Lenny Rachitsky

The discussion with Jason Fried, Co-Founder and CEO of 37signals, during a recent ProducTea with Leah episode (#42) was great. A particular one-minute segment (~14:20) really stuck with me. When asked about failures, Jason states: 🗣 “I don’t really think about failure. How does evolution work? Evolution is not sitting there using energy thinking about the wing shape that didn’t work. The wings that worked let’s make them even better...What’s working? Let’s do more of that. Let’s evolve.” This aligns perfectly with the concept of “slightly better problems” from The Subtle Art of Not Giving a F*ck by Mark Manson. It’s a never-ending upward spiral. The key takeaway? Don’t ignore failures, but don’t dwell on them either. Focus on what works, refine it, and evolve. This approach ensures continuous improvement without getting bogged down by setbacks. #ProductManagement #Leadership #ContinuousImprovement

5

Vox Populi, Vox Dei! 🙌 What should we build next? With our public roadmap, now YOU can decide! 👀 View current, planned and backlog projects that Speakeasy is working on ↕️ Upvote and downvote for your favorite projects 📥 Submit new ideas and feature request Just make sure you do it fast, because our team at Speakeasy ships quickly! 🚢 Check it out in the comments 👇 Also a shoutout to Raphael Fleckenstein and team for shipping a very useful product! #roadmap #sdks #linear

22

6 Comments

Product for Engineers dropped another awesome post recently: How we decide what to build. It’s a beautiful building-in-public look at how PostHog goes about product development. There is so much to learn from this newsletter. I love it. https://lnkd.in/ewNBSpu4

60

45 Comments

Raising $1.5M as a solo woman founder is not an inspirational story; it’s a practical one. It's less about fairy-tale success and more about the subtle art of persistence, resourcefulness, and knowing how to make Google Docs do what entire dev teams are paid for. This cheat sheet isn't advice; it's a survival manual, polished by exhaustion and necessity, for those navigating their own absurdly convoluted journeys in building and funding a startup.

11

1 Comment

"Probably the worst decision that I made as CEO of Loudcloud was to take ona. 10-year lease with a $30M letter of credit. We leased the space right before the dot-com crash. Rents dropped 10-fold in Menlo Park... worse yet, due to the recession, we couldn't hire so the space we already had was sufficient and we never ended up moving into the new building. So I basically lost $30M for nothing." This is from Ben Horowitz's recent interview with Big Think. Even a quarter century ago the typical office lease construct is misaligned with how a company needs to operate. Our industry could get away with this when office spaces had a monopoly on how and where work got done, but those days are dead. As global markets are crashing, we'll be looking at another disruptive cycle for businesses which will create more cost cutting pressure and the first thing to go when they roll will be inflexible leases. CRE needs to develop new offerings and fast. https://lnkd.in/ePuZgzkU

20

2 Comments

It's Day 1️⃣0️⃣ of Building in Public. Today brought more progress on Kounslr's course recommendation engine -- we're working hard to wrap this feature up. We are super close to finalizing the feature and pushing it, but keep running into small roadblocks in our pipeline. The current issue we are handling relates to one of my previous posts -- reading in course data for a school/county. There are so many exceptions that a course can have, and it is almost impossible for us to search through or account for all of them. We are currently going through a lot of trial and error to try and see what we can do. 🛠️ As I wrote about previously, the current approach is a Python script which organizes data to fit into our Firebase structure (attached below). However, if we cannot seem to handle every nuanced piece of information that we are given, an LLM might be the way to move forward. If anyone has other suggestions, they would be greatly appreciated. 🙂 May 20 2024 3:32pm Ashburn, VA #startup #productdevelopment #python

26

2 Comments

𝗪𝗵𝘆 𝗣𝗮𝘁𝗵𝗙𝗮𝗰𝘁𝗼𝗿𝘆, 𝗙𝗼𝗹𝗹𝗼𝘇𝗲, 𝗠𝘂𝘁𝗶𝗻𝘆, 𝗮𝗻𝗱 𝗛𝘂𝘀𝗵𝗹𝘆 𝗔𝗿𝗲 𝗦𝗶𝗺𝗶𝗹𝗮𝗿 𝗯𝘂𝘁 𝘄𝗶𝘁𝗵 𝗧𝘄𝗼 𝗞𝗲𝘆 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀 As the co-founder and CEO of Hushly, I’ve spent a lot of time looking at how our platform compares to others in the B2B 𝗠𝗮𝗿𝗸𝗲𝘁𝗶𝗻𝗴 𝗣𝗲𝗿𝘀𝗼𝗻𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 space, like PathFactory, Folloze, and Mutiny. Each of us competitors emphasizes a different message: PathFactory emphasizes 𝗰𝗼𝗻𝘁𝗲𝗻𝘁 𝗶𝗻𝘀𝗶𝗴𝗵𝘁, Folloze emphasizes 𝗰𝗮𝗺𝗽𝗮𝗶𝗴𝗻 𝘀𝗽𝗲𝗲𝗱, Mutiny emphasizes 𝟭:𝟭 𝗽𝗲𝗿𝘀𝗼𝗻𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻, and Hushly? Well, we emphasize a 𝗰𝗼𝗺𝗽𝗿𝗲𝗵𝗲𝗻𝘀𝗶𝘃𝗲 𝗼𝗳𝗳𝗲𝗿𝗶𝗻𝗴 with our three-in-one platform. But here’s the thing: just because each on us emphasizes a different area, it doesn’t mean each of us don’t offer strong capabilities in the others. For example, just because PathFactory highlights content insight, it doesn’t mean the other three—Folloze, Mutiny, and Hushly—don’t offer similar or even better content insight tools. The same goes for speed, personalization, and efficiency. At the end of the day, we’re all aiming for the same thing—helping marketers deliver more personalized experiences that drive results. That said, there are two key differences that set Hushly apart. 𝗧𝗵𝗲 𝗙𝗶𝗿𝘀𝘁 𝗞𝗲𝘆 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲: 𝟯-𝗶𝗻-𝗢𝗻𝗲 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 Now let’s get into what makes Hushly different. We’re the only platform that brings together a 3-in-one solution. I like to think of it like this: Hushly operates cars, boats, and planes, with each vehicle specialized for its own terrain—land, water, and air. In more practical terms, we offer a website-personalization platform, a content-experience platform, and a demand capture platform, all in one seamless system. By comparison, PathFactory, Folloze, and Mutiny each have their own individual platform, but Hushly offers a unique blend of all three capabilities, unlike other platforms. That makes our offering broader and more flexible. 𝗧𝗵𝗲 𝗦𝗲𝗰𝗼𝗻𝗱 𝗞𝗲𝘆 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲: 𝗖𝗮𝗽𝗶𝘁𝗮𝗹 𝗜𝗻𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝗲 Another big difference? What happens after you purchase the software? Hushly is the only platform that’s completely capital-independent. While other platforms have investor backing, which fuels their growth, Hushly's capital independence allows us to make decisions solely in the interest of our customers. 𝗪𝗿𝗮𝗽𝗽𝗶𝗻𝗴 𝗜𝘁 𝗨𝗽: 𝗦𝗮𝗺𝗲 𝗚𝗼𝗮𝗹𝘀, 𝗧𝘄𝗼 𝗞𝗲𝘆 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀 At the end of the day, PathFactory, Folloze, Mutiny, and Hushly are all B2B Marketing Personalization Software platforms with nearly identical functionality. But if you’re looking for a platform that combines website personalization, content experience, and demand capture all in one, and one that’s truly customer-first, Hushly stands out with these two key differences. Visit us at https://lnkd.in/esKX_A5b to learn more!

30

2 Comments

Innovation for doers: Focus on Impact, Not Jargon Your customers don't care if your innovation is labeled as disruptive, adjacent, or incremental. They won't ask which "Horizon" it fits into or whether it aligns with the latest innovation framework. It's time to cut through the noise. What matters most is the tangible impact your innovations have on their lives and businesses. They care about solutions that solve real problems, enhance their experiences, and drive meaningful change. So, let's shift our focus from terminology to transformation. Instead of getting bogged down by jargon, concentrate on delivering results that truly matter. 🔍 Stop overanalyzing. 💡 Start making a difference. Your next breakthrough could be just around the corner—if you let go of the labels and embrace the possibilities. #Innovation #CorporateInnovation #BusinessTransformation #InnovationStrategy #ImpactfulChange #venturebuilding #CVC #openinnovation #accelerators

19

We've gotten fine-tuning down to the point where it can just take minutes, not days. And we've just released the details on how to the public. A few months ago, our normal fine-tuning tasks took a dozen hours or more and it wouldn't be uncommon for training to run for multiple days. That kills iteration velocity and increases customer churn when someone needs to come back 24h later just to test out their model. Today, fine-tuning is 15x faster meaning the jobs that took half a day previously are now running in less than an hour. We've now published the details on how we got there in one of the substantive blog posts I've seen on fine-tuning best practices, from our very own Arnav Garg. It's full of best practices like: ▶️ How switching our optimizers saved us nearly 40% in time. ▶️ Why disabling gradient check-pointing with our other memory optimizations was able to save us 25% – 30%. ▶️ The other secret sauces that go into a great fine-tuning experience, from sample packing to custom cuda kernels. If you fine-tune models or are getting educated on the space, I think this is a must-read. https://lnkd.in/gxJg8ERW

61

1 Comment

"Do you think I should be a founder?" is a question I get asked a lot. It's #SeattleTechWeek and I'm seeing great quotes and advice being shared, so I wanted to share an early story before we had any funding. I had been trying for our initial funding for months. I was learning, we continued to work on the product, marketing, and investor pitch. We were improving but it was still one rejection after another. I will never forget the moment after a pitch that went particularly poorly. I found myself staring at my calendar realizing I didn't have anything next. No more pitches, no meetings... just nothingness. I had called every favor, exhausted every network connection I had, and I felt no closer than the day I quit my job. I opened my slides to iterate and look for improvements, but what was the point? Onward, let's go. There is always the next person you can approach. This person I didn't know well, he was a tech exec I'd worked with previously. I had a lot of respect for him, and noticed he was connected to an investor that I hadn't reached. Requested an intro and the he told me "Ozan... you have no customers, no revenue, and your team is quite small. Unfortunately there isn't enough here for me to justify the introduction." It seemed reasonable enough, and I thanked him for his time anyways. I thought... maybe asking for the intro was disrespectful. The next day I was sitting at my kitchen table across from my advisor and mentioned my failed attempt at securing the intro. I'll never forget him saying "Hmmmm, I don't know that person, but let me say something. Fuck that guy, fuck him. That's such a load of bullshit! You've quit your job where you were a top performer, you're putting everything on the line, you're dipping uncomfortably deep to the end of your savings, you're out there hustling with very little support, and he can't even do the bare minimum? No that's not reasonable, that's bullshit." It didn't matter whether or not the person was justified in not making an intro. It didn't matter if my advisor was biased towards my perspective. What did matter was that it was an extremely dark week for me, after so many rejections, with what seemed like no light at the end of the tunnel, and it reminded me that I'm not always wrong. "Do you think I should be a founder?" is therefore such a hard question to answer. It was mentioned this week that startups are hard and they're not for everyone. I don't say it, but the first thought in my head is always "If you have to ask, then probably not." It's a long, lonely journey. It's filled with dark days, doubt, very little support, and more often than not it doesn't feel logical. But - there are some amazing moments as well. When I have to sum it up I mention that it's an endless string of high highs and low lows. If you do it, the number one thing I'd say is never ever stop asking for favors. I'm still asking for them 6 years later, and the answer I usually get hasn't changed either. Push.

97

15 Comments

I'm very worried about the current cohort of stream processing startups. No one wants stream processing to succeed more than me, but I fear there aren't as many use cases as we'd hoped. In 2015, the thesis for stream processing was that because people are collecting their data in real-time, they'll obviously want to process it in real-time, too. But that wasn't happening. We believed the problem was the programming model—perhaps it was too limited. So we worked for a few years on on making the APIs more expressive. Well, the tech got better, but the adoption still wasn't coming. We thought perhaps the tech is just too complicated. So we worked on making it operationally simpler. I think we're there now, and mainstream adoption still looks uncertain. The reality is that stream processing's main benefit, lower latency, just isn't that critical outside of a handful of mission critical use cases. As someone who's invested a huge chunk of his career in stream processing, I would love to be wrong about this. But I do think startup-land needs a different strategy.

405

145 Comments

Very energized by the customer workflows I’m seeing incorporating Zapier core, Tables by Zapier, Interfaces by Zapier and Central by Zapier! Whole applications and ecosystems are being built and they are all powered by Zapier! If you haven’t heard of the solutions above, you’re not late but folks are certainly bush whacking no code and it’s just a whole new category of automation! #zapier #builders #automation #saas #central #tables #interfaces #chatbot #zapconnect #automationforeveryone

21

1 Comment

In the last 18 months of building Clappit we have tested, broken and rebuilt the product extensively to rethink how we can impact engineers at scale. Today we move with a calm confidence into the journey and pivot of truly serving world class companies. The conversations in the last few months and ahead are leading us to the center of the cloud and AI world. For those who are in the cloud space, there are exciting times ahead with the added cherry of AI on top transforming the way things are done from here on.

12

There is a lot to chew on in this great article by Maggie Appleton (Maggie Appleton), called Barefoot Developers. About how large language models might enable new kinds of software, free of the demands of industrial scale, unicorn-ness, and venture capital. Software that might be made for, say, a dozen people. If you are involved in software, this is worth your time to read. https://lnkd.in/eZA_uVYf

4

2 Comments

I'm using my own product to scale intros in our community 🎉 When I created Britepeer, I was solving a problem for a community a few of us made called StartupNash. StartupNash is really good at getting everyone on a chat and talking about the everyday real-world challenges of being an entrepreneur in Nashville. But once everyone is in there, then what? It takes A LOT to make sure the right people are meeting each other. Britepeer does the job of connecting peers in the group. It scales the intros for me! Some recent asks in our group: 🤔 Someone is looking for a professional with experience dissolving a TN registered LLC. 🤔 Someone is looking for help with SEO/Analytics. 🤔 Someone wants to learn more about the VC community in Alabama. There are peers in StartupNash who can help these folks, but I'd love to add more. So I redid StartupNash's website and put Britepeer, front-and-center. I want more people to get in the group, use Britepeer, and meet founders, funders, peers, mentors, and services. Join our community and use Britepeer to help you get connected. And if you're a community manager who faces the same problem, come talk to me!

9

1 Comment

Here is what I think of RAG: you want to ask a question about a book for example, but you tear a few pages off (retrieval) and give it to the LLM and ask the question (generation). Long context is nice (and slow), but it forgets the book as soon as it finishes reading. Continued pretraining was supposed to help to help transfer knowledge into the LLM, but nobody seems to have been able to make it work well. But the recent work from Tatsu Hashimoto's lab at Stanford on "Synthetic Continued Pretraining" gives me a lot of hope: https://lnkd.in/gRUkhCwd. Here is the crux of the idea from the paper (their method is called EntiGraph): "EntiGraph first breaks down a text corpus into a list of entities and then uses a language model to generate text descriptions about relations among the extracted entities, iteratively “filling in” the knowledge graph underlying the corpus." They get the model to understand the corpus better, and so doing RAG on that beats the vanilla base model + RAG.

128

6 Comments

"I've talked to Founders who have been in LA for years and never felt like they had a sense of community within like a group of other Founders who were doing the same things." In this episode, we sit down with ⁠Zack Oliver⁠, the innovator behind ⁠Founder's Table⁠, a community initiative that connects LA Founders over intimate dinner settings. Zack shares his journey from finance in New York to creating a network in LA through curated gatherings, aiming to foster relationships and exchange knowledge among local entrepreneurs. As Zack details the evolution of Founders Table from casual meetups to structured, meaningful interactions, listeners will gain insights into the significance of community in the entrepreneurial landscape. This episode of Founders In LA⁠ explores how personal connections can amplify professional journeys and the role of shared experiences in overcoming business challenges. Thanks again to our sponsors: Nearsure Unità Prodhub Check out links to the full episode in the first comments. ---------------------- Follow Ethan Cole PhD for content & insights on Building Faster and Smarter; new trends in Product Development, the latest the Talks and Workshops from TED AI, and to learn more about Founders in Los Angeles.

20

3 Comments

I recently talked with Paz Yanover, VP of Product at Amplication, about creating and managing effective docs. Here is his developer docs success formula (check 1st comment): In this interview, Paz shares his formula to creating developer docs that drive product adoption. You'll learn: - The important role docs plays in support and attracting devs - Tips for balancing technical accuracy with user-friendliness in your docs - Strategies for effective collaboration between technical writers and product teams Check the first comment for the full interview. --- Enjoy this? ♻️ Repost it to your network and follow Derick Ruiz for more.

19

1 Comment