately started using it.</p><p>Very soon, everyone at Bell Labs was using Thompson&#39;s program to search for patterns in files.</p><p>Thompson eventually decided to release the program to the public, and he called it #Grep, which stands for &quot;Global Regular Expression Print.&quot;</p><p>Grep quickly became one of the most popular #Unix programs, and it is still widely used in millions of machines today.</p><p>This story never ceases to amaze me of how once a hacky program developed by one person is now used by thousands of people every day.</p><p>Small problems that we solve today can have great impact! Let’s keep building things and solving problems!</p><p>#SoftwareEngineering #Programming #SoftwareEngineer</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ffde784858d1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Effective Reading Techniques I wish I could teach my Younger Self]]></title>
            <link>https://laxmena.medium.com/effective-reading-techniques-i-wish-i-could-teach-my-younger-self-54f0555cebaa?source=rss-91dc7710e42------2</link>
            <guid isPermaLink="false">https://medium.com/p/54f0555cebaa</guid>
            <category><![CDATA[ideas]]></category>
            <category><![CDATA[books-and-authors]]></category>
            <category><![CDATA[books]]></category>
            <category><![CDATA[advice]]></category>
            <category><![CDATA[reading]]></category>
            <dc:creator><![CDATA[laxmena]]></dc:creator>
            <pubDate>Tue, 15 Nov 2022 06:17:50 GMT</pubDate>
            <atom:updated>2022-11-15T07:53:24.790Z</atom:updated>
            <content:encoded><![CDATA[<h3>Effective Reading Strategies I wish I could teach my Younger Self</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*I1Sw5T5PyGoRJCPsYHVjfg.jpeg" /></figure><p>Over the years, I learned, adapted, and improvised my reading techniques. The way I read now is completely different from how I used to read before.</p><p>If I could travel back in time, I will share these techniques and insights with my younger self to become an effective reader. These techniques can help you be one as well.</p><h3>Write, Scribble, and Dog-Ear the books</h3><p>Without any doubt, this statement would have extremely offended my younger self.</p><p>I used to consider books as prized possessions, marking and folding their pages are punishable offenses. Got disappointed even with the slightest damage. I even severed friendships for damaging my books, No kidding.</p><p>Several years forward, I now believe ‘Writing, Marking and Dog-earing’ a book is not a sin, but an act of showing true respect.</p><p>By Interacting with the books, it is no longer a one-way monologue. But a two-way communication between the author and the reader.</p><p>Here are my suggestions to interact with the book:</p><p><em>- Highlight the passages<br>- Note down thoughts and opinions next to the passages<br>- Question something that you don&#39;t believe in<br>- Dog-ear the pages so you can revisit your favorite parts later</em></p><p>I call this <strong>interactive reading</strong>. This can significantly improve the reading experience, and enable us to connect with the content in a more depth and meaningful way.</p><h3>Books-per-year metric is Nonsense</h3><p>I used to have <strong>Reading Goals. <em>Read X number of books in a calendar year.</em></strong> It used to be a driving force, and a way to compete with friends.</p><p>Now I strongly feel ‘Books-per-year’ is an inappropriate wrong metric to measure reading experience.</p><p>Though It pushed me to read more I seriously wish I hadn’t rushed through a lot of books, just for the sake of the count. Instead, I should have <strong>focused more on how the books improve thoughts and imagination.</strong></p><p>Mindlessly sprinting through the pages, defeats the actual purpose of reading.</p><p>Each book (fiction or non-fiction) is written to serve the reader with a unique experience. Savor and enjoy them, instead of forcefully swallowing them.</p><h3>You don’t have to complete all the books you start</h3><p>My Younger version had to complete every book I started, whether I liked it or not. <br>My Present version understands, There’s no Ironclad rule that states we need to finish every book we start.</p><p>All books are not meant for everyone.</p><p><strong>If a book doesn&#39;t work out, feel free to move on. You don&#39;t have to put yourself through pain to finish the book.</strong></p><p>An author may have contradictory ideas to you, but that should not be the sole reason to skip the book. That takes us to the next wisdom:</p><h3>No need to agree with everything a Book says</h3><p>This is the most important piece of information that I would like to give my younger self.</p><p>A reader doesn&#39;t have to agree with everything the author says.</p><p>A <a href="https://www.ytamizh.com/thirukural/kural-423/">Thirukural</a> couplet emphasizes this idea — <strong>“Whoever may present information, discerning its truth is real wisdom”</strong>.</p><p>We may not always align with the author&#39;s perspective, and it is perfectly fine. <br><strong>Only the ideas that make sense to our rational brains should be taken.</strong></p><h3>Read a book twice</h3><p>I will definitely recommend reading books twice and following this technique.</p><p><strong>Read interactively for the first time.</strong> Highlight passages and core ideas. Write down questions, thoughts, and opinions. Note topics that need further exploration.</p><p><strong>The second time, only go through the highlighted passages</strong> and writings. These highlights can be filtered and built <strong>to create valuable notes.</strong></p><p>These notes can be digital or physical. These notes are real treasures and immensely valuable ideas. Organize them in a way that the ideas are readily available and handy, so great value can be extracted from them.</p><p><strong><em>If you find this article useful and would love to read more of my ideas, consider Following me and subscribing to my Newsletter by </em></strong><a href="https://laxmena.com/posts/effective-reading-techniques?utm_source=medium-article#:~:text=my%20articles%20useful%2C-,subscribe,-to%20receive%20more"><strong><em>clicking here</em></strong></a><strong><em>!</em></strong></p><p>Originally Published on <a href="https://laxmena.com/posts/effective-reading-techniques?utm_source=medium-article">my Blog</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=54f0555cebaa" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>