I sold my first startup before I turned 20. Since then, I’ve scaled businesses to $35M ARR, sold 2 startups, and built 3 micro SaaS products to $10K MRR in under a year—while going through YC & TechStars. Now, I’m building BodyPro, @getdewey , @2SlashAI , and Content Zen. Over the next 3 years, I aim to launch 5 more products, grow to $50K MRR, and sell 2 more startups. Follow me for lessons on launching quickly, finding product-market fit, and scaling smart 🚀

The day has come... Instagram saved posts are now LIVE on Dewey. Plus, bsky animated gifs, qts, tiktok galleries, pdf improvements, and more... Let's dive in:

How much does it cost to hire someone to create a simple MVP for an idea these days? I am talking about authentication, an API or two to connect to and a few simple services, with UI.

Shutting the Door. After 10 years of building companies and experiencing a mix of emotions, I'm at a major crossroads. I've decided to close my latest venture (unless something significant happens by month's end), and I'm debating daily: return to the steady, familiar corporate career with its comfy salary, or roll the dice again with another startup? I reflect on my passion for building and the unpredictability of starting anew. What I know for sure is that whatever happens next will bring new, exciting adventures.

Do we trust AI too much? Alphabet, just dropped its longtime “Don’t be evil” promise on AI for weapons or surveillance—citing “national security.” But with investors eyeing profits, you have to wonder: did they do it because they need more money? Meanwhile, experts like Stuart Russell warn that autonomous weapons could ignite a global arms race. Some data: Alphabet’s shares fell 8% at the open, and the tech giant plans to pour $75B into AI expansion next year. With $96.5B in quarterly revenues, Google’s influence is massive—and so is the risk. Not sure if this is suppose to make us feel better or worse about AI.

So many choices, so little time. With the release of o3, I felt overwhelmed by the number of ChatGPT models available. I decided to run a few tests to better navigate my options. If you’re on the paid version, you’re currently faced with 7 different models, each boasting descriptive features like “Advanced Reasoning” (ah, now I know when to use it 😦). However, there’s no explanation of what these features mean for my daily tasks. Instead of a seamless experience where the system picks the right tool for your prompt, users are expected to become technical writers for OpenAI, deciphering each model on their own. About 90% of users stick with the familiar GPT4o—the default ChatGPT option they know and trust. The remaining 10% decide to risk it all and try the o models. These o models (Reasoning) offer an edge: they are faster and provide smarter responses that push the boundaries of what we thought was possible with AI. Here’s my rationale for choosing each model: 1. o3-mini-high: My default model for most tasks. It’s fast, reliable, and allows about 150 messages per day before it locks you out. 2. GPT4o: Essential for features like file uploads and image generation. 3. o1: Handy for coding tasks in Canvas mode, where it still performs well despite its strict 50 messages per week limit and its turtle-like speed. One final tip I discovered today: you can switch models mid-conversation. For example, start with GPT4o for file uploads, and then switch to o3-mini-high for your other queries.

What do poker and startups have in common?Everything. Back in 2012, I packed up my life (for 3 entire months) and moved to LA with a head full of dreams and a heart set on becoming a professional poker player. It was crazy leaving behind the everything to chase such a stupid dream. But looking back, it wasn’t just about the cards for me or the quick cash, It was about the grind, the strategy, the highs and lows, and the thrill of building something from nothing. Years later, I see just how much poker is similar to building a startup. Both are games of calculated risk. In poker, you don’t win by playing every hand, you wait for the right moment to play. Startups are sort of the same, you can’t chase every idea or opportunity, you need to know when to bet big and when to fold. Both require resilience too. In poker, you’ll lose hands you think you should’ve won, but it’s not about one particular hand, it’s all about the long game. Startups? Same story. Failure isn’t the end, it’s just part of the process and the daily grind. Moving to LA (closer to Las Vegas) felt like stepping into the unknown, but that’s where growth happens. Whether you’re sitting at a poker table or pitching investors, success comes down to reading people, trusting your instincts, and constantly learning. Chasing my dream of becoming a poker player taught me two important lessons: the first, I am not a great poker player, the second, I love the feeling of a good chase.

Looks like ChatGPT pulled a classic American move, outsourced its job to a cheaper Chinese AI knockoff.

AI is going to push the stock market today, down, way down. DeepSeek claims to have trained a top-tier language model for just $6M using outdated H800 GPUs, this is compared to the hundreds of millions (or billions) spent by giants like OpenAI and Google. DeepSeek say they pulled it off in just 3.7 days with 2048 older GPUs, hardware that’s significantly weaker than the cutting-edge H100 GPUs others use. H800 GPUs have half the performance of H100s in bandwidth. Assuming DeepSeek claim is true, it’s a 99% cost reduction compared to the massive budgets of OpenAI and Google for models like GPT-4 and Gemini. That’s a potential game-changer for the entire AI ecosystem, especially for startups that rely on software development to bridge generic models with specific business applications. This changes the entire AI value chain: from the energy demands that power automation to the infrastructure behind it, like Nvidia GPUs, and finally to the software that most startups are so heavily focused on. Lower hardware costs could level the playing field towards smaller players, but it's not clear if DeepSeek’s numbers are true or not. So taking all of that into account, the major players in the GPU manufacturing are going to drive the market down, at least for today.

Decided to test out OpenAI launch spree, and the results are ... Yesterday, two major upgrades were made to ChatGPT: 1️⃣ Added Artifacts creation in the Canvas feature (copied from Claude). 2️⃣ Connected the Canvas to their most powerful model so far, o1 (o3 in the works). To be honest, by itself they aren't super impressive, but combined they are fairly cool, allowing you to create Artifacts using o1 (arguably the best model ATM). There is a small catch (as always), the $20 per month version limits you to 50 o1 messages per week, which also limits what you can create with it. To create bigger Artifacts (more messages) you will need the $200 per month version. The $200 version also gives you access to the new Operator agent (browsing the web), o1 Pro (slightly better than o1), Unlimited access to Sora’s advanced mode, and soon o3 model. I decided to give it a try and built a baby name generator, initial version wasn't great but few prompts later it became good. I also made it pull the names out of a baby name site. I recommend giving it a try for those that have the Plus ($20 per month), here is a quick how: 1️⃣ Select O1 mode in the top-left corner. 2️⃣ Click the toolbox icon and choose Canvas. 3️⃣ Enter your prompt to create something. 4️⃣ Hit Preview in the top-right after it generates the code. Bottom line: OpenAI is continuing to push the limit with what you can do with AI, making it easier to create things quickly, however, so far the pricing is a little steep for most, but the power of these new features is hard to ignore.

My SaaS got attacked and made $550 MRR from it. Today, I woke up to the sound of notifications jumping on my phone. I opened it and saw that I was getting subscription purchases on Dewey left and right. I scratched my head and tried to understand—did Lady Luck just knock on my door? Then I remembered my luck sucks, so it's probably not it. I started looking around and saw that purchase after purchase all came from users who just signed up, never used the product, and when trying to purchase, they went through dozens of credit cards until one went through. It was a very weird attack—are they trying to attack my SaaS for me to make more money? Why go through the trouble? Send me an email and I'll happily send you a payment link. Eventually, I decided to refund the money and close these accounts, but I still wonder, what was the point of this attack? Just checking a bunch of stolen credit cards? Lesson learned: @getdewey is so good, people don't need to try before you buy.

Just in case you want to be ready for TikTok doomsday: techcrunch.com/2025/01/22/de…

TikTok is now live on Dewey The U.S. government almost shutdown TikTok on Jan 19. Now, they have 75 days to figure out a solution. And you have 75 days to protect all your TikTok Favorites and Collections:

Doing some testing to see conversions on various product launch platforms and identify what drives the highest quality traffic. Any recommendations besides Product Hunt, of course?

TikTok goes dark, with a glimmer of light The ban on TikTok is somewhat weird , but even more strange is the fact that ByteDance not interested in selling TikTok, currently they lost 20% of their user base - basically decreasing the value of TikTok. I’m not too thrilled about the Chinese government controlling a massive social platform, even if I’m not really convinced it’s used for anything malicious. Meanwhile, creators and businesses that rely on TikTok have lost their revenue completely. Still, we probably won’t see any huge gap in the space - other apps can will and already are filling the gap. As for TikTok, they’ll likely are waiting to see what Trump will do hoping for an extension, or a fresh deal. It's like a tech casino, where the house always wins.

Calculating valuations is science? science fiction. I’ve raised money for several startups both my own and helped others. Each time, I keep understanding the same thing - valuation is nothing but a guess. Investors talk about multiples, TAM, and comparables, but it’s about throwing a number out there and seeing if you can get someone to agree to it. I sometimes get questions about how to calculate valuation. My advice is simple, grab two dice. Roll them, and whatever number you land on, that’s your new valuation. Treat valuation as a flexible negotiation, not a scientific formula. Don’t lose sleep trying to find a right number, there is no right number.

From Criminal to Entrepreneur I was born in Ukraine and left when I was four years old (thank you Chernobyl). As an immigrant I grew up in poverty, I became fascinated with earning money from a very early age. By eight, I was tinkering with BASIC (a very old programming language), at the age of ten I launched my first “business” reselling slime-like figures. I’d buy them cheap at a local market and flip them for double the price outside my school and several others (100% margin - most profitable business I ever built). By the age of eleven, I started working at a computer store, assembling computers. At thirteen, my life changed, I found out that to make more money, I can do things that are less "legal", I’d joined a hacking group, but I didn't just hack for fun, I spotted another business opportunity - selling ripped games. That turned a real business, with nine sales reps (all high schoolers) and one “CFO” in charge of money collection. Before I knew it, I was pocketing about $8,000 a month. Not bad for a kid barely old enough to drive. But at fifteen, I decided to take a turn and got involved in car theft, something I’m not proud of. After about three arrests, I decided that this is not the path I want to take. But I learned a valuable lesson about myself - I had a gift for making money, uniting teams, and spotting opportunities. That’s why at eighteen, I finally started my first legit business.

Never raise money. I’ve pitched VCs enough times to know it's the most annoying work in the life of a founder, and for complete transparency I was successful in getting funded more then once. When VCs do come aboard - everything goes faster, right? More money, more resources. However, the time and energy it takes to get them on-board, the horrible feelings in the process and finally the new "partner" you gain is sometimes, not great. In the last couple of years, I chose to build some bootstrapped SaaS products. Yes, I still built another VC backed business but I love working on my products. My time, my vision, and answering to… myself - because I am an awesome boss. At the end of the day, startups can survive without VCs, but VCs can’t survive without startups. So yeah, you can keep the designed pitch decks and huge rounds, but it's not as fun as building a bootstrapped product to $50K MRR.

When I first started building a team in my previous company, I made one simple decision - hire people I know and trust. My team was made up entirely of close friends (ofc all professionals). During the process I learned a few things, both good and bad. The Good: - Your friends will have your back, no matter what. - They’re the first to go the extra mile because they genuinely want you to succeed. - Communication feels natural, and you build a team culture fast. The Bad: - Navigating disagreements is a challenge. - Balancing personal and professional relationships, what to discuss and when. - When things get tough (and they will), how do you give honest feedback without hurting close friends? Looking back, working with friends was both rewarding and tricky. The key is setting clear boundaries and keeping communication open - separating work problems from personal ones isn’t always easy. In the end I will probably do it again - just maybe more cautious about who the friends are.

My first startup wasn’t in SaaS or software. It wasn’t even in tech, really. It was hardware. A friend and I came up with a simple ideaת a small part for electrical sockets that could extend them without needing bulky extenders. Think of it as a cleaner, more aesthetic solution for something no one had really thought to improve. My friend was a hardware engineer but for me stepping into the hardware world was completely new. Eventually I dove in, helped design the part, and somehow we managed to build a "working" prototype. Here’s where the story gets crazy. Before we even figured out what to do next, a well known corporation approached us. They wanted to buy the patent. We were young, very unexperienced in navigating such situations, so we sold it. A seven-figure deal. Life-changing money for us at the time. For years, we wondered if we’d made the right call. What if we’d held on longer? Could we have scaled it ourselves? Now, with more experience under my belt, I see how much luck played into that moment. We had no distribution, no strategy, and no clue how to grow it further. That offer was pure timing, and honestly, we were lucky to get it. Two things I took from this: Sometimes, the best decision is taking the win in front of you. Even the best ideas need execution, timing, and a lot of luck to succeed. Looking back, I wouldn’t change a thing. That little hardware project kicked off my entrepreneurial hunger.

We just dropped a ton of new features to kick off 2025 From Threads to Truth Social, Date Filters and a revamped Export Pass Let's dive in: 👇