Kaushik Mahorker

To all the founders about to quit and dreaming of a FAANG job or equivalent in pay & prestige, listen up: 1. What you're doing now-- trying to build & grow something new-- is the only "job" that AI and robots won't be taking. 2. The job you are dreaming about is already gone. Push through to fight another day. Be relentlessly resourceful as PG puts it. Your problem needs to be solved and no one is going to solve it for you. If you want help/advice on cutting expenses, sharing upside with community in exchange for build & grow collaboration, and raising capital from your users-- DMs open. There is no greener grass than the type that grows on PMF. Using Slyk IO to launch, grow, and raise fast puts you on the fast track to finding out. Your community can help you get there. I know because I'm doing this with Cadoo . #founders #startups #community

5

3 Comments

My YC batch was remote due to Covid in 2021. That led to 100s of calls with other YC founders. We all had two main issues: 1. Fundraising 2. Growth One affected the other. Without funds it’s difficult to grow. Without growth it’s VERY difficult to raise funds. This is where a well-established personal brand for a founder works like a charm. People trust in people. When trust is established, you can sell whatever you want. You can sell your idea to investors. You can sell your products to customers. Building trust through a strong personal brand will take time, but it is definitely worth it.

30

6 Comments

I spent last week exploring the internals of DragonflyDB - a drop-in replacement of Redis, and it does a staggering 6.43 million ops/second on a single machine 🤯 DragonflyDB is a multi-threaded implementation of Redis with a bunch of insane optimizations. Some of the architectural design decisions and internals of data structure implementation just blew my mind. To understand it better, I went through their blogs, docs, and even some fragments of source code. I compiled all my learnings in a series of videos and trust me you will absolutely love going through it. the videos will change how you look at data structures :) subscribe to my channel, and the video drops every Friday evening. youtube.com/c/ArpitBhayani ps: Until then watch other System Design and Database Internals videos on my channel; they are all no-fluff. #AsliEngineering #DatabaseInternals #SystemDesign

1,040

24 Comments

I've recently heard two different stories that illustrate the stark differences in how AI is affecting the job market: - One startup founder said he laid off two junior engineers. Now, he's involved in coding himself (he has a background in software engineering) and uses GPT/Copilot. He mentioned that product development is much faster now compared to when he had the two junior engineers. - Out of this year’s graduates from a top 50 computer science master’s program, only one person received a job offer upon graduation. One student applied to 60 companies and received zero interview requests—not zero job offers, but zero interviews. It seems like AI is making experienced, senior professionals more effective at their jobs, reducing the need for junior employees. But if no one hires juniors, how can anyone become senior? The job market continues to be very challenging. I hope this is just a short-term, interim phenomenon as we figure out how to integrate AI into the workforce. In the meantime, we might also see the rise of tiny companies, including one-person businesses or small (less than 3 people) startups.

73

12 Comments

Hello Friends. Today, we launch Krut AI on Product Hunt. After all the hustle for the first time, you can sign up use Krut’s trial version and upgrade yourself. This is a massive milestone for us— and we could use your help. Please support, comment, share, etc. A bit about how we thought of building Krut: a year and a half ago, I noticed a challenge: E-commerce SMBs struggle with brand optimization due to disconnected AI image creation processes -- even as they invest deeply into sophisticated brand optimization. So we asked, “Why?” What we realized: today’s decision-makers aren’t well connected with the prompt writing in AI tools to make their decisions right for their brand-building exercise. AI Image Creation lives on prompting island in tools, software, & platforms. But the team learns, decides, & executes in traditional way(photoshoot, content writer, etc) Krut AI brings together AI Image Creation + Traditional Way. We have built an AI-first company to drive business growth, not just creating images that no one looks at. We’re built to eliminate the traditional way of doing photoshoots and content creation, paying thousands of dollars and more than a week for someone to write back to say “What am I looking at?” Because, you know, this is not what you expected. What we’re launching today represents small and medium businesses that had a bad impact with their e-commerce listing due to a lack of prompt writing skills. And Krut is far from perfect. We're putting the finishing touches on things and features that we’re excited to build. We’d love your feedback, good & bad. But, beyond that, we’d love to work with you. We are dedicated to working with some of the world's top and SMB e-commerce founders & operational teams as they use Krut AI to make better decisions, learn, & execute faster. Krut AI will transform businesses and we’re ready to accelerate things for you. I’m so proud of my team Raka RAKSHITH KUMAR KN and MitthunDa MITHUN MOHANDAS. And I’m so thankful to you, our community, for all the support along the way. ❤️ On PH: https://lnkd.in/gxBqevhZ And to try Krut AI yourself today: https://krut.ai/

30

3 Comments

How to ship fast as a small company looking for PMF by Varun Sirinivasan, founder of Farcaster: 1. Avoid side quests. If you aren't coding, selling or talking to users, it's a side quest. 2. Eliminate process. Don't cargo cult your team into sprint planning, offsites or recurring 1:1s. Gather once a day, talk about what to ship and ship it. 3. Hire people who love to ship fast all the time. Most people cannot do this correctly or do not enjoy doing it. 4. Don't hire people who can't ship. It's temping to get an ea, bd rep or community manager to offload work. This usually just creates more work for you. 5. Have a simple and fast tech stack. A new hire should be able to write a feature in a day and deploy it to users in 30 minutes.

22

5 Comments

Do you also Bring Your Own AI (#BYOAI) to work? 🤖 It is safe to say that we are well past the initial days of generative AI where all it led to was confusion more than anything else at most large scale companies. The 2024 Work Trend Index Report from Microsoft & LinkedIn is out and it confirms a paradigm shift into the future, with AI here to stay. However, it dives into some very interesting insights and nuances as well, which I will break down over the new few days. Today, a whopping 75% of knowledge workers use AI at work, with 46% having just started less than six months ago. A large majority of users agree that AI helps them 🔸 Save time (90%) 🔸 Focus on important work (85%) 🔸 Be more creative (84%) 🔸 Enjoy Work From Home (83%) But where does BYOAI come into this and what does it mean? Now AI progressing this rapidly was something not many companies had planned ahead for, leading to very slow & confused undertaking of initiatives pertaining to AI. This is where a lot of employees (78% of AI users) started taking matters into their own hands, using AI tools to get their work done quicker- especially at small to medium size companies! Even more surprisingly, the penetration is across all age groups, not just GenZ as many might presume (see illustration below). With AI enabling rapid change and faster pace than ever before, it becomes imperative for leaders to understand how to best handle and synergize AI with tasks at work, all while still ensuring quality, privacy and security. Is there something you are doing to ensure your workplace gets the best of AI? How have you been making the most of AI and conversely, has it ever gotten you into trouble?! Image Credit: 2024 Work Trend Index Report by Microsoft & LinkedIn #artificialintelligence #ai #aiatwork #generativeai

9

3 Comments

𝗟𝗹𝗮𝗺𝗮 𝟯 𝘂𝗻𝘃𝗲𝗶𝗹 𝗮 𝗳𝘂𝗻𝗱𝗮𝗺𝗲𝗻𝘁𝗮𝗹 𝗰𝗼𝗺𝗽𝘂𝘁𝗶𝗻𝗴 𝘁𝗿𝗮𝗱𝗲𝗼𝗳𝗳 𝘁𝗵𝗮𝘁 𝗳𝗲𝘄 𝗮𝗿𝗲 𝘀𝗲𝗲𝗶𝗻𝗴⬇️ 🦙Llama 3 just came out and to the surprise of many people, the L3-8B version can match the L2-70B one on several benchmarks. Several people also argued that the chinchilla paper (https://lnkd.in/dXaRqyea) was wrong because the llama 3 scaling is very far from chinchilla-optimal. 🤔 In my opinion, this is due to a misunderstanding of what the chinchilla paper is about and more generally of the effects of scale in deep learning. The bitter lesson (https://lnkd.in/d3sVYYgp) has become common knowledge in the field thanks to the GPT family that popularized it. We hear everywhere that "bigger models are better": Yes it's true, but in what sense? Many people have a mental model along the lines of "bigger models are more capable" meaning they would be able to achieve levels of performance that smaller models can't. ❌ This is not true with LLMs and is mainly because we train only once on tokens (most of them at least) and therefore overfitting is not possible. The 𝘳𝘦𝘢𝘭 advantage of bigger models over smaller ones is that they converge *faster*. In other words, you can reach a given level of NLL (modeling performance) with fewer training iterations. ❔ Then a natural question arises: Should I train a smaller model with more or a bigger model with fewer iterations? That's where the chinchilla paper brings a response. They say that the question they are answering is: "Given a fixed FLOPs budget, how should one trade-off model size and the number of training tokens?" (implying "to achieve the best possible NLL"). 🧐 Coming back to Llama 3, it's obvious that the Llama team is aware of this work and there is a reason why they didn't follow this recipe. My guess is that they are taking into account the inference cost in the balance (which is completely out of the scope of the chinchilla paper). When one "overtrains" an 8B model, one ends up with a worse "NLL to training compute" ratio, but also with a better "NLL to inference compute" ratio. ✅ Now if we think about it from a business perspective, it's a very powerful tradeoff because you can choose to buy performance either at a fixed cost (training) or at a variable one (inference). Hope this post helped you! #Llama3 #Meta #LLM #AI

9

3 Comments

AI SOFTWARE GTM FOR FOUNDERS Target unbillable human productivity. Look for structural mismatches where end users have a strong incentive to get more efficient. Selling productivity gains to time tracking business models is generally hard. Buzzwords like "AI agent" and "copilot for X" may help with fundraising but can hurt your email outreach and landing page conversion. Focus on the actual pain points you are solving and use language familiar to your buyers, not your VC! AI curiosity is a double-edged sword. Prospects may be excited to see a demo but intimidated to buy AI systems. Shiny AI can get you the meeting, but a conservative software solution closes the deal. Don't forget that decision makers are human. Products that help companies make more money will be more popular than cutting costs and reducing staff. Be careful in positioning AI as a labor-displacing solution.

28

India’s Jivi MedX has beaten OpenAI's GPT-4. It has also outperformed Google's Med-PaLM 2 in the medical-specific LLMs category. And still, some people will ask where India is in the AI race. Let me tell you the story of Jivi. Jivi is an Indian healthcare AI startup co-founded by former BharatPe Chief Product Officer Ankur Jain. Jivi has scored an average of 91.65 across the leaderboard's nine benchmark categories. But what purpose is it exactly going to serve? It is going to revolutionize primary healthcare through generative AI, making top-quality care accessible 24x7 at a fraction of the cost. Imagine a doctor available for you at any time and at the cheapest cost, with higher accuracy, enabling timely and precise treatment. This makes your healthcare solution error-free. #HealthcareAI #AIRevolution #JiviMedX #InnovativeHealthcare #AIinHealthcare #GenerativeAI #PrimaryHealthcare #IndiaTech #TechInnovation 

124

14 Comments

I’ve been keeping a running list on things I wish I knew or did while working in product before transitioning to founding Inari (YC S23) and here's the 2nd part! Note: this pic is me on our 1st day building in SF for YC. Smiling through how fun all the incorporation paperwork was 😂 Mastering growth and distribution 💌 I never truly internalized the adage “first time founders obsess over product and second time founders obsess over distribution” until we launched Inari for the first time last year and initially heard crickets. I hadn’t worked on growth in any of my prior roles and now I probably spend 60%+ of my capacity catching up, learning basic tactics, and experimenting with new channels for Inari. I think we could be much farther along if I had mastered in-product growth and activation tactics, had reps setting up outbound and SEO systems, and built a community and audience much sooner (I also never realized how fun engaging with people online could be 😄). No excuses not being “technical” in 2024 💻 If you’re considering working on a SaaS startup and you can’t code, just go and learn to. With the recent advances in LLMs and coding assistants, there’s literally never been a better time to learn to code, unblock yourself when you run into issues, and build an MVP. There’s nobody on the planet that will care about your product as much as you will. If you’re unable to be deep in the details, identify defects, and make fixes quickly yourself, then the overall clock speed and quality bar of your product will suffer. The quality expectations for ALL startups have increased over the last few years and it’s on you as a founder to keep that rigor and velocity extremely high, so being able to unblock yourself makes a world of difference. Managing your own psychology 🧠 Surprisingly, the toughest part when transitioning to founding was managing my own mental state. There’s a constant internal uncertainty on whether you’re building something that people actually want and spending time on the right priorities. Your view on finances also completely changes. Money used to be a resource I spent on fun discretionary spend but now money feels like “oxygen”. The stakes of each day feels much heavier when you feel like you’re running out of oxygen for your startup which you're pouring your heart into. Overall, be prepared for these internal influences before making the jump! That could be better carving out the right runway, validating your idea or building an MVP before leaving your job, or bumping up your skills around building and selling beforehand.

80

7 Comments

One thing that working directly with LLM’s teaches you is that indeed, an LLM is just a computer file. A doc. A file. It can do damage. Think of its pedigree. A usefull but limited LLM currently is going to run you 4 gigabytes. If someone tries to send you a 4 gigabyte Word doc, face it, your day is not improving. PowerPoint was used to start a war. Files contain viruses. It is a serious issue. It’s not the file that is going to destroy us. I have been seeing people saying “AI superiority” in reference to an intelligence “arms race”. It’s scary. That kind of thinking will destroy us well before any file will. Injuries will be self inflicted. It’s not the file. A file that contains encoded copies of all of our knowledge is going to be big. It’s also immensely stupid to try to cram it all into one file. We are using intelligence to do stupid, make stupid and ultimately make stupid acceptable. But more importantly today, time is not your enemy. It is your gift. Regardless if AI is helping you save time or not, you should definitely be learning more about it and how it works. #files #productivity #nowar #agi #ai

1

Our founding team spotted an untapped product opportunity in a fresh market Hiring a new team to build out this product was not feasible We had to make the call Our leap? Shift our entire engineering team to work on this new product Over a single weekend! Meanwhile, I was the new kid on the block in my role Still carving out my niche But I took on the challenge My approach was straightforward: - Dive deep into discussions with my team, our GTM team, and founders (To grasp this opportunity fully) - I needed to convince myself first (The risk/reward had to be clear) - I confided with 4 veteran engineers (Who poured their hearts and souls into our existing product) 3 major worries emerged: - They saw a big risk to the company by putting all eggs in this "new basket" - They lacked the tech skills needed to succeed in the new product - They were worried they will keep serving two masters: 1. Old Product 2. New Product I decided to address these concerns head-on… Next steps? - I pressed our CEO for an extra week before rolling out this change - I organized small, strategic meetings Started with team-level discussions Then escalated to full engineering team briefings - I presented the potential and risks with utter honesty and vulnerability - I shared how I would support And take the tough journey alongside the teams The results: - We completed the major pivot in a record 10 days! - I earned the trust and credibility of the team, Turning a critical corner in our company journey P.S.- I am looking for my next opportunity If you think I can make a difference in your organization Drop me a DM!

39

4 Comments

There was a time when people paid $200-$300 for a single prompt. Now, LLMs themselves are better prompt engineers than humans. It’s interesting how quickly things change in less than a year. LLMs went from small 7-50B parameter models, all the way up to +1T parameters (parameters determine how “smart” the LLM is). With it, the bar for writing good prompts has dropped to nothing… P.S. For those shaking their heads, yes, LLMs in production still require significant work. Open-Source LLM prompting remains tough and requires thorough testing.

10

2 Comments

Are you a founder using Notion/Airtable/GSheets as your ATS? I'll give you $100 to talk to me :) Context: Dover's building the ultimate ATS for startups. We get a lot of companies switching from legacy ATS's like Lever, Greenhouse, and Workable, but the most common ATS companies switch from isn't even an ATS at all – it's Notion, Google Sheets, and Airtable. These tools are great (we use all 3 at Dover!) since they're easy and fast to set up and get some basic organization. But you quickly run into problems once you reach any sort of hiring scale. - Can't post to job boards - Can't ingest candidates from job boards - Can't sync your email correspondence to each candidate - etc etc I suspect people like these tools because they're familiar and they don't want to learn a new product. Anyways, if your company is still hiring using Notion, GSheets, or Airtable as your recruiting infrastructure, comment here on what you use + send me a DM. Once we chat on the phone, I'll send you a $100 Amazon gift card.

22

9 Comments

I wrote a quick blog post: "Should you ditch notion for a real ATS?" Check it out here: https://lnkd.in/eytPC5pr

10

3 Comments

For this week’s episode of The Lobster Talks podcast, I had the privilege of going behind the scenes with Laduram Vishnoi, the founder and CEO of Middleware - a pioneering observability platform for cloud-native applications. Laduram's entrepreneurial journey is nothing short of unconventional. He shares candid details about bootstrapping Middleware by self-funding and building a team in India, allowing him to kickstart development without burning through millions. Even more impressive? Raising an outstanding $6.5 million seed round amid the chaos of the Silicon Valley Bank crisis. But that's just the tip of the iceberg. Laduram also dives deep into: -His unique strategies for managing a distributed workforce across multiple timezones -How he persevered to get into Y Combinator after 5 previous rejections -Actionable advice on nailing the fundraising process as a solo founder What truly sets this conversation apart is Laduram's vision for leveraging cutting-edge AI to transform observability and tackle skyrocketing cloud costs head-on. I was blown away by the insights and inspirational story shared in this episode. It's a must-listen for any aspiring entrepreneur or anyone passionate about innovation in cloud technologies. Check out the full episode here: https://lnkd.in/eqawwxcX And let me know - what's your biggest observability and monitoring challenge in today's cloud environment? Share your thoughts in the comments! #startups #entrepreneurship #podcast #ycombinator

40

1 Comment

Build/learn in public type of post that I shared on LearningLoop.com: We want different networks on LL to have their own membership criteria so that you can only join one if you fit its criteria (e.g. to join LL for tech startup founders you must be a tech startup founder now, to join LL for first-time parents you must be a parent in the first 24 months of having a kid now, etc.). Why? To maximise relevance of every member to every other member. Why? The less relevance between members, the less members talk and the less engaged LL becomes. Why does it matter? Well, we're building a consumer app. We have to design for scale & engagement. Example: If we put someone who just became a parent (their first child) in a group with other first-time parents who also just had their first children, for many months to come they'll have things to discuss on LL, whether emotional stuff or questions that they have. Once their kids are older than 2-3 year old, the parents will probably not wanna talk as frequently with parents who just became a parent. Human attention always moves to the next most important thing in one's life. Grouping parents of 1 year olds with parents of 10 year olds to have daily/weekly conversations will probably cause friction because the parents of 10 year olds may get impatient and perceive the younger parents as noobs or overthinkers. Maybe 1-10% of people are infinitely patient and but the majority aren't. (I'm personally not and if someone asks me daily about something that i mastered few years ago, I get really annoyed and super avoidant, and I've seen this pattern as one goes up the competence hierarchy in any domain). There's a reason for classes and levels and progression at school. You get matched with people at your level (by some definition). Since we're building a peer to peer learning app that you can use both as a social network and as a learning app, we need to build an environment in which every individual (the majority) can be absolutely selfish and optimise for their self-interest, but in doing so, they also happen to be helpful to others. Meaning, members are not on LearningLoop to teach others, but to grow themselves, but in doing so (and sharing their questions and learnings and feelings and thoughts) they also grow others. When members move on, they leave behind a lot of wisdom that new members can benefit from (assuming we do a good job of capturing that wisdom and making it available easily) Scaling curation is one of the 2 fundamental technical problem statements of our company and it's deeply rooted in what's best for the business

14

1 Comment