Life on Hard Mode with Pratik Karki

Life on Hard Mode #5 - The Financial Stack: From Debt to Founder

Ramen Profitability: My Personal Journey DISCLAIMER: I am not a certified financial planner, so take my advice with a grain of salt. But I am someone who had to figure this out with zero safety net, with no prior knowledge.As they say, “Doing it yourself beats reading about it in a book 100% of the time.” When I started at Google in 2021, I was $22,500 in debt. I was on a visa, meaning if I lost my job, I would lose my apartment and would be on the next flight home. I was sending (and continue to send) money back to my family. Between my job offer at Google and my first paycheck, I primarily subsisted on Ramen fried rice [https://www.seriouseats.com/cup-noodle-fried-rice] to make ends meet. Fast forward three years: * Paid off all student loans. * Fully paid for my wedding. * Paid for my immigration process. * Built a mid-six-figure net worth. * The Ultimate ROI: Saved enough “Runway Cash” to quit a $250k/year job to start Anthromind [http://anthromind.com] with $0 income. The main takeaway is that I treated personal finance like an engineering problem. I debugged my spending, optimized my savings rate, and built a system that runs on autopilot. The Engineering Approach to Finance: Building the System Please don’t take this guide as a guide to getting rich quickly; it’s about building a reliable machine that turns income into freedom. I broke my financial life down into components, just like a software architecture. * Component 1: The “Visa Tax” & Playing Defense Most financial advice assumes you start at zero. Immigrants often begin at a disadvantage. We don’t just have standard bills; we have a “Visa Tax.” It means a lot of money in visa fees, inability to work side hustles, no social benefits, high remittance fees, etc. Even getting a credit card was a pain in the ass. I remember the Discover student card was the only one I qualified for in my first semester of college. And that allowed me to build credit very quickly, thankfully. I didn’t think about investing until I was a couple of years into college, when I was putting money into GameStop and Dogecoin, lol. But before that, I was patching the holes, using summer internships to pay down my credit card debt and cover my tuition the following semester. * Component 2: The Automation Engine There’s a good financial lesson: “Pay yourself first.” It might sound like it means give yourself money first, but it actually means put money into your retirement accounts before you see any of your paycheck. I usually set up automatic transfers on payday: * 401k Match (free money and Google had a very generous match) * HSA (triple tax advantage, don’t touch the principle) * Roth IRA (tax-free growth) * High-Yield Savings in a Brokerage * Contribution to a 6-12 months “emergency savings” account * Student loan debt, if you have any. I recommend using the snowball approach [https://www.ramseysolutions.com/debt/how-the-debt-snowball-method-works], which is to pay off the highest-interest debt first. By automating this, I removed myself from the equation. The money was gone before I could spend it. I also received a quarterly bonus, which I would aggressively put into a Roth IRA and 401(k) to save on taxes. I recommend people do the same for year-end bonuses. * Component 3: Grow, grow, grow baby! If you aren’t aware of the magic of compound growth, let this be your lesson. If you contribute to all of these funds and set aside a good portion of your paycheck, with compound growth and smart investing, you can achieve up to 20% year-over-year returns and double your money every 5 to 7 years. That can afford you a lot of things by cutting back on your lifestyle. Especially if you’re in your 20s, you should be saving as aggressively as possible because you give your money the highest amount of time to compound. I believe the general goal is to have: - 1 year’s salary in your 401(k) by the time you’re 30 - 3-4 years’ salary by the time you’re 40By working those years in your twenties, playing defense, and compounding your money, you are in a solid position in your thirties to buy a house, travel more, have kids, and retire early, etc. Friday Positivity: Looking On The Bright Side Today’s positivity piece is brought to you by this perfect illustration of perspective. I am what you call an “optimistic nihilist.” I believe it was the French philosopher Sartre who first spoke extensively about nihilism. Later on, a whole movement developed about being optimistic despite nihilism. I tend to be in that camp because if nothing really matters, then every single thing you do is valuable, right? It might also be because I am playing Clair Obscur: Expedition 33, an existential masterpiece, that I am discussing this for this week’s topic. It took me about a year to fully grasp the logic that I have one life, and if I do not take as many risks and leap as far as I can in this one life, what is the point? That’s why I’m very daring. I do not care about putting myself out publicly. I do not care about making ostentatious goals for myself. Because if nothing really matters, then everything I do really matters. I recommend that everybody think through things this way. Topic of the week: The Immigrant Financial Stack In today’s podcast, I break down the exact “order of operations” I used to allocate every dollar I earned, expanding on the engineering approach above. I cover: * Defense: Living like a student even when earning a tech salary (and how to deal with lifestyle creep). * The “Visa Tax”: Specific financial hurdles for immigrants (credit building without history, sending remittances, legal fees). * The Google Strategy: How I utilized 401k matching, HSAs, and the Mega Backdoor Roth (the most underrated tool in tech). * Buying Freedom: Why I bought the “Anthromind Runway” to fuel the growth of the next stage of my career. Resources: To help you build your own system, I’ve provided the spreadsheets I actually use: * My Personal Budget Sheet: This is the actual template I used to track my income, expenses, and savings rate. It’s not fancy, but it forces you to see exactly where your money is going. * The Office Benefits Guide: A compilation of the most common workplace benefits (HSA, Student Loan Matching, FSA) that people ignore. I’ve included common misconceptions and strategies for each. * The “Pay Yourself First” Protocol: A simple checklist for your payday routine. It ensures your savings and investments are funded before you start spending. * The “Leaky Bucket” Tracker: If you don’t know where your money is going, use this simple daily log for 30 days. It helps catch the small expenses that drain your wallet. To access them all, go to this Google Drive link. [Link [https://drive.google.com/drive/folders/1ONdVo3567TXknhZOaNePcS_iXlqAFN8w?usp=sharing]] P.S. What have I been reading/watching/playing? * [Paper] EchoLeak: The First Real-World Zero-Click Prompt Injection Exploit in a Production LLM System: I’ve been more and more fascinated with data and memory leakage using LLMs, and this is an excellent paper on this topic. The researchers conducted a white-hat vulnerability exploitation of Microsoft Copilot using techniques such as Markdown to retrieve assets, all without any user interaction. The lesson here is how to build stable co-pilots without these security vulnerabilities. Source [http://ojs.aaai.org/index.php/AAAI-SS/article/view/36899] * [Movie] Fackham Hall: A hilarious comedy movie that’s also very satirical about downtown Abbey, hoodunit mysteries like Knives Out, and the style of comedy is very much like the Naked Gun series. It was a great time and just made me laugh. * [Book] The Captain Class: I was gifted this book by a friend (thank you, Paras dai). It’s a compelling, data-driven case study of what an actual glue leader looks like. Primarily written about sports teams like the legendary teams of Ferens Puskas, Tim Duncan, Tom Brady, etc. I believe there are many good business use-case translations as well. What stood out is that it’s not the “superstars” that are the best captains. A lot of them are actually antisocial and hard to get along with. But their conviction and win-at-all-costs mentality are what make them outliers. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pratikkarki.substack.com [https://pratikkarki.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

Life on Hard Mode #4 - How VC Math Works, No Cap

Explain Like I’m Five: WTF are Venture Capital Firms? Earlier in 2025, I had the opportunity to present on raising capital to a few universities in the Bay Area and Southern California. It was an incredible experience because they were wide-eyed, curious economics/business students who really leaned into the lecture.Find the full slides that I presented over in this Google Drive [https://docs.google.com/presentation/d/1TS5NYidDOG-SGCIqxi72BHnI5wVpdT8TmIcW8aLAbo8/edit?usp=sharing].I literally knew nothing about the venture capital world a year ago when I co-founded Anthromind. So, in six months, being able to present and talk about it in depth required a lot of learning. I was on the ground, building relationships and absorbing as much knowledge as I could on this topic. The main ideas I touched upon were: * How do VCs operate and make money? * Should you even bother raising venture capital money? * What are the benefits of raising venture capital money, and what are the drawbacks? * What do the terms 83B election, QSBS, preference tax, etc., mean? * How do you get noticed by VCs? Myth: Every Successful Startup Should Raise Venture Capital Money People think, “let me form a startup, raise some capital, and then I’ll start working,” But in fact it’s the opposite. You typically build something for a while until it’s “venture-ready”. In fact, validating your idea and getting customers is a no-brainer before even attempting to get funded. The best time to raise capital is when you really need a big lever. Like the Archimedes principle - “Give me a big enough lever and I can move the world,” venture capital funding gives you the biggest possible lever for your start-up. Anthromind [http://anthromind.com] has a massive market, and VC funding has a great lever, for example. So, should you raise capital for your startup? First, consider the outcome that you want. If you wish to build a billion-dollar company, are dead set on it, and willing to risk anything for it, then probably yes. Think about the total market that you have. If you’re doing Uber for dogs, what is this world market for that? That is very different from, like, an AI-native startup building SREs for companies. That is a very different market, with very different potential revenues, so you need to index for that. In the chart below, I show why the venture capital business model is a little skewed: they expect one company out of a hundred or even thousands to become the next Tesla or Apple, and that’s what floats the whole VC portfolio. There are also other considerations. You could raise $2M of seed at a $20M valuation. But what happens if your company does not actually meet the expectations? What if you’re stuck at $2M of annual recurring revenue for the next five years? The reality is you’ve stalled, the VCs won’t pick up your calls, and you have nowhere to go, so you effectively become a zombie company. There are so many of these companies that get seed funding and never grow beyond that point because they are effectively zombie companies. In fact, a lot of the time, it would have been better never to raise any capital at all and to continue growing sustainably on angel checks. In conclusion, it’s a hard call, but please weigh all your options before considering raising venture capital. Friday Positivity: It’s Okay To Be Disliked For this week’s Friday Positivity, the topic is “It’s okay to be disliked.” There’s a great saying: “Pleasing everybody means earning nobody’s respect.” I found that to be very true throughout my life. I allude to the crabs-in-a-bucket phenomenon, and it is very accurate in how it works in society. If you’re not aware of the crabs-in-the-bucket mentality, it’s when folks would pull down for others trying to better themselves. There are a lot of reasons for it: jealousy, culture, we’re all in this together, but it’s so detrimental to progress. I feel like this is common among immigrant households, based on personal experience seeing friends and family from Nepal and other immigrant communities. They really treat folks who are trying to ameliorate their situation with a lot of disdain, actually, and it really stifles them.Don’t be an a$$hole, but don’t be a pushover either. If you want the life that you wish to have, be ready to step on some toes. As they say, it’s lonely at the top. Topic of the week: How VC Math Works, No Cap I explained a lot about the VC math situation above, but the full podcast talks about a lot more. We go over the QSBS tax incentive, what the VC fund model looks like, and how they actually make money. So listen to the full podcast episode and give me your comments, please. P.S. What have I been reading/watching/playing? * [TV Show] Pluribus: Vince Gilligan is the creator of Breaking Bad and Better Call Saul, and he has not missed so far. I really like his new show Pluribus; it’s a lot of things. It’s an introspection on AI usage in this world. It talks about communism because it’s a direct homage to The Invasion of the Body Snatchers. I feel like it’s a great character study of a highly individualistic person, Carol, who thinks she’s better than a hive mind but can't even do the most basic things, like taking out the trash, without help. So the actual story is, is everybody an island, or do we rely on each other more than we think? * [Paper] Generalist foundational models are not clinical enough for hospital operations [https://www.arxiv.org/abs/2511.13703]: This is an excellent paper on how taking a general foundation model, even the ones specifically built for clinical operations, is not that great when it comes to building trust and accuracy among clinical use cases. The authors of this paper also created one of the world’s largest open-source datasets for medical images. That’s why the agent over here, CheXagent, is really great. It aligns with what we’re doing in Anthromind and with building the best possible post-training datasets. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pratikkarki.substack.com [https://pratikkarki.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

Life on Hard Mode #3 - Thankfulness and Looking Ahead

“A hungry mouth is not dissuaded by a moustache” - Nepali proverb Friday Positivity: Top Things I’m Thankful For This Year * Eternally thankful to my wife and family for supporting me all throughout the process * Thankful for my co-founder and team at Anthromind [http://anthromind.com] * Thankful for my mentors, friends, and YOU for tuning in to this podcast, which I’m trying to get better at every week * I’m thankful for the good weather in California * Finally, I’m thankful for the life I get to live Topic of the Week: Looking back and ahead Please listen to the full podcast above! I talk about what’s led me here, what’s in store for me in the future. I touch on my roots in Nepal, how I grew up loving books and video games, and how I even developed my own video games using GameMaker 7. I talk about getting into high school and college in the US on full-ride scholarships, and about taking a gap year after high school because of the Nepal earthquake. I also talk about trying to get a job while juggling extra-curriculars and 20 hours of work during college, the struggles that come with that, and how I was able to overcome them. Finally, what it was like working at Google in San Francisco, and all the cool stuff I did. Finally, what led me to go into the full founder mod,e hence starting Anthromind P.S. What have I been reading/watching/playing? Clair Obscur: Expedition 33 is a genuine masterpiece of a game. How can a team of 25 led by a former-Ubisoft developer create something that puts every other video game that costs hundreds of millions of dollars to shame? Clair Obscur is an exciting game that explores themes of mortality and grief. The gameplay is fascinating because it combines the classic role-playing game mechanics of Chrono Trigger or Final Fantasy with a 3D environment like God of War. The turn-based combat is super fun. I was worried because I find some turn-based games a little slow, but in this one, it’s actually very action-packed because you have to time your parries and dodges really well, otherwise you’re gonna die very quickly. The difficulty is fair; it’s a bit punishing, but that’s what the fun is. If you don’t die to a boss a handful of times, are you even learning anything right? The visuals are also astounding. Probably the prettiest graphics I’ve ever seen on screen. They capture the Belle Époque aesthetic of France really well. Finally, the music is terrific! I thought the music from Silksong was incredible, but this one blows it out of the water. Really happy I started playing this game, and I recommend everybody play it as well. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pratikkarki.substack.com [https://pratikkarki.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

Life on Hard Mode #2 - How to Get into Big Tech

Life on Hard Mode # 2 - How to Get into Big Tech There is a saying: “Don’t be afraid of the man who can do 10,000 different karate moves. Be afraid of the person who has done the same move 10,000 times.” From 2016 to the present, I’ve probably interviewed at hundreds, if not thousands, of companies. I also interviewed quite a few folks, too. But I started pretty modestly. Here are my lowlights from college, for example: * 4 rejections from Google for internships and jobs (2016-2019) * Multiple failed final stage interviews at companies like Meta, Databricks, Microsoft, NVIDIA (I would have been filthy rich by now if I got that offer in 2020), etc. * Well over 3000 applications over the course of college * Rescinded internship offer during the summer of COVID What qualifies me is the plethora of experience I’ve accumulated. A lot of understanding of the fundamentals and how hiring practices are conducted in big tech and at my company, Anthromind [http://anthromind.com]. Even though the underlying technology changes (e.g., more agentic frameworks, prompt engineering, GDPO, or PPO), the core methods of hiring and evaluating candidates remain the same. Since so many people asked about getting job offers, and in today’s market, it is brutally difficult, here is my no-BS guide. Friday Positivity! Today’s positivity piece is “showing up is half the work.” Topic of the week: How to get into Big Tech In today’s podcast, I cover everything from my background before Google to my technical preparation, personal development, interview prep, and more. Talk about the main things that big companies look for: * Strong fundamentals * Ability to break down problems * Clear communication * X-factor or culture fit Finally, since the job market in 2025 is so different from even two years ago, I have my own principles on what I would do differently. As I am recruiting, I look for this myself. Hence, I think I have a good grasp of it. Resources: * A resume teardown * A complete hiring rubric guide * A worksheet to develop your “stories” from experiences that fit into any behavioral question pattern * A step-by-step checklist for the entire process from first application to post offer * A form where you can submit your resume for feedback To access them all, go to this Google Drive link [https://drive.google.com/drive/folders/1Fnns5bQraTj3WxJvTblsPuq32V_yyuco?usp=sharing]. And the anonymous resume feedback form is here: https://forms.gle/x1tXLagYvE5agA1v5 [https://forms.gle/x1tXLagYvE5agA1v5] (Please provide a URL. There is no upload feature.) P.S. What have I been reading/watching/playing? * [Paper] MiroThinker: Pushing the performance boundaries of open-source research agents via model, context, and interactive scaling [https://www.arxiv.org/abs/2511.11793]: this is a fascinating paper about how a model itself can be scaled internally vs. by increasing the model size or context length * [Paper] Generalist foundational models are not clinical enough for hospital operations [https://www.arxiv.org/abs/2511.13703]: This paper shines a light on what we are seeing daily at Anthromind. For example, companies, primarily in the clinical space, using base or even “healthcare” pre-trained models are not seeing the results that they want. In fact, the results are horrendous. Proper post-training is so necessary for clinical operations. * [Podcast] RL is even more information-inefficient than you thought [https://www.dwarkesh.com/p/bits-per-sample]: This is the most recent issue of the Dwarkesh podcast. And in this one, Dwarkesh talks about the jagged nature of RL and the true cost inefficiencies. Very fascinating stuff. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pratikkarki.substack.com [https://pratikkarki.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

Life on Hard Mode #1 - Pilot

Why did I start a podcast? I started this podcast because I wanted a place to speak honestly about the work I am doing, the ideas I care about, and the challenges that come with building Anthromind. My inbox has been full of questions about careers, AI, visas, startups, and life choices. I cannot give everyone long personal replies (trust me, I tried), but I can share what I am learning in a format that scales. A weekly podcast forces me to slow down and think. It lets me break down research papers, explain how I build and sell products, talk about what I see in the AI world, and tell the truth about what it feels like to live life on hard mode. It is also a way for me to stay accountable. Every Friday, I have to show up, speak clearly, and share what I worked on. It’s a no BS method of sharing and building in public as much as I am able to. I want this to be useful to people who are trying to carve their own path. If someone can learn from my mistakes or get a clearer picture of how to build something in AI today, then this will be worth it :) The podcast is my way of opening the door a little wider and letting people see how all of this actually works.Please continue to send me your questions, thoughts, and concerns. I’ll tackle them every week. Also, I’m open to ideas on areas I should focus upon. Topic of the week: Post-training for LLMs explained Post-training is the set of steps that happen after an LLM finishes its main training run. The base model (think GPT 4 or Deepseek R1) learns general patterns from huge amounts of text, but that alone does not make it safe, helpful, or aligned with what people want. Post-training is where we shape the model into something usable. And this is why I built Anthromind [http://anthromind.com]specifically. There are three major parts worth understanding. 1. Preference training This is where humans show the model examples of good and bad behavior. We rank outputs, score them, and teach the model what we prefer. This shifts the model toward clearer reasoning, safer answers, and more predictable behavior. Think of ranking by bias, trustworthiness on a scale of zero to ten. Recent work, like step-wise preference training, focuses on rewarding the thinking process rather than just the final answer. 2. Reinforcement learning and policy shaping Here we adjust the model using reward signals. The idea is simple. If the model gives answers we like, we push it to produce more of that behavior. If it produces something we do not want, we correct it. Newer techniques use faster and more stable methods than the original RLHF. Techniques like Reinforcement learning through Verifiable rewards (RLVR). The industry is moving toward training the model to explain its reasoning chains, which makes its internal decisions easier to steer. 3. Tool use and structured behaviors Large models today do more than predict text. They plan, call APIs, write code, and follow rules. Post-training is where we teach the model how to use tools, how to follow steps, and how to interact with real systems. This is also where we reduce errors, hallucinations, and unsafe actions. In a sense, post-training is most important here, and organizations adopting AI cannot survive without it. The short version is that post-training is the bridge between raw capability and real-world usefulness. It is how we take a powerful yet chaotic system and shape it into something that can help people, answer questions safely, and work seamlessly within products. Without post-training, even the strongest base model is not something you can ship. In conclusion, AI adoption will not be possible without post-training techniques. A recent article by MIT reported that 95% of generative AI pilots failed, and the lack of post-training techniques and awareness is cited as the reason. Sources: * https://alphaxiv.org/abs/2502.21321 * https://x.com/tszzl/status/1948907851508056495 * https://alphaxiv.org/abs/2509.25300 * https://alphaxiv.org/abs/2509.17866 Questions and answers! I received 24 questions across LinkedIn, X, FB, Insta, and Substack. I’ve compiled and answered quite a few in the episode! I’ll get to more in future episodes. Please keep them coming. P.S. What have I been reading/watching/playing? * [Book] The Hard Thing About Hard Things: To be honest, I heard about this book for the longest time and didn’t really see a point in reading it, but I’m thankful that I did. Ben Horowitz goes through the details about why founding and starting a company feels like it’s rainbows and whiskers, but is not really true. It is brutally tough, man. However, it is also inspiring because he walks us through how to navigate difficult scenarios in a very pragmatic manner. * [Book] The Fall of Hyperion: I read the first Hyperion book by Dan Simmons, and immediately I was like, “I need to read the sequel.” And the sequel does not disappoint. It’s a different storytelling structure than the first one, but it has a very complex story and touches upon topics such as AIs and systems so complex that humans cannot comprehend them. It is truly ahead of its time. * [Movie] Guillermo Del Toro’s Frankenstein: I’ve been a fan of Guillermo del Toro since Pan’s Labyrinth and the Hellboy days. Frankenstein was released this year, sadly, only on Netflix, with a limited theatrical release. It’s mind-blowingly good and should be watched in theaters like I did. The less said, the better, but you truly feel for the monster by the end. Probably my favorite movie of the year so far. * [Game] Hotline Miami 1 & 2: I played the first Hotline Miami game when I was a kid, and it was really fun. The music was good, but I didn’t really understand the story. I finally played the second one, which serves as a proper conclusion and includes a lot of backstory for the characters. The story is so compelling; it explores themes such as violence, escalation, and legacy. I’ve been a massive fan of it ever since I played it again, and have been listening to music on repeat. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit pratikkarki.substack.com [https://pratikkarki.substack.com?utm_medium=podcast&utm_campaign=CTA_1]