Lazar Jovanovic 🤖's Linkedin Analytics

Day 4 of #vibecoding my new project, #16 out of 50 for the year. And it starts with a ba...ups, it's a *bug* after all! And in *bed* too. And yes that is a pink colored bed cover, courtesy of my girlfriend. 😂 Just because it started this way, doesn't mean it won't be a *banger*! Lots of words that start with *B*... is this the one project that makes me a billionaire? 💡😂 Stick with me for a few more early mornings and nights and witness it yourself!

Vibe Coding requires the right vibes 😎 Especially on Day 1 of my #50in50Challenge week 16! That's where it starts - the ideation 💡 What should I be building this week?

One of the most interesting experiments I have seen on prompting 🤠❤️

THIS IS ONE OF THOSE MOMENTS WHERE YOU WILL REMEMBER WHERE YOU WERE WHEN IT HAPPENED! Cloudflare is hosting probably the biggest event on showcasing the power of MCPs by having some of the largest companies in the world like Asana, Atlassian, Intercom, Linear, PayPal, Sentry, Square, Stripe and Webflow hold demos of how they are using the power of AI assistants and MCPs in their workflows. Banger! 💣

If money was the problem to enter the #vibecoding tribe, that's also barely an excuse anymore. Lenny Rachitsky just published a monster deal for Lovable, Replit, Vercel v0, Notion , Perplexity and bunch of other tools + access to arguably the best newsletter in the world for just $200 for a full year of pro subscription! I know I am buying that deal any day 🥇💪 https://lnkd.in/esCDJQ-V

Day 5 of the project build in #50in50Challenge. Remember how I got the project done last night at 10:00 pm? Well, I lied. I was up until 3am debugging it. And I'm still not done. Reality kicks in this time. This is going to be the first project in my challenge that I'm not going to be able to deliver by the deadline. 😶‍🌫️ But I also feel pretty okay with it too because I know that this could be potentially the biggest one yet. Stick with me

Day 2. Part 2. But back to square one. 😭 When building a new project, there's so much that you don't know that you don't know. And so, when you see you are hitting the wall on very basic things, don't brute force it. Start over. 💪 That's why I build first and design later - I waste no time or credits, or gain any emotional attachment to the build, it was ugly to begin with. Same idea - only this time, I KNOW EXACTLY HOW TO BUILD IT. And more often than not, by starting over knowing everything you need to, you will make more progress than fixing the Frankenstein you started off with. Mikko Käär each time this happens, I think of you and put a note I owe you a drink. By the time I build 50 apps, I may need to buy you a brewery 🤣🤣🤣

Christmas comes early this year, very early! I launched an app as a part of Lovable Hackathon to create your avatars based on your own image as input + carefully crafted prompt, but could not get it to work. The API just wasn't available. Until today. Can't wait to wake up tomorrow and launch my app v2, only this time with support to the Image Generation API from OpenAI 😍

Working on something new this week compared to my usual small SaaS projects 🔎 My first Lovable prompts were things like: - "Build me a landing page for an app for personal finance!" - "Please fix this, nothing is working!" - "I want this Start button to be bigger!" - "Build me a cool dashboard that looks modern!" Short. Vague. Zero context. It took me a few months to learn how to talk to AI, and now I am happy to say that conversational English is no longer the challenge. But I recently stumbled upon something that looks like wizardry to me a bit - building cognitive protocol designs. So this week I am working on building symbolic systems that let AI reason, self-correct, abstract patterns, and simulate workflows like an IDE. It’s how agents, compilers, and AI tools think under the hood. At least I think that's the case. I may just be stupid as usual! 😃 But it does feel like a ton of fun trying to figure this out.

I created a monster! But instead of a Powerful Godzilla, it's a Frankenstein! https://lnkd.in/eX3cRH9P The good part is - along the way I actually built a few useful tools along the way during the Project #15 creation - a Library of CustomGPTs for Lovable! Here are the links to all 9 of them: Lovable Base Prompt Generator - https://lnkd.in/eFG94xzG Lovable PRD generator - https://lnkd.in/ezDcMNMV Lovable Prompts Wizard - https://lnkd.in/ePbqrsGG Lovable Design Assistant - https://lnkd.in/eHX-U4KW Lovable Idea Validator - https://lnkd.in/e_Ss35Gu Lovable Bug Fixer - https://lnkd.in/edMA7gfz Lovable API Implementation Assistant - https://lnkd.in/eKbhc8Qr Lovable + Stripe Wizard - https://lnkd.in/eU2iuaiq Lovable SEO Assistant - https://lnkd.in/erCAbvir (experimental) ❓ Why this project? Doing a challenge every week from scratch simply doesn’t make sense. So I: 1. Documented my whole setup process 2. Extracted all the docs, chats and any other useful information 3. And converted it into a collection of 9 CustomGPTs for each step of the Lovable builder journey ❓ How does it work? Users needing assistance can: - Brain dump their idea into Base Prompt GPT - Visit Idea Validator and see if what they are thinking makes sense - Take the base prompt into PRD generation GPT and use it to create docs and Design Guidelines - Start their project in Lovable and setup their SSG logic for optimal SEO - From there, if they need any help with APIs or Stripe integration - there’s a GPT for both! - Same goes for general prompting and debugging ❓ Tech Stack I only needed to use ChatGPT this week. KEY TIP: Use JSON to format your GPT instructions for most optimal structure. Include as many relevant docs as needed. ❓ Things I did the first time I’ve built CustomGPTs before but never did I use JSON and math formulas to generate their instructions. ❓ Things I plan to improve I believe that calibration for some of these assistants is definitely going to be needed at a later point I think that the SEO assistant definitely needs to offer more than just the code for SSG ❓ Challenges This week was fun but had a lot more distractions than what I am used to somehow, so I was jumping between things a lot ❓ Final Thoughts Grading it at 6/10, as I think this is a very much needed resource for people getting started, but I also could have built all of these into a Lovable app perhaps so that’s why I am taking 4 points off.

Most times, the best step forward is to actually take a step back. I pretty much did not do anything for my project yesterday. And it helped me immensely. Instead of hitting my head into the wall, I sat down and took some time to think about how can I make this simpler. And that is the biggest lesson that I've learned overall in the last six months of #vibecoding - Simple is always better. I think I finally see the light at the end of the tunnel with this one. It may be late, but I think it's going to be well worth the wait. 😎 💎

No more vibe coding for today! Just good vibes and vball for me until next week. Releasing a video tomorrow as usual though, not on the project but still something very interesting 😁

Day 3. And I will say it again - RESTART YOUR PROJECT EARLY! I made more progress in 30 minutes than I did in the first 2 days of the Week 16 in my #50in50Challenge. It's ugly. It uses mock data and mock flows. But damn it - IT WORKS!!! Can't wait for 5 PM to start connecting the pipes. The stack for this week is: - Lovable for the UI - Supabase for auth, database, edge functions - Apify for data scraping - OpenAI for data processing - Stripe for handling payments - Vercel for deployment There may be some more additions but for now, this is what I am going with. Nothing better than that feeling of seeing everything working exactly the way you want it to! 💎

20k and not even 20 projects in, nice! ☺️

I single shot prompted 11 different apps last night trying to build a simple fitness app. Here's the verdict! 🔎 - DreamFlow - definitely a great start, but super low limit on token numbers. I was able to build the whole app, but I couldn't continue to do anything else. I feel that this is the most promising mobile app builder I used so far.  - Lovable - as usual, this is my absolute favorite because I had the most amount of success using it. I was able to build the app in nine prompts total.  - a0.dev (YC W25) - I tried it for the second time and last time ever. I will say absolute worst experience building a mobile app. - v0 - I love Vercel for all of my project deployments, so I definitely always have to give v0 a chance. And quite honestly, the progress since I last used it has been remarkable. - Bolt(.)new - I think again, yet another overrated app, at least from my point of view. I just felt that I've spent way more time and a lot more prompts and got way less compared to the direct competitors. - Rork- a little bit better than the one above, but still wasted 4/5 free messages for the day on bug fixing. Definitely not an optimal user experience for somebody who is using it for the first time. - Create - unexplainable bugs keep happening to me when I'm using this app, and I think it's again only a matter of prompting style and navigating it that's the big problem for me. - Creatr - not a bad result, as usual. There needs to be some tweaking, I think, design-wise and a little bit on the functionality side. But definitely one of the more promising builders that I think is very underrated. - Databutton - I used it for the first time, and I got to say I'm pretty impressed with the planning feature it has built-in. It has a weird bug though which allowed me to process a prompt without actually adding the URL and the API key that were necessary to integrate the backend. So I felt that that was a bit of an issue. But other than that, pretty cool looking app. - Manus - definitely took by far the most amount of time of all the apps to produce a working prototype, and it was pretty underwhelming compared to the amount of hype people have given to Manus for all these months. I have to say, I expected more. - Trae - I've used other terminal apps before, and I gotta say I really like the experience. I'll probably look to learn more about how to use it properly. A couple more notable mentions: - Von - it crashed on the first prompt - Shellbula - definitely an interesting concept, but they want you to import your own API keys - Softgen - definitely the wrong strategy in my opinion to ask people to pay immediately before they can even test it out. - Bind - this app is still in beta, so it doesn't have all the functionalities and features built out. 16 apps tested in less than 2h. I remember it took me 2 days to do the same exercise a few months ago. I am getting much better at this!

I love Lovable. 🤖 But this is funny and true for some since the 2.0 launch. Meme of the week...or perhaps a true state of affairs? What's been your experience #vibecoding with Lovable these days?

Is IP dead in the age of AI? I really think it is. Elon Musk and Jack Dorsey recently "advocated" that intellectual property actually stifles further growth of AI systems. I personally feel like nobody's protected even the AI tools themselves. Speaking of which - we've just learned that there's a leak of system prompts for the most prolific AI coding tools on the planet. Lovable , Cursor , Devin, Replit , Manus and many more. Trust is the only remaining moat.

OpenAI has entered the chat 😎 Vibe Coding Chat Mode to be precise. They just released 3 things that could change the game: - A terminal code editor that's open source - Two models announced - 4.1. as API and o4 which is excellent at coding - A rumor that they may acquire Windsurf If they do this, they could pretty much overnight dominate the Vibe Coding space. Overall, the amount of innovation makes my head hurt a lot these days and makes me feel like I am so far behind it... Not to mention that OpenAI is also behind, and going backwards it seems, at least according to their model names haha 😃 Even they can't keep up!

Vibe coding becomes just CODING by the end of 2025. Mark my words. We will always need engineers. But probably only the best ones. However, this is the first time in my life that I would advise highschool graduates NOT TO go to college to study engineering. And rather opt into something about product architecture, AI, project management and learning prompting and critical thinking on the side

IT WORKS!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! https://lnkd.in/ebV95D6F