<?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 Ankit Sharma on Medium]]></title>
        <description><![CDATA[Stories by Ankit Sharma on Medium]]></description>
        <link>https://medium.com/@ankitsharma61016?source=rss-a4769bef3762------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*OIgaZ27CrBt-fN36R5SXTw.png</url>
            <title>Stories by Ankit Sharma on Medium</title>
            <link>https://medium.com/@ankitsharma61016?source=rss-a4769bef3762------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:26:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ankitsharma61016/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 weight of a promotion isn’t more work]]></title>
            <link>https://medium.com/@ankitsharma61016/the-weight-of-a-promotion-isnt-more-work-a083a60ed742?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/a083a60ed742</guid>
            <category><![CDATA[self-improvement]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[self-awareness]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sun, 19 Jul 2026 09:08:32 GMT</pubDate>
            <atom:updated>2026-07-22T02:55:24.820Z</atom:updated>
            <content:encoded><![CDATA[<p>I joined Amazon for the first time as an SWE Intern in Summer of 2023. I was grateful and nervous about the new experiences, people I will interact with and work that I’ll do. I have a whole arc about it written here: <a href="https://medium.com/@ankitsharma61016/my-internship-experience-at-amazon-and-tips-for-other-swe-interns-6aa54e8a363a">My experience</a> and <a href="https://medium.com/@ankitsharma61016/part-2-flow-state-arc-leveling-up-398279c67814">No-PPO arc</a>. Like everybody else who joins Amazon, I scraped every public feed about work-life balance and work environment, and not to my surprise I had heard terrible things about it. Toxic managers, and back-stabbing colleagues and what not… and so I had to be cautious too and I was REALLY cautious with everything I said or did, initially.</p><p>But then something changed, I experienced what I didn’t expect. People who are unlike anything I read about. Ready to hangout after work, doing awesome fun stuff and being super zealous about life really! If you haven’t read my other blogs then know that I got super lucky. Some of my colleagues are now the closest friends that I know.</p><h4>Grateful for what I got…</h4><p>My job at Amazon made me eligible to hop on <strong><em>first flight of my life</em></strong>, <strong><em>first time trying variety of food items (Pancakes, Sushi etc.)</em></strong>, <strong><em>first time GoKarting</em></strong>, <strong><em>Karaoke night</em></strong>, <strong><em>concert, gifting my parents and brother </em></strong>and <strong><em>staying at a 5-star hotel</em></strong>. These might not seem like a big thing to you as a reader, but these were the things I used to think about as a kid and being eligible enough to do it all is a dream come true and I am grateful and this is all the more reason that I gotta keep doing the good work with dedication.</p><h4>I got promoted to SDE-2 this month</h4><p>This July, I got promoted to SDE-2 at Amazon. It would not have been possible without the support of people around me. I am lucky enough to have friends here to guide me for each unique work or life-related situation and having them around me in Bangalore, where I don’t have anybody else, makes my life a lot easier.</p><p>This articles lists what have I learned in these two years:</p><h3>Learnings</h3><h4>Everything you learn NOW has compounding returns.</h4><p>Notice that YOU CAN LEARN ANYTHING now, for the first time in history, faster than ever. You have access to the smartest thing Humans ever created. Anything you learn and do during age 21–30, has compounding returns. You can develop and apply skills that can change the course of your life and those around you… How empowering is that?</p><p>But there’s a catch nobody tells you about compounding: you have to actually start the loop. And the biggest thing stopping most people from starting isn’t laziness. It’s who they think they already are.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*gmphXQWtzbc6y-lF.jpg" /></figure><h4>Your past is not the contract for your current present to follow.</h4><p>This is the mindset that everybody should have! When your environment keeps changing then it’s not optimal to stick to your old self like a contract. Rediscover yourself at 1:57 am, or on a random Tuesday or whenever you realize that your identity is not serving your purpose or who you want to be. The point is to not wait for a perfect time: “<em>I’ll change next year</em>” or “<em>This is who I am…</em>” or “<em>I will change from next week</em>”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/712/0*cdiYA0mOAuW-y9A9.jpg" /></figure><p>Labels are really powerful, be very careful labeling your behaviors! <strong><em>Introvert</em></strong> is a label, and you using it as an excuse while avoiding conversation is limiting yourself to mere labels. YOU ARE NEUROPLASTIC! you become what you believe.</p><blockquote>Internalizing your weakness will hurt your vision more than your (perceived) lack of talent ever would!</blockquote><p>The power that you gain when you let go of the fear of change is tremendous. YOU ARE FREE TO DISCOVER YOURSELF EVERYDAY! I can be a completely new person tomorrow if I pretend to be ‘it’ long enough — You can read more about it in my article about <a href="https://medium.com/@ankitsharma61016/but-why-arent-we-working-backwards-from-the-perfect-version-of-self-2162a935f252">“working backwards from your ideal self”</a>.</p><p>And if your identity is up for renegotiation, so is the most fossilized belief you carry about adulthood: that work is the thing you survive and life is the thing you enjoy.</p><h4>It should be called Work-life symphony, not work-life balance</h4><p>I read it somewhere that I shouldn’t be chasing <em>work-life balance, </em>I should instead be chasing <em>work-life symphony</em>. work-life balance assumes that one component is fulfilling and second component (often work, if you haven’t got the hint…) is draining.</p><p>If you are spending 8 hrs per day doing something then it shouldn’t be draining. 8 hours is about 1/2 of your waking time, imagine spending half of your waking time for 40 years, doing something that you hate! That’s pathetic!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*6qY5O3bjqTtOf40I.jpg" /></figure><p>What I should be instead chasing is <em>work-life symphony</em>! Your work should be so fulfilling that you wake up excited to learn, implement and build the life that you want. This demands sacrifices.<strong><em> If you are lucky enough to find your purpose that won’t seem like sacrifices but a cost for the future that you can reach, double down on it</em></strong>. Another thing that this mindset shift does is that your life is not linear, it’s filled with uncertainty, sometimes you gotta spend more time at work and sometimes at life, and this idea makes it easier for you to allocate time.</p><p>Once you convince yourself that everything that you learn and apply, has compounding returns, you get more motivated to work on yourself.</p><p>Waking up excited to build sounds nice as a sentence. But excitement is a skill, and the best people I know at it are five years old.</p><h4>Create like a child, inspect like a scientist.</h4><p>A child doesn’t care what an adult thinks of their art. If a child gets an idea, they don’t wait for approval from an adult before grabbing a crayon and just start drawing… They have highest <em>Bias for action</em>. That’s what you have to cultivate internally, but only creating without retrospection is almost always like setting yourself up for disappointment… and thus the second part: “<em>Inspect like a scientist”. </em>Always retrospect on your creations. Your creations without iterations are just POCs(Proof of concepts).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*lXXStJXafdYt1gec.jpg" /></figure><p>Being childlike will likely reduce the time between your Idea and Action… and Action produces results… Being scientist like will analyze those results and produce better actions. You will have no choice but to improve, and as long as you are improving, you are guaranteed to win.</p><p>I had 10x’d my life since I stopped performing and started doing whatever the heck I want. Through my writing, I, now don’t want to gain audience, but just consolidate what I want to believe. This is my little corner on the internet after-all. I build stuff that I need for fun, not for pleasing anybody else, and when I solve my problem, I do it as if I will use it and that’s the win!</p><p>Inspecting like a scientist needs one piece of equipment though. Scientists keep lab notebooks. Mine is just a blank page at 1 am.</p><h4>Write more often</h4><p>Writing makes me calm down. The thoughts that bother you sometimes are just exaggeration. <strong><em>When in tension, just start writing</em></strong>. The benefit of writing is that you can only write one line at a time, this forces your brain to stop overthinking and focus on one thought at a time and then at the end you can retrospect your patterns and about 99% of time you will end up calling yourself stupid for making big fuss about small stuff.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*-iM9lYEa3qwWohXc.jpg" /></figure><p>Another pro of writing is that AI can read it if you want. AI is super powerful and it will get even better. Optimize your life, tell it about yourself so it can help you get better.</p><p>Writing is also how I caught my oldest lie. For my whole life I told myself everybody is my friend. The page disagreed.</p><h4>Prioritize YOUR people</h4><p>For about my whole life I have seen people as equal. <em>Mom, everybody is my friend… I gotta go on their B’day he is my friend… </em><strong><em>blah blah. </em></strong>This is BS. At any moment you can’t have more than 4–5 close friend circle. Prioritize those and guard your time from others. Your time is the only currency nobody can pay you back and every lost moment comes with a cost (monetary and opportunity)…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*usteHDldyEfFO4IO.jpg" /></figure><p>Make friends that you remember when something really good happens to you or something really bad happens to you. A certain friend changed how I see video calls. I used to hesitate to video call even my closest friends, because I was scared that I am being too much… the thing that is meant to be for you, will stay with you regardless of how you are or who you will be, <strong><em>be okay to let loose and be unapologetically yourself.</em></strong></p><p>I got my zeal of child like curiosity back. I believe with the people around me I am starting to forget what embarrassment is. This blocker had cost me more opportunities than I can count.</p><p>Guarding your circle raises an obvious question: how do you decide who gets in? For friendships I go by feel. For collaborations I learned the hard way to go by evidence.</p><h4>Commitment to collaborate without trial is setting you up for failure</h4><p>I have realized that collaboration should not be committed without trial. Try, spend some time with the person and ensure that they know you are evaluating and observe. Bad thing about Humans is that they can say something and act like something else… in short, they can lie.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/550/0*NrTm6Vl-9rJUtw4B.jpg" /></figure><p>It’s not that they are doing that on purpose, they want to do what they said they will, but THEY won’t because they are not yet ready. <strong><em>Ensure you stay with the people who do what they say they would.</em></strong></p><blockquote>Actions carry more weight than words.</blockquote><p>Reading that back, it sounds calculating. Evaluate people? Watch them before you commit? So let me say the other half, because without it that advice turns you into someone I wouldn’t want to work with.</p><h4>Be Kind</h4><p>Lose everything, but don’t lose your kindness. God sees everything! DO NOT HURT ANYBODY FOR YOUR GAINS. By hurt I don’t mean that you have to please everyone, ensure that you are communicating what you want clearly and are providing feedback. I used to believe that I was being kind by not communicating my expectations to the person whom I am working with, but it’s wrong. Having expectations but not communicating those expectations, brews hate. The other person can’t improve, because you are not telling them what they have to improve, while you don’t give them even a chance to redeem themselves or improve.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*EWMkyeizVQLLhsEE.jpg" /></figure><p>Which brings me back to the promotion, because everything above is now the job. People will hold expectations of me by default. The least I can do is what I just preached: communicate mine clearly, and meet theirs.</p><h4>… So, What changed after promotion?</h4><p>Nothing really… The weight of promotion is not the additional work, but the expectation that people will have with me by default.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/0*VlYvvBaCCgR4DoQa.jpg" /><figcaption>Metamorphosis</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a083a60ed742" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[…But why aren’t we working backwards from the perfect version of Self?]]></title>
            <link>https://medium.com/@ankitsharma61016/but-why-arent-we-working-backwards-from-the-perfect-version-of-self-2162a935f252?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/2162a935f252</guid>
            <category><![CDATA[neuroscience]]></category>
            <category><![CDATA[personal-development]]></category>
            <category><![CDATA[self-awareness]]></category>
            <category><![CDATA[self-improvement]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sat, 13 Jun 2026 07:48:49 GMT</pubDate>
            <atom:updated>2026-06-13T07:48:49.031Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>If repetition rewires the brain, the only question that matters is: rewire it toward what?</em></h4><p>If I wake up every morning and run, what exactly separates me from a runner? Not the title or how I feel about it. <strong><em>I run, therefore the word fits</em></strong>. The label is just a description of what I keep doing. So if labels are earned by repetition, then the person I want to become is not a destination I arrive at someday. It is a set of reps I either do today or I do not.</p><p>That sounds motivational. It is actually mechanical and it has a name.</p><h4><strong>You can change.</strong></h4><p>Neuroplasticity is the brain’s lifelong ability to reorganize itself by forming, modifying, and strengthening neural connections. The brain is not a fixed organ you were issued at birth. It rewires its pathways based on what you learn, what you experience, and most of all, what you repeat i.e. <strong>I become what I repeat.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*B-d_GYIdjrI07Yyn" /></figure><p><strong>Eileen Gu</strong>, the Olympic skier, said it better than any textbook:</p><blockquote><em>“You can control what you think, and you can control how you think, and therefore you can control who you are. Especially as a young person, with neuroplasticity on my side, I can literally become exactly who I wanna be. How cool is that? How empowering is that?”</em></blockquote><p>How cool is that, really. There’s a hidden trap that comes with it, If repetition is the tool, repetition is also the threat. If I keep <em>not-doing</em> the things I said I would do, I do not stay neutral. I actively rewire myself into a slacker. The brain does not care about your vision or your feelings. It tallies the reps. A repeated thought that talks you out of action carries almost the same weight as the inaction itself. That is the whole reason willpower exists: to win the gap between the thought and the rep.</p><p>So the brain is adaptable, and it is always recording. Which surfaces the only question worth asking.</p><h4><strong>Rewire toward what?</strong></h4><p><em>“You become what you repeat”</em> is true and it is also useless on its own. Repeat <em>what</em>? Toward <em>which</em> version? Most people answer this forwards. They look at who they are today, pick a slightly better version, and inch ahead. That is optimization, and it keeps you orbiting the person you already are.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qXkALZptDn4THWOmlYhENw.png" /></figure><p>I am a software engineer, so let me reach for the move I use every single day at work i.e. Working backwards.</p><blockquote><strong>You do not design a system by starting from the current code and hoping it improves. You start from the spec, the finished thing, the press release, and you work backwards to the first task.</strong></blockquote><p>The endpoint defines the path. Not the other way around. So do that with yourself. Write the document for the finished version of you. Be embarrassingly specific. Mine reads like this: Ankit pauses three seconds before he reacts. Ankit ships imperfect work in public without flinching. Ankit comes on camera and the discomfort fades with reps. Ankit protects his hours like they are non-renewable, because they are. That document is a target you reverse-engineer.</p><blockquote><em>Work backwards to choose the rep. Drill forwards to make it stick.</em></blockquote><p>Once the finished version is on paper, the gap between him and me stops being a vague anxiety and becomes a list. And a list has a first item. That first item is today’s rep. <strong><em>Neuroplasticity</em></strong> is just the reason the rep eventually holds.</p><h4><strong>The rubber meets the road at one rep</strong></h4><p><em>Bob Bowman</em> coached Michael Phelps. In a talk on the characteristics of champions, he says the most important goal an athlete has is the immediate one, not the medal or the season. The quality of a single turn, right now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dd_8X3LEW7l12Pf4" /></figure><p><em>Champions</em>, he argues, obsess over the controllable process and ignore the scoreboard. The scoreboard is an output. The turn is an input. You only get to touch inputs.</p><p>This is the forward half of the method, and it is why working backwards does not turn into another planning exercise. You work backwards exactly once, to find the rep. Then you stop analyzing and you drill that rep like an athlete who does not negotiate with the alarm clock.</p><p>What if I pretend I am a runner and run every single day? At some point the pretending becomes indistinguishable from the thing. The outcomes have no choice but to bend, because the external exists whether or not the internal has caught up yet. <strong>Action is more trustworthy than feeling.</strong> Show up as though the title already belongs to you, and the structure of what you do grows stronger than what you feel.</p><h4><strong>The part nobody warns you about</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HDpc2Ws0GYJ7gSPU" /></figure><p>Change has a price, and the price is people. I hesitated to reinvent myself for a long time. I kept wondering what the people who knew me would think if I shifted gears too fast. That fear stops just as many people.</p><p>Here is what I made peace with. You can decide, right now, to be more curious, or more disciplined, or more ambitious, and nobody gets to pull up your old grades and say you are not allowed. The people ready to grow with you will grow with you. The rest will drift, and it will not even come with a fight or a clean goodbye. One day there is simply nothing left to talk about, because your priorities stopped overlapping.</p><p>That is going to hurt. So you handle it the way you handle any expensive decision: <strong>before you start, write down the price, and agree to pay it.</strong> You reach the goal owning the cost you already accepted. That is how you arrive without regret. And you have to go immune to rejection, because people will dismiss you at first. If a dismissal is enough to stop you, you were already stopped. Social pressure is a real force. Everyone who has ever done anything ambitious had to learn to resist it.</p><h4><strong>Anxiety is just missing training data</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*84pckf_lpRXfPh0v1ab0Jg.png" /></figure><p>Here is the reframe that changed my life more than any other. I used to get anxious about new things. Then I started treating anxiety as a instinct rather than a verdict. My hypothesis: <strong>anxiety is your brain reporting that it lacks training data.</strong> You feel it about things you have not done enough of. The fix is not to think your way calm. The fix is to run at the uncomfortable thing and stay there long enough for neuroplasticity to log the reps and update the model.</p><p>Posting things that scare me is part of that program. Speaking my mind is part of that program. Writing this is part of that program. Every rep is one more labeled example telling my brain this is survivable, this is normal, this is who we are now.</p><h4><strong>So where do I actually start?</strong></h4><p>I do not have a universal answer. I can tell you what worked for me: write about yourself, a lot.</p><p>Your patterns. Your end-of-day reflection. What you could have done better. How your ideal life looks twenty years out, and how it looks right now. Who you care about most. What you want to avoid at all costs. What you are grateful for. The page is where the spec gets written.</p><p>Charlie Munger said the most successful people are “continuous learning machines” who go to bed every night a little wiser than they woke up. Diligence and raw intelligence lose to that compounding. <strong><em>So the floor is simple. Go to bed slightly smarter than you woke up.</em></strong></p><p>I journal in <strong>Notion</strong> now, and Notion has an MCP server, which means I can connect it directly to Claude. So the reflection stops being something I do alone. I hand the model my actual daily logs and ask the only question that matters: A<strong>re my daily actions taking me toward the spec, and if not, what is the smallest rep that course-corrects tomorrow?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/546/1*a2xJ3HsSC33PJ6A7FoY36w.png" /></figure><p>That is the full closed self-improving loop. You write the finished version of yourself. You log what you actually did. And you build a quiet system that reads both, every day, and tells you the one turn to fix next.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AY113YdphDxGB5mSxLylmQ.png" /></figure><blockquote>… this way I have built a hyper-personalized god, watching over your life from far away, whose only scripture is the person you swore you would become.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2162a935f252" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[LLM Council: I Asked Five AI Personalities to Tear Apart My App Idea. Here’s What They Found.]]></title>
            <link>https://ai.plainenglish.io/llm-council-i-asked-five-ai-personalities-to-tear-apart-my-app-idea-heres-what-they-found-9ea52e182d81?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/9ea52e182d81</guid>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[anthropic-claude]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Thu, 28 May 2026 18:51:00 GMT</pubDate>
            <atom:updated>2026-05-28T18:51:00.433Z</atom:updated>
            <content:encoded><![CDATA[<p>I was ready to build. I had a product concept I’d been circling for weeks, a dark anime lock screen wallpaper app for the gym-bro, grindset, 5AM-alarm demographic. I’d done the research. Market analysis. Monetization model. Flutter MVP plan. Two to three weeks to ship, I told myself.</p><p>I ran it through Claude. Got a thoughtful, coherent answer. Felt validated. Then I ran it through a council of five.</p><p>It found the product-killing flaw in three minutes. A flaw I’d documented myself, written off as “manageable,” and then genuinely forgotten to worry about. That’s what this article is about.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cTN3YYXrWwXIa9XLj4yfkQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZlR-MwdpcFOdPvnAk_Dovw.png" /></figure><h4>The Problem With Asking One AI</h4><p>You ask Claude (or GPT, or Gemini) a hard question and it gives you a good answer, which is well-reasonedm and coherent… but doesn’t it feels like it’s a syncophant’s answer to your question? It feels as if it uses “<em>Yes, You are right…</em>” or “<em>I’m sorry. I was wrong…</em>” too frequently, for a trusted reviewer. This is what I hate! AI designed towards avoiding conflict than as an expert.</p><p>What you actually want, when the decision is real and the cost of being wrong is high, is what you get from a good room of smart people who genuinely disagree with each other. One person hunting for the fatal flaw. One zooming out to the bigger picture. One who just wants to know what you’re doing Monday morning. And crucially: none of them deferring to whichever voice sounds most confident.</p><p>Andrej Karpathy, on a random Saturday afternoon in November 2025, built something that does exactly this.</p><h4>What Karpathy Actually Built?</h4><p>In his own words from the GitHub README: <a href="https://github.com/karpathy/llm-council">Link</a></p><blockquote>“This project was 99% vibe coded as a fun Saturday hack because I wanted to explore and evaluate a number of LLMs side by side in the process of reading books together with LLMs.”</blockquote><p>He called it llm-council. The idea is instead of asking one model your question, you assemble a council of multiple LLMs via OpenRouter, run them all in parallel, then have them <strong>peer-review each other&#39;s answers anonymously</strong> before a Chairman synthesizes the final response.</p><p>The three stages:</p><ol><li><strong>First opinions</strong> — every model answers independently, no coordination</li><li><strong>Anonymous peer review</strong> — each model sees all the responses labeled A through E (not by model name), and ranks them for accuracy and insight</li><li><strong>Chairman synthesis</strong> — one designated model gets everything and produces the final answer</li></ol><p>The anonymization is the key insight. If the models know <em>which</em> model gave <em>which</em> answer, they play favorites, deferring to GPT-4 over a smaller model regardless of what was actually said.</p><p>Karpathy built it as a Saturday hack and explicitly said he wasn’t going to maintain it. But he published it anyway. The internet did the rest.</p><h3>The Claude Code Adaptation</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6CdpUiy4ID0TLnwJ4Cbc0w.png" /></figure><p>A builder named <a href="https://github.com/tenfoldmarc/llm-council-skill">@tenfoldmarc</a> took Karpathy’s core insight and rebuilt it as a Claude Code skill. He used same methodology, but instead of different models, it uses five sub-agents with different <em>thinking styles</em> running inside Claude code.</p><p><strong><em>This is the version I installed. And it’s the version that found my flaw.</em></strong></p><p>The five advisors:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7cSLVysZahyG66DOgVAuxQ.png" /></figure><ol><li><strong>The Contrarian</strong> actively hunts for what will fail. Not pessimism — the friend who saves you from a bad deal by asking the questions you’ve been avoiding.</li><li><strong>The First Principles Thinker</strong> ignores the surface question and asks: what are we <em>actually</em> trying to solve here? Sometimes their most valuable output is: <em>you’re asking the wrong question entirely.</em></li><li><strong>The Expansionist</strong> looks for upside everyone else is missing. What could be bigger? What’s being undervalued? They don’t care about risk — that’s someone else’s job.</li><li><strong>The Outsider</strong> has zero context about you, your field, or your history. Responds purely to what’s in front of them. This is the most underrated advisor — experts develop blind spots, and the Outsider is the one who says “wait, what does that word even mean to someone who doesn’t already know this?”</li><li><strong>The Executor</strong> only cares about one thing: what do you do Monday morning? If an idea sounds brilliant but has no clear first step, the Executor will say so.</li></ol><p>After all five respond, five more peer reviewers evaluate the answers anonymously. Then a Chairman synthesizes everything, and crucially, <em>the Chairman can disagree with the majority</em> if one dissenter’s reasoning is stronger.</p><h4>What It Found on My App</h4><p>I gave the council my full brief: the app concept, the market research, the technical architecture, the monetization model, the iOS constraint I’d already documented.</p><p>Four of five advisors independently flagged the same thing.</p><p><strong>The iOS limitation is not manageable. It kills the core loop.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v3crzMcjDeNigPAsmB61ng.png" /></figure><p>My entire value proposition was “daily lock screen delivery”, you wake up, your lock screen has a new dark anime quote, you feel locked in. <em>But iOS doesn’t let apps set the lock screen programmatically</em>. Every single day, the user has to manually save the image and set it themselves. That’s 4–7 steps. Most users do it once, maybe twice, and then forget the app exists.</p><p>I <em>knew</em> this and previously I had talked myself into believing that it could still work. That was my builder brain speaking, which I mentioned <a href="https://medium.com/@ankitsharma61016/ankit-stop-waiting-start-ai-maxxing-e09a7e13b58c">here</a>.</p><p><strong>“Anime wallpaper” is a graveyard App Store category.</strong> Zedge has tens of millions of users. HiPaper, Walli, and a dozen others dominate organic search with hundreds of thousands of reviews. The App Store discoverability problem is harder than the product problem — and the council had barely touched it. And then this, from one reviewer:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*otPdynS8uhvTeoIVpjWoww.png" /></figure><blockquote>“The wallpapers are a commodity. Ankit-as-persona is not. The fastest growth path for this demographic is a builder-in-public narrative: a 22-year-old Amazon engineer building his escape. That story is shareable. The wallpapers alone are not.”</blockquote><p>It again points me to the same thing, I’ve been putting off — <strong><em>BUILD FOR DISTRIBUTION!</em></strong></p><h4>How to Install It (Takes 30 Seconds)</h4><p>Install the Claude Code skill:</p><pre>git clone https://github.com/tenfoldmarc/llm-council-skill ~/.claude/skills/llm-council</pre><p>Then just say <strong>“council this”</strong> or <strong>“pressure-test this”</strong> followed by your question. It scans your workspace for context files first — CLAUDE.md, memory folders, anything relevant to your question — so the advisors have real grounding, not just your raw prompt. Then it runs the full council, produces an HTML report you can open in a browser, and saves a full markdown transcript.</p><p>Takes about four minutes. Produces something you can actually act on.</p><p>If you want it scoped to a specific project instead of globally:</p><pre>git clone https://github.com/tenfoldmarc/llm-council-skill /your/project/.claude/skills/llm-council</pre><h4>When to Use It</h4><p>TLDR: High stakes questions with trade-offs and un-certainity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FdBn2rLD33-s9TUsh8o2AA.png" /></figure><h4>Conclusion</h4><p>I’ve been using AI as a mirror for a while now. Ask it something, get a response, feel understood. The council broke that habit for good. I am now using llm-council for high-stake decisions and applying my own judgement over it.</p><p><strong>Credit:</strong> <a href="https://github.com/karpathy/llm-council">Andrej Karpathy</a> for the original Saturday hack. <a href="https://github.com/tenfoldmarc/llm-council-skill">@tenfoldmarc</a> for the Claude Code adaptation. <a href="https://x.com/olelehmann">@olelehmann</a> for the inspiration chain.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9ea52e182d81" width="1" height="1" alt=""><hr><p><a href="https://ai.plainenglish.io/llm-council-i-asked-five-ai-personalities-to-tear-apart-my-app-idea-heres-what-they-found-9ea52e182d81">LLM Council: I Asked Five AI Personalities to Tear Apart My App Idea. Here’s What They Found.</a> was originally published in <a href="https://ai.plainenglish.io/">Artificial Intelligence in Plain English</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[You can only miss a place if it holds people you love]]></title>
            <link>https://medium.com/@ankitsharma61016/you-can-only-miss-a-place-if-it-holds-people-you-love-c153d1d1c68d?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/c153d1d1c68d</guid>
            <category><![CDATA[self]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[life-lessons]]></category>
            <category><![CDATA[self-awareness]]></category>
            <category><![CDATA[self-improvement]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Thu, 28 May 2026 12:48:01 GMT</pubDate>
            <atom:updated>2026-05-28T12:48:01.927Z</atom:updated>
            <content:encoded><![CDATA[<p>I went to Home for two weeks. I am guilty that with oceans full of love for my parents and my “Home”, I miss Bangalore and People I see every-single-day. <em>Is this because of the gap between my Home visits? Is this normal? I think not. </em>This is new. I have never felt this way.</p><h4>What Bangalore Built</h4><p>I have found my people. I didn’t know what that sentence meant until this year. I thought I did. But there’s a specific kind of night that teaches you what it actually means.</p><p>One time we spent whole night chatting in a restaurant during a farewell. We sat there till we were literally kicked out at 3 AM. That was the first time I felt as if I am back in college. Then I had multiple side quests, multiple night-out meetups at AnTeRa(a café in Bangalore) for Pancakes and outrageous combinations with coffee like <em>orange-coffee,</em> which I couldn’t drink, so I did what any normal person would do, <strong><em>that is put Maple syrup in the orange-coffee.</em></strong> One time we randomly decided to convene at one friend’s home, just chatting, singing and having good time till 2 AM.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*CZXV_1bC1SSU8ukKsOCu7w.png" /></figure><p>The things that stood out was that I said “No” to multiple meet-up at first and overtime I have realized what I would miss and then I stopped saying “No”. One person in particular would just auto-assume that I am in 😂. People inviting you to their flats, hosting you and accommodating you is so price-less and new for me. I feel so blessed to be surrounded by these people.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1023/1*foKw2pCisCJnSeKF-QKAGw.png" /></figure><p>We went to a karaoke nights, where you can’t watch the recordings back with sound on, because you sang so horribly; Concert where you don’t even know one song of the singer and have audacity to open lyrics and sing your lungs out. With the same people, you can just sit in a middle of tech-park, and just talk or play random games or in a party where DJ is blasting music and talk about tech. They’re people who match the frequency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yW9ZCmkxISA5YvVB6HH_vg.png" /><figcaption>The time when we went for Pizza @ Pizza 4Ps (you must try it if you are in BLR) on a MONDAYYY!</figcaption></figure><p>I have realized that I have to be more myself to find my people. Wearing masks is exhausting. My life’s has been so much better since then… It’s not like I was forcing myself to wear masks or if I have a hidden agenda, my natural tendency was like that. I was a people pleaser, I am still not there yet, but the change has initiated. Randomly. One day when I woke up. Quite literally. <strong><em>I just woke up different one day.</em></strong> My mind has been feeling ecstatic as it’s resting position. Everything — People, place and moments feel so magical to me right now.</p><p>I feel alive and human.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Tc1DBzEVj7G8ee8ilYKOQ.png" /></figure><blockquote>Right now, I’m like a prism of people around myself. I’ve acquired their music tastes, their habits, the way they talk, dress, and outlook towards life and I am glad I probably have the best evolving combination in my life right now.</blockquote><h4>What I took for granted for 20 years…</h4><p>I came back home at the time when temperature hits 46 Celcius casually, with lethal loo winds blowing during afternoon. My father goes to office in that heat. When I am home, he picks up sweets on his way back from office.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NIwlj4-pjqGq4cLZTMBcWg.png" /><figcaption>Kota, Rajasthan, India — Temperature in Celcius</figcaption></figure><p>He used to sometimes drive me to coaching (30 minutes each way) through summer afternoons that would make most people refuse to leave air conditioning. He wakes up at 5AM to pick me up from railway station, even though I said no multiple times. He always drops me to the railway station when I leave. He never mentioned any of it. None of it was ever presented as sacrifice. I think this is what they call <strong><em>un-conditional love</em></strong>? or perhaps it is the fact that my mother ensures she only cooks what I love eating every single day and during all the 12 meals (perhaps under-counting…) that I eat while I am at home. Doing this, in Kota heat, while I can barely feel like moving is… I don’t know what to even call that.</p><p>They are simple people, and I mean that as the highest compliment I know how to give. When you offer them comfort, they decline. When you try to repay even 0.01% of it, they look at you like you’ve said something strange.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dPeaYjrMuhem1JRHDz2HJA.png" /></figure><h4>Time moves slowly when I am at home…</h4><p>This time I made sure I spend time with them. This time however I forcefully took my parents out to dinner. They don’t eat out. Like, never. I had to convince them to come. These are people who spent their entire lives making sure I had everything I needed, and the idea of someone doing that for them didn’t quite look normal to them, especially my Father. With me, I bought my terrible habit of <em>BlinkIt (a 10-min delivery app) </em>and infected my Mom-dad with it. My Dad is a skeptic when it comes to these things, but then he also loves sweets. Noting that, <strong><em>Who would bring them ice-cream tub at 12 AM if not BlinkIt?</em></strong> — that is how I hooked my dad to <em>BlinkIt</em>.</p><p>My mom and I painted a bottle together. Pastel colors. It was my mom’s idea, though I had ordered paint few months ago, for her to have something NOT-electronic but she didn’t use it, and told that we will do it together, so that’s what we did this time. It felt goooooood… It felt really good to do something for them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MHyDsKY8aBs63vGx-_jrXg.png" /><figcaption>Might not look as impressive to you… but it’s a centerpiece of Living room!</figcaption></figure><p>Every time I visit home and leave it, my heart feel heavier. I see my parents aging, slowly becoming as if they are my own kids. Asking me stupidly-simple questions “<em>Ankit, How do I change music for this one reel?</em>” “<em>I can’t seem to find this button, Help!</em>” and I can’t shake off the feeling that these people spent their whole life nurturing me with richness and yet they are so willing to throw their life away (even still?) for me… <strong><em>what is it if not true love</em></strong>.</p><p>What used to feel like default, what I absorbed as background for twenty years, now looks like a bright light in a dark world. I can see it now because I’ve been away long enough to develop the contrast.</p><h4>The Guilt</h4><p>I can’t stop thinking about Bangalore really and I miss the people. With all the new experiences, place and food — everytime I explore something or do something new, I feel a quiet guilt “<em>Damn! I wish I could show it to my mom/dad</em>”!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vPfZ1L2xEjnlaCt1yuFkgw.png" /><figcaption>Purrsuit — first hardware project built over a weekend— <a href="https://medium.com/@mayankupadhyay398/a-phone-call-to-a-man-on-a-toilet-built-us-an-app-ca41eb456364">Read more</a></figcaption></figure><p>If I am going for a trip, that’s what I do exactly — Video call them &amp; If I am at a <strong><em>“not-so-parent-friendly”</em></strong> place then a simple “I’ll be out tonight! Sleep well”, and they reply back with “Send us the message when you get back safely”. <em>Like what??? Why everything that you do overflows with love!!?</em> Anyways, so I was saying that I am really missing Bangalore right now for the people but also for the weather. <strong><em>WHATTT do you mean that their peak temperature is lower than the lowest temperature in Kota.</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3KsnfcaXVCX4Xyn8ytHTog.png" /></figure><h4>Conclusion</h4><p><em>You can only miss a place if it holds people you love.</em> For two years of staying away from my home, coming back to ‘<strong><em>Home</em></strong>’ used to feel like little islands of joy in the sea full of unknown people and experiences. The fact that Bangalore now pulls at me the same way is probably… <strong>Expansion</strong>.</p><p>The fact that I don’t have any photos of <strong><em>just us</em></strong> together just proves that… I should probably click more photos of HUMANS! Stupid me.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c153d1d1c68d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Will Either Win or Evolve till I win! ]]></title>
            <link>https://medium.com/@ankitsharma61016/i-will-either-win-or-evolve-c2e186aa1f98?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2e186aa1f98</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[personal-development]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Mon, 04 May 2026 21:20:58 GMT</pubDate>
            <atom:updated>2026-05-09T20:10:47.861Z</atom:updated>
            <content:encoded><![CDATA[<h3>I Will Either Win or Evolve till I win!</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_BWZGKhuusuTQBA4MXzJmg.png" /></figure><p>I made a decision sometime this year that quietly changed everything.<br>Not a dramatic one. No journal entry. No vision board. Just a reframe that settled somewhere deep and refused to leave.<br>Every outcome is either a win or data. If I succeed, then great. If I fail, I upgrade and come back with better information. The only actual loss is the one where nothing changes. Where I fail <em>and</em> learn nothing <em>and</em> show up the same way next time.That’s the outcome I’m afraid of. Not failure. Stagnation.</p><p>This is what April taught me.</p><h3>I ran an experiment. On myself.</h3><p>At the start of April, I made a rule: say yes to almost every outdoor plan.<br>Not because I’m an extrovert. Not because I was bored. But because I realized<strong> I’d been optimizing my life based on <em>assumed</em> preferences</strong>. Things I thought I knew about myself without ever actually testing them.<br>That’s bad engineering. You don’t assume how a system behaves. You instrument it. You run tests. You look at the output.</p><p>So I ran the test.</p><p>Three concerts. Multiple outings. Uncomfortable situations, unfamiliar crowds, plans that pulled me away from my room and my thoughts. <br>I don’t love concerts as much as I imagined I might. There’s a word for what I felt <strong><em>Monachopsis, </em>the subtle, persistent feeling of being out of place.</strong> Like everyone around you is fluent in a language you never quite learned. I felt it at two of the three.</p><p>But the one concert I actually enjoyed? It came down to two things at once, an artist I genuinely loved, <em>and</em> the right people around me. Strip either one away and it’s just noise. Together, they made it something real.</p><p>That’s a framework, not just an observation. The experience matters less than <em>who you’re in it with</em> and <em>whether it connects to something you already care about.</em></p><p>I also confirmed something I’d suspected but never verified: I come alive in conversation. Not crowds. Give me a table, someone who’s genuinely thinking, and 2am — that’s where I want to be. I would never have known this so clearly without April.</p><p>The experiment worked. Not because every experience was good.<strong> Because I now <em>know</em> — not assume — what I am.</strong></p><h3>The things that broke open</h3><p>Running experiments on your external life eventually forces you inward. April didn’t just teach me about concerts. It cracked open a few things I’d been avoiding.</p><p><strong>Silence isn’t neutral.<br></strong>I work with engineers. Some of them do things the wrong way. For a long time, I watched it happen and said nothing — a people-pleasing instinct dressed up as “not wanting conflict.”<br>I stopped doing that.<br> staying silent when you know better isn’t politeness. It’s cowardice wearing a polite mask. Letting someone do the wrong thing — when you have the knowledge to correct it — is a debt to the people around you. This sounds narcissistic, but it is what I feel.<br>I spoke up. It was uncomfortable. It was right. Plan is to do it more often.</p><p><strong>Your time is the only thing you can’t get back.<br></strong>I started noticing how casually I was giving it away. To people who didn’t value it. To conversations that led nowhere. To plans made out of obligation. I started saying no. Not aggressively. Just clearly.</p><p>Then something unexpected happened — I found people who <em>also</em> treated their time as sacred. Who showed up with conviction. Who spent a whole day building something with me and meant every hour of it.</p><p>That changed me. When you’re around people who don’t waste time, you stop wasting yours. Mind you, I am not saying spending time with people doing random shiz is “Time waste”, I consider it a time-waste, if it drains my energy. If I am coming home exhausted instead of excited about telling my parents about my day, then it’s “Time waste”.</p><p><strong>Mediocrity is a choice, not a circumstance.<br></strong>I have a fear of being mediocre that borders on obsession. Some people find that arrogant.Wasting the potential I’ve been given doesn’t feel like a personal failure. It feels like a disrespect — to the opportunities I’ve had, to the people who believed in me, to the simple fact that I <em>can</em> do more than average. That comes with a responsibility.</p><p>I will not be mediocre. Not because the world is watching. Because I am.</p><p><strong>The missing wheel.<br></strong>High agency isn’t one skill. It’s three:<br><strong>Clear thinking. Bias to action. Disagreeability.<br></strong>For a long time, my tricycle had a missing wheel. I had the clear thinking. I had the disagreeability — I’ve always been willing to hold an unpopular position if I believed in it. But bias to action? That wheel was loose.<br>I could see the destination. I could argue for it. I just wouldn’t <em>move.<br></em>Ideas stayed ideas. Potential stayed potential. The gap between knowing and doing was embarrassingly wide — and I watched myself maintain that gap for longer than I’d like to admit.</p><p>Then last week I woke up different.</p><p>Not because of one thing. The April experiments had been stacking up — data point after data point, uncomfortable moment after uncomfortable moment — until something broke open. The old model couldn’t hold. A new one came through.</p><p>The third wheel clicked into place. I don’t know if it’ll stay there. But I know what it feels like now. And I’m not going back.</p><p><strong>I used to talk about building things. I never built them.<br></strong>This one is specifically for builders. If you’re not one, skip it. Actually — don’t. You probably need it most.<br>If you want to build, you should be building. Not planning. Not architecting. Not telling people about the thing you’re going to build someday.</p><p>Building.</p><p>There’s a particular kind of person — I was this person — who is fluent in the <em>language</em> of building without doing the actual work. They know the frameworks. They have the ideas. They can describe the product in precise detail, map the tech stack, explain the market gap. And then they open a new note and start planning the next thing.</p><p>That’s not building. That’s performing building to yourself.</p><p>Here’s the question that broke the pattern for me:</p><p><em>If you want to be a builder and you’re not building — why aren’t you? Because that’s what builders do. They build. And if the answer is “I might fail” or “I’m not sure it’ll work” — that’s not a reason. That’s the entire point.</em></p><p>Every meaningful thing you could build has a real probability of failing miserably. Not a small one. A real one. The market might not want it. The tech might break. You might build entirely the wrong thing.<br>Accepting this isn’t pessimism. It’s the entry fee.<br>Most people find a way to never pay it. They stay in the planning phase indefinitely. They call it “validating the idea” or “doing research” or “waiting for the right time.” What they’re actually doing is protecting themselves from a very specific kind of pain — trying your best at something real and watching it not work.</p><p>That pain is real. It’s also the only path.</p><p>There’s something you’re afraid to ship right now. That’s exactly what you should be shipping. Everything you want as a builder is on the other side of that specific fear — not a general fear, not someday’s fear. That one. Today.</p><blockquote>The builders who win aren’t the ones with the best ideas. They’re the ones willing to be wrong in public, iterate fast, and stay in the game long enough for something to stick.</blockquote><p>Plans should be measured in decades. Execution should be measured in weeks.<br>Stop describing what you’re going to build. Go build the embarrassing first version. Ship it before it’s ready. That’s the job.</p><p><strong>You are what you consume.</strong></p><p>Not metaphorically. Literally.</p><p>Negative reels. People who drain. Content that makes you feel like you’re moving when you’re actually just scrolling. None of this feels dangerous because it’s not dramatic. Nobody ever felt one reel destroy their ambition. It’s death by a thousand small cuts.</p><p>But here’s what I started noticing: when I’m around people who complain, I start complaining. When I’m consuming content built to trigger and outrage, I feel vaguely agitated and unproductive for hours without knowing why. When I spend time with people who aren’t serious about their time, I stop being serious about mine.</p><p>Your inputs shape your outputs. Always. Whether you’re tracking it or not.</p><p>Eliminate the negative inputs. Not out of judgment — out of self-preservation. Optimize your life for the goals you actually want, not for the comfort of people who don’t share them.</p><p><strong>Ragebait. Aimed at myself and others around me!<br></strong>This one sounds strange. It works.<br>When I want to put something off, I ask myself one question: <em>“If you do that, you’d be repeating the same old pattern. But you wanted to change. So do it.”<br></em>The anger at the old version of me is more motivating than any quote I’ve ever read. I’m not running toward something abstract. I’m running <em>away</em> from someone I refuse to be anymore.</p><h3>The Purrsuit</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*01H7ve3v9OaytXdeV61UHg.jpeg" /></figure><p>Then came the rover. No plan. No brief. Just four engineers, a problem, and less than 24 hours.</p><p>We built a working rover.</p><p>But that’s not the part that mattered. The part that mattered was <em>how.</em> We laughed through the hard parts. Someone took charge when it was needed, without being asked! <strong>I love when everybody in a group is narcissitic! That’s a differnt energy all together!</strong></p><p>They showed up the whole day. With conviction. Like what we were building <em>mattered</em> — even though it was a prototype, even though nothing was at stake except pride and the joy of making something real out of nothing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*77ZNJfkqzdUIj3JcOJubzg.jpeg" /><figcaption>Momo!</figcaption></figure><p>I work best in the right container. Put me with ambitious, execution-focused people and I force-multiply. Put me anywhere else and I keep my spark — but it costs something to maintain it. These people don’t cost me energy. They add to it. They made me want to be more serious about my time, not less. That’s rare. I woke up in mornings of long-weekend with conviction and excitement — I know I will win with this team.</p><p>We’re building together now. Apps, ideas, things that don’t exist yet. We’re figuring out the marketing piece. But we’re good engineers with the hunger to execute.</p><h3>The belief I keep coming back to</h3><p>For decades, sports scientists believed running a mile in under 4 minutes was physically impossible. A hard biological ceiling. Experts said so. Data said so. Everyone accepted it.<br>Then on May 6, 1954, Roger Bannister broke it.<br>Within the next few years, multiple runners broke it.</p><p>The barrier was never physical. It was a model — a story people told themselves about what was possible. Once one human proved it wrong, everyone else’s model updated. Immediately. <br>Most “impossible” things work this way. They’re not physics problems. They’re belief problems.</p><blockquote>If your problem doesn’t defy the laws of physics, it’s not unsolvable — regardless of what anyone tells you.</blockquote><p>The moment you stop accepting other people’s ceilings as your own, the problem usually turns out to be solvable. Sometimes embarrassingly so.</p><h3>This year</h3><p>I said yes to April to collect data. I got it.<br>I know what I am now. What I like. Who I work best with. Where I come alive. What I refuse to be.<br>The April experiments stacked up — one after another — until something broke open. The old model gave way. A new one came through.<br>I will either win or evolve.<br>I’ll speak when I know I’m right, even when my voice shakes.<br>I’ll respect my time enough to say no clearly and without guilt.<br>I’ll ship things — not describe them, not plan them. Ship them.<br>I’ll build with people who treat their time as sacred.<br>I’ll treat every failed experiment as a successful data point.<br>And I will not be mediocre. Not because the world is watching.</p><p><strong>Because I am.</strong></p><p>This is my year.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2e186aa1f98" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ankit, stop waiting.
Start AI-Maxxing.]]></title>
            <link>https://medium.com/@ankitsharma61016/ankit-stop-waiting-start-ai-maxxing-e09a7e13b58c?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/e09a7e13b58c</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sat, 18 Apr 2026 13:43:30 GMT</pubDate>
            <atom:updated>2026-04-18T13:43:30.772Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>A public letter to myself</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ExXQ7_QqgJUOUkD8oapCMg.png" /><figcaption>A visual metaphor for the space between ideas and shipping</figcaption></figure><p>Yes, AI will take jobs. You know this. The question isn’t whether — it’s whether you’re going to be the person who saw it coming and still did nothing.</p><p>The doomers aren’t entirely wrong. But they’re missing the larger point: this is the single biggest opportunity equalizer in human history. Right now, today, you have access to the same tools as the billionaire in New York. AI, a laptop, an internet connection.</p><blockquote><strong>The gap between you and the most resourced people on the planet has never been smaller. What’s your excuse going to be?</strong></blockquote><p>Here’s the thing you keep telling yourself you’ll get to:</p><ol><li><strong>“I don’t have the knowledge to build things.”<br></strong>Use Claude. You can learn anything in the world — from no-code tools to system design to writing copy — through conversation. Your knowledge gap is now a 10-minute problem.</li><li><strong>“I don’t know how to sell what I build.”<br></strong>Distribution is the game. Build in public. Create content around your process. The product and the audience can grow together — AI helps you do both faster than ever.</li><li><strong>“I don’t have enough time.”<br></strong>AI is your leverage. What used to take a team of five takes you a weekend now. You’re not constrained by hours anymore. You’re constrained by how long you keep saying this.</li><li><strong>“I don’t know where to start.”<br></strong>That one’s on you, Ankit. Action is the only thing AI can’t do for you. Everything else is solved.</li></ol><p>The tools are equal. The opportunity is now. The only variable left is <em>whether you move.</em></p><h3>Part 2: How do you find ideas worth building?</h3><p>The framework is simple: take two things that don’t obviously belong together, and build the thing that lives at their intersection.</p><p>Music × tech. Finance × gaming. Logistics × social. Most good products aren’t invented from scratch — they’re transplanted. AI is unusually good at spotting these intersections, surfacing patterns across domains that no single human reads deeply enough to connect. But knowing the intersection exists isn’t the same as knowing whether it’s worth building.</p><p>That gap — between “AI says this is an opportunity” and “this is actually worth my time” — is where <em>you</em> come in. Your judgment, your taste, your read on whether real people would pay for this. AI handles the surface area. You make the call.</p><blockquote><strong>AI can consume data at scale. Reddit threads, market signals, tool gaps, user complaints. Feed it the noise. Your job is to find the signal.</strong></blockquote><p>Practically: before you build anything, use AI to run demand research. Scrape a niche on Reddit. Ask it to synthesize what people are paying for versus what they’re complaining is broken. Then gut-check it against your own read. The research takes minutes. The call still takes you.</p><blockquote><strong>Honest confession for one of my projects!<br></strong>You got Polar.sh working. Payments, sandbox, the whole setup. And then you stopped. Why? Because going to production means using real card details. That’s not a technical problem. That’s fear with a technical costume on. It’s costing you real momentum and you know it.</blockquote><p>This is worth naming because it’s a pattern. The last 10% of shipping has nothing to do with code. It’s the moment you stop treating your project like an experiment and start treating it like a business. <strong>You’re stuck in that 10%. Get out.</strong></p><h3>Part 3: You have edges you’re not using.</h3><p>You’re not starting from zero. Stop acting like it.</p><ol><li><strong>Unlimited resources: </strong>Free compute, Amazon-scale context, and a job that expects you to build. You have no bottleneck on experimentation — only on deciding to start.</li><li><strong>Human judgement: </strong>AI surfaces everything. You filter it. The taste layer — what’s worth building, who it’s for, why now — is still entirely yours.</li><li><strong>Time</strong> <strong>— and this one is underrated</strong>: AI is so new that genuinely novel ideas are being had for the first time in human history, every single day. Think about that. <strong>You could be the first person alive to connect two dots in a way no one ever has.</strong> Experience in the industry still matters — but obsession can beat experience when the experienced people are buried in meetings and you have free evenings. They have responsibilities. You have bandwidth. That asymmetry is a weapon if you use it.</li><li><strong>Adaptability — the real one: </strong>Multiple people across Amazon and Prime Video have said it unprompted: you moved first on AI. In an org of thousands, you’re the frontier runner. You don’t wait to fully understand something before using it — you learn mid-flight. By the time others are onboarding, you’re three iterations ahead. <br>There’s a character in Jujutsu Kaisen — Mahoraga — who adapts to any attack after seeing it once. Never loses to the same thing twice. That’s the edge. Not that you’re the smartest in the room. It’s that the room keeps changing and you keep being ready.</li></ol><h3>Part 4: So why haven’t you made any money, Ankit?</h3><p>Here’s what’s actually true: you are an idea machine with an execution debt.</p><p>You get genuinely world-changing ideas constantly. People know you for it. And not one of them has made you a single rupee. Every new idea arrives more exciting than the last, and the last one never quite reached the “someone paid me” moment before you moved on. You have a growing graveyard of interesting ideas, each abandoned at roughly the same stage — right before it gets uncomfortable.</p><blockquote><strong>The ideas aren’t the asset. The ideas are the distraction.</strong></blockquote><p>The person who takes your second-best idea and actually charges someone for it has beaten you. Every time.</p><p>And here’s the specific trap with “world-changing” ideas: the bigger the idea, the easier it is to never ship. A world-changing idea has infinite scope to hide behind — it’s never ready, never small enough to test, never the right moment. A small, specific, slightly boring idea forces you to ship it by Thursday or admit you won’t.</p><blockquote>Somewhere right now, someone who can’t write a for-loop is automating workflows, selling the output, and making real money. <strong>That should bother you.</strong> Not because they’re doing something wrong — but because they’re doing something you could do in a weekend, and haven’t. The difference isn’t skill. It’s that they didn’t wait until the idea was perfect.</blockquote><p>You don’t need a mentor. You don’t need more ideas. You need a forcing function — and you need to understand the real gap:</p><blockquote>You have builder brain. You don’t yet have founder brain. <em>Builder brain asks: “How do I make this work technically?” Founder brain asks: “Who is this for, will they pay, and what’s the smallest version I can sell tomorrow?” The Polar.sh sandbox moment is textbook builder brain — you solved the interesting problem and lost energy the moment it became a business problem.</em></blockquote><p>The line between project and business is terrifyingly thin. It’s one real person. One real payment. Even ₹99. That’s it. You’ve been engineering around that line instead of crossing it.</p><p>Your floor is a non-technical person’s ceiling. You have the compute, the judgment, the adaptability, the ideas, and the time. <strong><em>The only thing missing is the Tuesday version — the embarrassingly small thing you could put in front of someone this week and ask them to pay for.</em></strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e09a7e13b58c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5. Dining Savage Problem]]></title>
            <link>https://medium.com/@ankitsharma61016/5-dining-savage-problem-2eb7335195ef?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/2eb7335195ef</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[java]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sun, 15 Mar 2026 13:57:13 GMT</pubDate>
            <atom:updated>2026-03-15T13:57:13.777Z</atom:updated>
            <content:encoded><![CDATA[<p>A tribe of savages eats communal dinners from a large pot that can hold <strong><em>M servings</em></strong> of stewed missionary. When a savage wants to eat, he helps himself from the pot, unless it is empty. If the pot is empty, the savage wakes up the cook and then waits until the cook has refilled the pot.</p><p>Any number of savage threads run the following code:</p><pre>// Unsynchronized savage code<br>1 while True:<br>2     getServingFromPot()<br>3     eat()</pre><p>And one cook thread runs this code:</p><pre>// Unsynchronized cook code<br>1 while True:<br>2     putServingsInPot(M)</pre><p>The synchronization constraints are:</p><ol><li>Savages cannot invoke getServingFromPot if the pot is empty.</li><li>The cook can invoke putServingsInPot only if the pot is empty.</li></ol><h4>Solution Approach #1:</h4><p>To me it seems like the sequence of operation would be something like:<br>SSSSSSS…SSS(M times) C SSSS…SSSS(M times)</p><p>This resembles light-switch pattern that we talked about in previous posts. Incoming Savage turns on the light, and outgoing savage turns off the light, for Cook to use.</p><pre>Servings = M // A variable to keep track of left servings<br>mutex = Semaphore(1) // mutex for safe operation on Servings variable<br>isFilled = Semaphore(1) // implies that the pot is not empty.<br>isEmpty = Semaphore(0)</pre><p>Here is the Cook and Savage code (My solution):</p><pre>// Cook<br>isEmpty.acquire()<br>putServingsInPot(M)<br>servings = M<br>isFilled.release()</pre><pre>// Savage<br>mutex.acquire()<br>isFilled.acquire() // ensures that Savage doesn&#39;t move forward if pot is empty<br>getServingsFromPot()<br>servings--;<br>if (serving == 0):<br>    isEmpty.signal(); // signal that the pot is empty and don&#39;t release isFilled semaphore.<br>else:<br>    isFilled.release(); // release isFilled, so other savages can use it. <br>mutex.release()</pre><p>Here is the book’s solution:</p><pre>// Cook<br>1 while True:<br>2     emptyPot.wait()<br>3     putServingsInPot(<br>4     emptyPot.signal()<br><br><br><br>// Savage<br>1 while True:<br>2     mutex.wait()<br>3     if servings == 0:<br>4       emptyPot.signal()<br>5       fullPot.wait()<br>6       servings = M<br>7     servings -= 1<br>8     getServingFromPot()<br>9     mutex.signal()<br>10<br>11 eat()</pre><p>However I think the book’s solution is slightly better because it avoids forcing every thread to process an unnecessary semaphore (i.e. isFilled)</p><h3>Conclusion</h3><p>I am getting better at this!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/907/0*E1-feqbEO794f2Ww" /><figcaption>Mahoraga: Its defining ability is to adapt to “any and all phenomena”</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2eb7335195ef" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Fixxate: I built a distraction-free speed-reading app for the Infinite Web]]></title>
            <link>https://medium.com/@ankitsharma61016/fixxate-i-built-a-distraction-free-speed-reading-app-for-the-infinite-web-1b877ccdc37e?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b877ccdc37e</guid>
            <category><![CDATA[apps]]></category>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[javascript]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sat, 14 Mar 2026 20:11:58 GMT</pubDate>
            <atom:updated>2026-04-27T08:39:11.522Z</atom:updated>
            <content:encoded><![CDATA[<h4>How Rapid Serial Visual Presentation works, why most reading tools miss the point, and how a weekend project turned into a full reading system built for the age of endless content.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*u9UATPZA3nDBZy9rgYfkQw.png" /><figcaption>Fixxate</figcaption></figure><blockquote>Try it: <a href="https://fixxate.vercel.app/"><strong>https://fixxate.com/</strong></a><strong><br></strong>Contribute/Raise bugs<strong>: </strong><a href="https://github.com/CrypticMessenger/Fixxate"><strong>https://github.com/CrypticMessenger/Fixxate</strong></a></blockquote><p>The internet produces more words every day than any human could read in a lifetime. Not just more articles — more good articles, more essential threads, more long-form essays from people whose thinking you actually want inside your head. The problem isn’t finding things to read. It’s that reading hasn’t kept up.</p><p>Fixxate is my attempt to close that gap. It’s an RSVP reader — a tool that shows you one word at a time, in a fixed position on screen, so your eyes never have to move. You drop in an EPUB, a PDF, a Markdown file, or paste any URL, and it reads to you at whatever speed you can handle. No ads, no tracking, no algorithmic feed deciding what comes next. Just you and the text.</p><p>This post covers the science behind why RSVP works, a full walkthrough of what Fixxate can do today, a technical section for the builders in the room, and where the product is headed.</p><blockquote>“Your eyes waste 80% of reading time just moving between words. RSVP gives that time back.”</blockquote><h3>1. The science: why you read slower than you should</h3><p>When you read a page of text normally, your eyes don’t glide smoothly across the line. They jump — short, rapid movements called saccades — landing on each word for roughly a tenth to a quarter of a second before jumping again. Research suggests that only about 20% of your reading time is spent actually processing the words. The other 80% is eye movement.</p><p>Rapid Serial Visual Presentation (RSVP) eliminates saccades entirely. Instead of moving your eyes across a line, words come to you — one at a time, in a fixed position on screen. Your eyes stay perfectly still. The result: an average reader who manages 250 words per minute on a printed page can comfortably hit 400+ wpm with RSVP after very little practice, and with training, 600–800 wpm is achievable with reasonable comprehension.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f41CrPeCI9jGzGEKPqRpmw.png" /><figcaption>Introduction slide#1: For first time users</figcaption></figure><h4>1.1 The Optimal Recognition Point</h4><p>The detail that separates a well-built RSVP reader from a bad one is the Optimal Recognition Point (ORP). Research into eye-tracking during reading shows that the eye doesn’t land at the centre of a word — it lands slightly to the left of centre. This is the point at which the brain most efficiently recognises the full word.</p><p>In Fixxate, every word is aligned so its ORP sits at a fixed horizontal position in the word frame. The ORP letter is highlighted in orange-red — so your eye always knows exactly where to fixate. For the word “reading”, the ORP falls on the letter “a”. For “the”, it falls on “h”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CVCgp_ViI7HwoLjjEXE6bg.png" /><figcaption>Introduction Slide#2 : ORP (Optical recognition Point)</figcaption></figure><h4>1.2 Smart pausing</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X9skLZA0slJSd7nMLfWTvA.png" /><figcaption>The UI</figcaption></figure><p>One of the most comprehension-killing mistakes in naive RSVP implementations is showing every word for exactly the same duration. A comma means nothing if you blast past it at the same speed as a normal word. Fixxate uses a variable delay system:</p><ul><li><strong>End of sentence (. ! ?):</strong> 2.2× base delay</li><li><strong>Comma, semicolon, colon:</strong> 1.5× base delay</li><li><strong>Long words (12+ characters):</strong> 1.5× base delay</li><li><strong>Medium words (8–12 characters):</strong> 1.2× base delay</li></ul><p>These multipliers compound with your WPM setting. At 300 wpm the base delay per word is 200ms. A sentence-ending period extends that word’s display to 440ms — enough for your brain to register the boundary without consciously noticing the pause.</p><h3>2. What Fixxate can do today</h3><p>The app shipped with more features than I originally planned because each one turned out to be simple once the core engine was solid. Here’s the current feature set:</p><h4>2.1 File formats</h4><ol><li><strong>EPUB — </strong>Full book support — spine items parsed in order, metadata extracted, chapter structure preserved.</li><li><strong>PDF — </strong>Text extracted page by page via PDF.js (Mozilla’s own library), loaded on demand.</li><li><strong>Markdown — </strong>Parsed in-browser — headings, links, bold/italic, code blocks, and list markers all stripped to clean reading text.</li><li><strong>URLs — </strong>Paste any article URL. Fixxate fetches it via a proxy chain, then runs Mozilla Readability (the same engine that powers Firefox Reader Mode) to extract just the article body.</li></ol><h4>2.2 The curate screen</h4><p>Web articles are messy even after Readability processes them. Before reading anything fetched from a URL, Fixxate shows you a paragraph-by-paragraph view of the extracted content. Each paragraph is a card you can dismiss with a single click. A “smart trim” button auto-removes the opening teaser, trailing CTAs, and any paragraph under 35 words (usually pull quotes or image captions that made it through).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q1c1rr29tHvbYtRT7vm8Jg.png" /></figure><p>The live word count and estimated reading time update as you remove sections. When you’re satisfied, you hit Read and the curated content flows straight into the reader.</p><h4>2.3 Reading controls</h4><p>Everything has a keyboard shortcut because stopping to reach for the mouse breaks flow:</p><ul><li><strong>Space:</strong> play / pause</li><li><strong>Up / Down arrows:</strong> speed ±25 wpm</li><li><strong>Left / Right arrows:</strong> jump ±10 words</li><li><strong>[ and ]:</strong> previous / next sentence</li><li><strong>N:</strong> drop a margin note at the current position</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RxOYJIBSzMraPmvb6aGbZA.png" /></figure><ul><li><strong>B:</strong> drop a named bookmark at the current word</li><li><strong>T:</strong> cycle between Dark, Light, and Sepia themes</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cmE2odul6BNYz_d5V4yM7w.png" /><figcaption>Sepia theme</figcaption></figure><ul><li><strong>F:</strong> cycle font size</li><li><strong>Esc:</strong> exit to the library</li></ul><h4>2.4 Margin notes</h4><p>Press N at any point during reading to drop a note anchored to your exact word position. A small input appears, you type a thought, hit Enter, and reading resumes immediately. At the end of a session your notes are shown in context — threaded through the article at the paragraphs where you wrote them. It turns passive consumption into active annotation without ever breaking the reading state.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OSFTvgzxr2DeS5O-Yt6puQ.png" /></figure><p>This matters more than it sounds. The act of writing a note — even a single sentence — forces a moment of processing that passively watching words flash by never does. Notes are saved alongside the article so they’re there the next time you open it.</p><h4>2.5 Comprehension pulse</h4><p>Every N paragraphs (configurable, default every 5), reading auto-pauses and a prompt appears where the word frame was. Not a quiz — a reflection nudge: “What was the main point of that section?” or “Did anything there surprise you?” You can skip by pressing Space or type a response and press Enter. Responses are saved and shown as threaded notes at the end of the session.</p><p>The research basis is solid. Elaborative interrogation — prompting yourself to explain what you just read — is one of the highest-impact learning strategies identified across cognitive science meta-analyses, consistently outperforming re-reading and highlighting. Comprehension pulse is what separates Fixxate from a speed-reading gimmick and makes it a genuine study tool.</p><h3>3. Onboarding</h3><p>First-time visitors see a three-slide modal before they can upload anything. Slide one explains the 80/20 saccade split with a live comparison graphic. Slide two is a live RSVP demo — words actually flashing — with a speed slider so users can feel the difference between 250 and 500 wpm before committing. Slide three is the keyboard shortcut reference.</p><p>The modal never shows again after dismissal. The state is tracked in localStorage with a single key. Returning users land directly on the library.</p><blockquote>“The onboarding’s live demo does more selling than any paragraph of text ever could.”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/956/1*gQxT0LJo_jjBeeh2yrm9rg.png" /></figure><h3><strong>4. How it’s built</strong></h3><p>Fixxate is built with Next.js. The interesting frontend challenge was designing a reading experience that would feel instant — no skeleton states, no loading spinners mid-session, no layout shifts. The RSVP engine itself is pure client-side logic: no server round-trips during reading, no state synced to a database between words. The entire playback loop runs in the browser.</p><p>That said, the RSVP core was originally prototyped as a self-contained HTML file — useful for rapid iteration since you can open it directly in any browser without a dev server. Some of that simplicity is worth preserving in the final architecture: the playback engine, parsers, and ORP logic have no framework dependencies and could be extracted as a standalone library.</p><h4>4.1 The playback engine</h4><p>The core of any RSVP app is the timing loop. The naive implementation uses setInterval with a fixed delay. The problem: setInterval fires at the same interval regardless of word length or punctuation, and it can’t be dynamically adjusted mid-read without clearing and resetting.</p><p>Fixxate uses a recursive setTimeout chain instead. Each word schedules the next one with its own computed delay:</p><pre>// Compute delay for current word<br>function wordDelay(word, wpm, smartPause) {<br>    var base = 60000 / wpm;<br>    if (!smartPause) return base;<br>    var m = 1;<br>    if (word.length &gt; 12) m *= 1.5;<br>    if (/[.!?]$/.test(word)) m *= 2.2;<br>    else if (/[,;:]$/.test(word)) m *= 1.5;<br>    return base * m;<br>}<br>// Schedule next word recursively<br>function schedNext() {<br>    var w = words[wordIndex];<br>    timer = setTimeout(function() {<br>        wordIndex += settings.chunkSize;<br>        render();<br>        if (isPlaying) schedNext();<br>    }, wordDelay(w, settings.wpm, settings.smartPause));<br>}</pre><p>This means changing the WPM mid-read is instantaneous — the new rate takes effect on the very next word, because there’s no interval to reset. It also means the pause at a full stop is computed per-word, not globally.</p><h4>4.2 ORP calculation</h4><p>The ORP position is approximated with a length-based heuristic — a simplification of the linguistic lookup tables used in the original Spritz patent:</p><pre>function orpIndex(word) {<br>    var clean = word.replace(/[^a-zA-Z0–9]/g, &quot;&quot;);<br>    var n = clean.length;<br>    if (n &lt;= 1) return 0; // single char: start<br>    if (n &lt;= 5) return 1; // short word: position 2<br>    if (n &lt;= 9) return 2; // medium word: position 3<br>    return Math.floor(n * 0.33); // long word: ~33% in<br>}</pre><p>The word is then split into three spans — before, ORP letter, after — and the ORP letter gets the accent colour. The container uses monospace font and a fixed-width frame so the ORP column never shifts between words.</p><h4>4.3 URL fetching and content extraction</h4><p>Fetching arbitrary URLs from a browser page has two layers of complexity: CORS, and content quality.</p><p>For CORS: the app tries four proxies in sequence, falling back automatically on failure. The first proxy (allorigins.win/get) returns a JSON wrapper rather than raw HTML, which sidesteps preflight issues and is the most reliable option for typical article URLs.</p><p>For content quality: Mozilla’s Readability.js is loaded on demand from jsDelivr when the user first fetches a URL. Readability is the same library that powers Firefox Reader Mode — it uses a scoring algorithm based on text density, tag structure, and link ratio to identify the main article body and discard everything else. This gets you clean article content from paulgraham.com, The Atlantic, Hacker News posts, and most major publications without any site-specific rules.</p><h4>4.4 EPUB parsing without a library</h4><p>EPUB files are ZIP archives containing HTML spine items and an OPF manifest. JSZip (loaded eagerly on startup, 100kb) handles the decompression. The parsing is:</p><ul><li>Read META-INF/container.xml to find the OPF file path</li><li>Parse the OPF to get the spine order and manifest item hrefs</li><li>Fetch each spine item in order, strip HTML tags, join text</li><li>Split the full text on whitespace to produce the word array</li></ul><p>The result is one flat array of every word in the book, with progress tracked as a word index. Chapters aren’t preserved as separate units yet — this is a known limitation and a planned improvement.</p><h4>4.5 Persistent storage with IndexedDB</h4><p>A full EPUB or PDF can be 100,000–500,000 words. localStorage’s 5MB limit makes it unsuitable for storing the full word array, but as a Next.js app Fixxate has access to IndexedDB — the browser’s built-in document store with no practical size limit for this use case. The full word array for any uploaded file is persisted to IndexedDB on first parse. On return visits, the file loads instantly from the local store with your position, bookmarks, and settings exactly where you left them. No re-uploading, no re-parsing. The file you read on Monday is there on Friday.</p><h3>5. What’s coming: the roadmap</h3><p>The current version is a solid reading tool. What it isn’t yet is a reading system. The roadmap is about closing that gap.</p><h4>5.1 AI-generated comprehension questions</h4><p>The generic pulse prompts are useful. Prompts generated from the actual paragraph text are significantly more useful. The roadmap includes an optional Claude API integration that reads the last N words of context and generates a specific question — “The author argues X is caused by Y. What evidence does he give for this?” — rather than a generic reflection prompt.</p><h4>5.2 Image support</h4><p>Web articles fetched via URL lose their images entirely in the current version. Two features are planned:</p><ul><li><strong>Images in the curate screen:</strong> during curation, images appear as cards between paragraph cards, exactly where they appeared in the original article. Removable like any other card.</li><li><strong>Image on pause:</strong> when you press Space to pause, if your current paragraph had an associated image in the original article, it fades in below the word frame with its alt text as a caption.</li></ul><p>This is the lowest-friction way to restore visual context without interrupting reading flow.</p><h4>5.3 Chapter navigation and TOC</h4><p>EPUB files have chapter structure that is currently discarded. The next EPUB parser version will preserve chapter boundaries, expose a table of contents, and let you jump by chapter. The reader topbar will show the current chapter name rather than just the book title.</p><h3>6. Design principles: why distraction-free is harder than it sounds</h3><p>Most reading apps have the same problem as the web itself: too much competing for your attention. Reading mode buttons buried in menus. Related articles loading alongside the text you’re trying to focus on. Notifications. Share counts. Comment threads one scroll away.</p><p>Fixxate’s design makes distraction structurally impossible. When you’re reading, the screen shows exactly one word. There is no peripheral content to drift toward. No sidebar. No recommendations. No sense of how long the article is — just the word in front of you and the faint progress bar at the top.</p><p>This sounds obvious, but almost every decision in the product flows from it. The context strip (the dimmed sentences above and below the word frame) exists because pure single-word display felt like reading in a sensory deprivation tank — your brain needs just enough surrounding text to stay oriented. The smart pausing at punctuation exists because without it, RSVP at speed feels like being yelled at. The keyboard shortcuts exist because touching a mouse breaks the reading state.</p><p>The curate screen — where you trim fetched web articles before reading — is the most important distraction-free feature of all. The web article arrives with its boilerplate intact: the author bio, the newsletter CTA, the related posts that snuck through Readability’s filters. The curate screen puts you in control before a single word starts flashing. You decide what’s signal and what’s noise. Then you read.</p><blockquote>“Speed without comprehension is just scrolling with extra steps. Fixxate is designed to slow your attention down even as it speeds your eyes up.”</blockquote><h3>7. Try it</h3><p>Fixxate is free to use. Drop in a book, upload a document, or paste any article URL. No account required. No tracking. The reading experience is entirely offline once the page has loaded — your books and articles never leave your device.</p><p><a href="https://fixxate.vercel.app/"><strong>https://fixxate.vercel.app/</strong></a></p><p>The project is actively in development. If you use it regularly and have feedback — a site that doesn’t parse cleanly, a feature that would change how you use it, a comprehension tool you wish existed — I want to hear it. The roadmap is shaped by what real readers actually need, not what sounds impressive in a product spec.</p><p>If you build on top of it, extend it, or find a smarter way to handle any of the technical problems described here, reach out.</p><p><em>The best ideas for where this goes next have come from people who just started using it.</em></p><p>… And as always, ending the blog with fuel:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QROn9Hdh7j9b3Tqd8kbfXg.png" /></figure><p>… and fixxate is my contribution to this. museum of passion projects!</p><p>— — — — — — — — — — — — — — — — —</p><p>Love it or Hate it? Hit me up on <a href="https://www.linkedin.com/in/ankit-sharma-dev/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b877ccdc37e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[4. Cigarette smokers problem]]></title>
            <link>https://medium.com/@ankitsharma61016/4-cigarette-smokers-problem-2a99edc435ad?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a99edc435ad</guid>
            <category><![CDATA[programming]]></category>
            <category><![CDATA[system-design-interview]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Sat, 14 Mar 2026 12:26:51 GMT</pubDate>
            <atom:updated>2026-03-14T13:07:48.905Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SpjFBnv_N3Mcyou-.png" /></figure><h4>The Core Problem Statement</h4><p>The challenge is to coordinate four threads (one <strong>Agent</strong> and three <strong>Smokers</strong>) so they can share resources without falling into a <strong>deadlock</strong> (where no one can move) or <strong>starvation</strong> (where one smoker never gets to smoke).</p><h4>1. The Participants</h4><ul><li><strong>Smoker A:</strong> Has infinite <strong>Tobacco</strong>. Needs Paper and Matches.</li><li><strong>Smoker B:</strong> Has infinite <strong>Paper</strong>. Needs Tobacco and Matches.</li><li><strong>Smoker C:</strong> Has infinite <strong>Matches</strong>. Needs Tobacco and Paper.</li><li><strong>The Agent:</strong> Represents the Operating System. It possesses all three ingredients but only provides two at a time.</li></ul><h4>2. The Interaction Loop</h4><ol><li>The <strong>Agent</strong> places two random, distinct ingredients on the table (e.g., Paper and Tobacco).</li><li>The <strong>Smoker</strong> who lacks those specific two (in this case, Smoker C, who has Matches) must wake up and grab them.</li><li>That smoker rolls a cigarette, smokes for a while, and then <strong>signals the Agent</strong> that the table is empty.</li><li>The process repeats.</li></ol><p>In the Cigarette Smokers Problem, the characters act as a metaphor for how a computer manages its assets.</p><ol><li><strong>Agent</strong> = Operating System i.e. Allocates resources to different tasks.</li><li><strong>Smokers</strong> = Applications i.e. Need specific resources to execute their code.</li><li><strong>Ingredients</strong> = System Resources Items like memory, disk space, or network ports.</li></ol><p>To complete the statement of the problem, we need to specify the agent code. The agent uses the following semaphores:</p><pre>1 agentSem = Semaphore(1)<br>2 tobacco = Semaphore(0)<br>3 paper = Semaphore(0)<br>4 match = Semaphore(0)</pre><p>The agent is actually made up of three concurrent threads, Agent A, Agent B and Agent C. Each waits on agentSem; each time agentSem is signaled, one of the Agents wakes up and provides ingredients by signaling two semaphores.</p><pre>Agent A<br>1 agentSem.wait()<br>2 tobacco.signal()<br>3 paper.signal()<br><br>Agent B<br>1 agentSem.wait()<br>2 paper.signal()<br>3 match.signal()<br><br>Agent C<br>1 agentSem.wait()<br>2 tobacco.signal()<br>3 match.signal()</pre><h4>The Two Critical Constraints</h4><p>To solve this problem correctly, your logic must satisfy these two functional requirements:</p><ol><li><strong>Liveness (Efficiency):</strong> If the Agent places Paper and Tobacco on the table, Smoker C <em>must</em> be allowed to proceed. The system cannot stay idle if the necessary components for progress are available.</li><li><strong>Mutual Exclusion &amp; Correctness:</strong> The system must not wake up Smoker A or B if only Paper and Tobacco are available. Waking the wrong process leads to “busy-waiting” or errors, as they cannot complete their task without the third ingredient.</li></ol><p>The main goal of this analogy is to ensure two things:</p><ul><li>If enough resources are available for an application to run, the system must wake it up.</li><li>The system should avoid waking an application if it still lacks what it needs to proceed.</li></ul><p>Q. If the Agent provides two resources at once, why do you think it might be a problem if multiple Smokers (applications) try to grab them at the same time without any coordination?</p><p><strong>A. </strong>In a “natural” but flawed solution, each smoker might wait for their two required ingredients one after the other. For example, the smoker with <strong>matches</strong> might run code like this:</p><pre>tobacco.wait() <br>paper.wait() <br>agentSem.signal()</pre><p>Imagine the <strong>Agent</strong> puts out <strong>tobacco</strong> and <strong>paper</strong>. Here is one possible path to a total system freeze:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JDiQn__jH4TONtis7RLoIw.png" /></figure><ul><li>The smoker with <strong>matches</strong> wakes up because <strong>tobacco</strong> is available and successfully executes tobacco.wait().</li><li>At the same time, the smoker with <strong>tobacco</strong>(who needs paper and matches) might wake up because <strong>paper</strong> is available and successfully executes paper.wait().</li><li>Now, the first smoker is stuck waiting for <strong>paper</strong>, but the second smoker already grabbed it.</li><li>The second smoker is stuck waiting for <strong>matches</strong>, which haven’t been put out yet.</li></ul><p>Neither can proceed, and they will wait forever. This is a classic deadlock because multiple threads are holding one resource while waiting for another that is held by a different thread.</p><p>To avoid this, we need a way to ensure that a smoker only starts grabbing ingredients if <strong>both</strong> are actually there. One way to do this is by using “Pushers” — helper threads that act as middle-men.</p><h4>Pushers Solution</h4><p>In this solution, the <strong>smokers</strong> never touch the ingredients directly from the agent. Instead, three helper threads called <strong>pushers</strong> respond to the agent’s signals.</p><ul><li><strong>The Scoreboard</strong>: The pushers use a “scoreboard” of boolean variables (like isTobacco, isPaper, and isMatch) to track what is currently on the table.</li><li><strong>Atomic Decisions</strong>: A pusher only signals a smoker when it confirms that <strong>both</strong> required ingredients are available.</li><li><strong>Prevention</strong>: This prevents a smoker from “half-starting” (grabbing one item and waiting) because the smoker is only woken up when the pusher knows the set is complete.</li></ul><p>When the <strong>Agent</strong> signals that an ingredient is available, the corresponding pusher wakes up. For example, if <strong>tobacco</strong> is put out, <strong>Pusher A</strong> wakes up and checks the other flags:</p><ol><li>If it sees isPaper is already <strong>True</strong>, it knows both tobacco and paper are ready. It then signals the <strong>Smoker with Matches</strong> 🕯️.</li><li>If it sees the table is empty, it simply sets isTobacco = True and goes back to sleep.</li></ol><p>Imagine the <strong>Agent</strong> has just put out <strong>tobacco</strong> and <strong>paper</strong> . <strong>Pusher A</strong> (the tobacco pusher) wakes up first, sees the table is empty, and sets isTobacco = True.</p><p>Q. Can you write code for this?</p><p><strong>A.</strong></p><pre>Pusher A (Tobacco pusher)<br>tobacco.wait()<br>mutex.wait()<br>if isPaper:<br>  isPaper = False<br>  matchSem.signal()<br>else if isMatch:<br>  isMatch = False<br>  paperSem.signal()<br>else:<br>  isTobacco = True<br>mutex.signal()<br><br>Pusher B (Paper pusher)<br>paper.wait()<br>mutex.wait()<br>if isTobacco:<br>  isTobacco = False<br>  matchSem.signal()<br>else if isMatch:<br>  isMatch = False<br>  tobaccoSem.signal()<br>else:<br>  isPaper = True<br>mutex.signal()</pre><ul><li><strong>Coordination</strong>: The pushers use a <strong>mutex</strong> to ensure only one pusher is looking at the “scoreboard” at a time.</li><li><strong>Precision</strong>: A smoker is only signaled via their specific semaphore (like matchSem) when a pusher has confirmed a complete set of ingredients is ready.</li><li><strong>Safety</strong>: The smokers themselves have very simple code now: they just wait on their specific semaphore, make the cigarette, and tell the agent they are done</li></ul><p>Each smoker waits for their specific “ready” signal from a pusher. For example, the <strong>Smoker with Tobacco</strong> (who needs paper and matches) would look like this:</p><pre>// Smoker A - who has paper and match<br>tobaccoSem.wait() <br>smoke()<br>agentSem.signal() // Signal the agent that the table is clear and more ingredients can be placed</pre><p>The reason the smoker’s code is so simple is that the <strong>pushers</strong> have already verified that the necessary resources are available before signaling the smoker. The smoker doesn’t have to check anything; if they are woken up, they know they have what they need to proceed.</p><h4>Variant 2: Accumulation</h4><p>Now we modify the agent, eliminating the requirement that the agent wait after putting out ingredients. In this case, there might be multiple instances of an ingredient on the table.</p><p>If the agents don’t wait for the smokers, ingredients might accumulate on the table. Instead of using boolean values to keep track of ingredients, we need integers to count them.</p><pre>1 numTobacco = numPaper = numMatch = 0</pre><pre>Pusher A modified code<br>1 tobacco.wait()<br>2 mutex.wait()<br>3 if numPaper:<br>4   numPaper -= 1<br>5   matchSem.signal()<br>6 elif numMatch:<br>7   numMatch -= 1<br>8   paperSem.signal()<br>9 else:<br>10   numTobacco += 1<br>11 mutex.signal()</pre><blockquote>This is an example of a pattern we will see several times, which Author calls a <strong>scoreboard</strong>. The variables numPaper, numTobacco and numMatch keep track of the state of the system. As each thread files through the mutex, it checks the state as if looking at the scoreboard, and reacts accordingly.</blockquote><h4>Conclusion</h4><p>The Cigarette Smokers Problem perfectly illustrates the real-world challenges operating systems face when allocating scarce resources to competing applications. A simple “two-out-of-three” strategy can trigger deadlock if threads grab ingredients one at a time — exactly what the flawed natural solution demonstrates.</p><p>The elegant fix uses <strong>pushers</strong> (scoreboard threads) as gatekeepers. Protected by a mutex, they check the table state (booleans or counters) and only wake the correct smoker when both required ingredients are confirmed present. Smokers then become trivial: wait on one semaphore, smoke, and signal the agent.</p><p>Even the variant allowing ingredient accumulation simply upgrades booleans to counters — showing that solutions must adapt to exact constraints.</p><p>Key takeaways for any concurrent programmer:</p><ol><li>Never acquire resources incrementally without a full-set guarantee.</li><li>Use helper threads/monitors for safe decision-making.</li><li>A scoreboard pattern (shared state + mutex) is the cleanest way to see the big picture.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*ik2ZoZu1ukps2DMX" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a99edc435ad" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[3. Dining Philosophers Problem]]></title>
            <link>https://medium.com/@ankitsharma61016/3-dining-philosophers-problem-ae2b70868867?source=rss-a4769bef3762------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae2b70868867</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[software-engineering]]></category>
            <category><![CDATA[java]]></category>
            <category><![CDATA[programming]]></category>
            <dc:creator><![CDATA[Ankit Sharma]]></dc:creator>
            <pubDate>Tue, 10 Mar 2026 01:01:01 GMT</pubDate>
            <atom:updated>2026-03-10T01:01:01.486Z</atom:updated>
            <content:encoded><![CDATA[<p>Proposed by Edsger Dijkstra in 1965, this thought experiment paints a picture of five philosophers sitting around a circular table. Between each philosopher is a single fork, meaning there are exactly 5 forks on the table.</p><p>These philosophers spend their lives doing only two things: thinking and eating. Here is the catch: to eat the spaghetti, a philosopher absolutely needs <em>two</em> forks — the one on their left and the one on their right. Once they finish eating, they put both forks back down on the table.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/840/1*FSKbBnq9PQiVOs6yE3BOTA.png" /></figure><p>The solution should be deadlock and starvation free.</p><h4>Naive solution</h4><p>Since we have to enforce exclusive access to the forks, it is natural to use list of Semaphores, one for each fork. Initially, all the forks are available.</p><pre>forks = [Semaphore(1) for i in range(5)]</pre><p>Let’s imagine we write a simple, identical program for every philosopher that looks like this:</p><pre>1. Wait for (acquire) the left fork.<br>2. Wait for (acquire) the right fork.<br>Eat!<br>3. Put down the right fork.<br>4. Put down the left fork.</pre><p>This seems logical, but it has a massive flaw.<br>Q. Can you guess the flaw?</p><p><strong>A. Deadlock: </strong>if everybody took their left fork first, then none of them would have their right fork i.e. A deadlock. Because the table is round and everyone follows the exact same symmetrical rules, we end up in a “circular wait.” Every single philosopher is holding a left fork and waiting forever for a right fork .</p><p>To fix this, we need to break the cycle. There are a few clever ways to do this, and the book introduces a couple of them.</p><h4>Solution #1 Bouncer</h4><p>Imagine we hire a bouncer to stand at the door of the dining room. The bouncer’s job is to limit the number of philosophers allowed to sit at the table at the exact same time.</p><p>Q. Since there are 5 forks on the table, what is the <strong>maximum</strong> number of philosophers the bouncer should allow to sit down at once to absolutely guarantee that at least one of them can grab two forks and eat?</p><p><strong>A. 4 Philosophers — </strong>If 4 philosophers sit at the table and they all stubbornly grab their left fork, that takes 4 forks. There is still 1 fork left on the table. That 5th fork <em>must</em> be exactly what one of the seated philosophers needs for their right hand. Once they eat, they put both forks down, opening the door for everyone else, completely avoiding the deadlock.</p><p>How would you implement this Solution? The end goal is to limit the number of threads accessing a critical region, and we do it using Semaphores initialized to 4. Let’s name the semaphore as footman</p><pre>footman = Semaphore(4)</pre><blockquote><strong>Multiplex pattern<br></strong>In concurrent programming, a multiplex is a generalization of a mutex. While a basic mutex (mutual exclusion) only allows <em>one</em> thread into a critical section at a time, a multiplex allows up to <em>n</em> threads to enter simultaneously.</blockquote><blockquote>To build this, we use a standard semaphore initialized to our maximum capacity, <strong>n</strong>.<br>When a thread arrives, it calls wait() (which decreases the semaphore count).<br>When a thread leaves, it calls signal() (which increases the semaphore count).<br>If the semaphore count hits 0, any new thread calling wait() is blocked and put in a queue until someone else leaves.</blockquote><p>Using footman semaphore, our new modified code would now look like this:</p><pre>0. footman.acquire()<br>1. Wait for (acquire) the left fork.<br>2. Wait for (acquire) the right fork.<br>Eat!<br>3. Put down (release) the right fork.<br>4. Put down (release) the left fork.<br>5. footman.release()</pre><p>By wrapping the fork-grabbing inside the footman multiplex, you perfectly restrict the table to 4 philosophers at a time.</p><p>Because those 4 philosophers can hold a maximum of 4 forks initially, there will <em>always</em> be 1 fork left on the table. That extra fork guarantees that at least one of the seated philosophers can grab their second fork, eat, and then put everything down so the next person can go .</p><p>standard solution! But the book offers another completely different way to solve this — one that doesn’t require a footman bouncer at all.</p><h4>Solution #2: The Oddball philosopher</h4><p>Think about the root cause of the deadlock: perfect symmetry. In our original broken solution (Naive solution), every single philosopher acted exactly the same way. We can think of them all as “lefties” — they all stubbornly grabbed their left fork first, then waited for their right. That symmetrical behavior is exactly what created the circular wait.</p><p>Q. What do you think would happen if we changed the rules so that four of the philosophers are “lefties”, but the fifth philosopher is a “rightie” (meaning they always try to grab their <em>right</em> fork first)? Would that break the deadlock cycle?</p><p><strong>A. </strong>By having just one philosopher reach for their right fork first, they get blocked <em>before</em> picking up any forks at all. That leaves their left fork sitting on the table, which allows their neighbor to grab their second fork, eat, and start a chain reaction that clears the table.</p><p>Q. Prove that if there is at least one leftie and at least one rightie, then deadlock is not possible.</p><p><strong>A. </strong>The proof works by contradiction. First, assume that deadlock is possible. Then choose one of the supposedly deadlocked philosophers. If she’s a leftie, you can prove that the philosophers are all lefties, which is a contradiction. Similarly, if she’s a rightie, you can prove that they are all righties. Either way you get a contradiction; therefore, deadlock is not possible.</p><p>These first two solutions (the bouncer/multiplex and the leftie/rightie asymmetry) are fantastic, but the book also introduces a third famous approach by Andrew Tanenbaum.</p><h4>Pseudo-Solution #3: Tanenbaum’s solution</h4><p>Instead of just looking at the forks, Tanenbaum’s solution tracks the actual <em>state</em> of each philosopher using three labels: &#39;thinking&#39;, &#39;hungry&#39;, or &#39;eating&#39;.</p><p>Q. If a philosopher is currently &#39;hungry&#39;, what specific state must <em>both</em> of their immediate neighbors be in (or NOT be in) before our hungry philosopher is allowed to transition to the &#39;eating&#39; state?</p><p><strong>A. </strong>They must <strong>not</strong> be in the &#39;eating&#39; state. If both neighbors are either &#39;thinking&#39; or &#39;hungry&#39;, our philosopher is clear to grab the forks and transition to &#39;eating&#39;.</p><pre>1 state = [’thinking’] * 5<br>2 sem = [Semaphore(0) for i in range(5)]<br>3 mutex = Semaphore(1)</pre><pre>// Tanenbaum’s solution<br>1 mutex.wait() // to ensure safe updates to state array<br>2 state[i] = ’hungry’<br>3 test(i) // function to check i-th philosopher is eligible to eat, if yes, signal sem[i]<br>4 mutex.signal() <br>5 sem[i].wait() // based on line 3, blocks or allows philosopher to eat.<br><br>eat()<br><br>// while leaving the dining table<br>6 mutex.wait() // acquire mutex for safe concurrent updates<br>7 state[i] = ’thinking’<br>8 test(right(i)) // check if right and left neighbours are eligible for eating, if so, turn their semaphore to 1<br>9 test(left(i))<br>10 mutex.signal() // exit<br><br><br><br>// The test function checks whether the ith philosopher can start eating, which<br>// he can if he is hungry and neither of his neighbors are eating. If so, the test<br>// signals semaphore i.<br>def test(i):<br>    if state[i] == ’hungry’ and state[left (i)] != ’eating’ and state[right (i)] != ’eating’:<br>      state[i] = ’eating’<br>      sem[i].signal()</pre><p>In Tanenbaum’s solution, when a philosopher finishes eating, they look at both their left and right neighbors. If a neighbor is &#39;hungry&#39; AND that neighbor&#39;s <em>other</em> neighbor is not eating, the departing philosopher actually hands them the go-ahead by signaling their personal semaphore to wake them up .</p><p>This solution brilliantly avoids deadlock because no one locks a single fork and holds it hostage. However, the book points out that it has a subtle flaw when it comes to <strong>starvation</strong>.</p><p>Imagine five philosophers numbered 0 through 4 around the table.</p><p>Q. What do you think would happen to Philosopher 2 if Philosopher 1 and Philosopher 3 perfectly coordinated to take turns eating, passing the “active” state back and forth between just the two of them?</p><p><strong>A. </strong>If Philosopher 1 and Philosopher 3 alternate perfectly with Philosopher 2 and Philosopher 4, they can effectively trap Philosopher 0 (or any other specific philosopher) in a permanent state of waiting . The hungry philosopher will just sit there while their neighbors constantly pass the forks back and forth over their head!</p><p>Because of this starvation risk, Tanenbaum’s approach is brilliant for avoiding deadlock, but it doesn’t strictly satisfy all the requirements of a perfect synchronization solution.</p><h4>Conclusion</h4><p><strong>Deadlock (Circular Wait) :</strong> A scenario where every thread is stuck waiting for a resource held by another, bringing the entire system to a complete halt due to perfect symmetry.</p><p><strong>The Multiplex Pattern:</strong> Using a bouncer (a semaphore initialized to a maximum capacity) to limit access. By restricting the table to 4 philosophers, we guaranteed an extra fork was always available.</p><p><strong>Asymmetry :</strong> Intentionally breaking identical thread behavior (like introducing a “rightie” philosopher) to mathematically prevent a circular wait.</p><p><strong>State Tracking &amp; Starvation:</strong> Tanenbaum’s approach of checking neighbors’ states before acting. This avoids deadlock but introduces the risk of starvation, where a specific thread might never get a turn if its neighbors perfectly alternate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/0*zLIr3sUNbhwGVzFU" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae2b70868867" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>