Throwback to my first entrepreneurial hustle. Selling these one-wheeler things. I would get the strangest looks from people on the street 😂. We were among the first to distribute these things in Poland. Handing out the flyers was super easy because everyone wanted to know what we were riding on.

How to promote effectively on Reddit without getting banned? New blog: www.maxrohowsky.com/blog/how-to-...

Refining my learnings: If a startup idea has been done exhaustively, it's not an immediate red flag. It's a sign that there's demand. But I personally would prefer a working solution over a shaky MVP. Most people do. So quick and dirty MVPs work better for stuff that's new with few competitors.

DB Chat PH Launch Learnings: - Desktop apps need code signing otherwise users get annoying install warnings - First marketing attempt showed that it's hard to find people who share my excitement for chatting to a DB - The use case of chatting to a database has already been covered exhaustively

Didn't want to overthink the product hunt launch. So I just launched the mvp. DB Chat connects to your database and lets you chat with it. The schema is passed as context to the Ai, so the SQL is spot on. Link: producthunt.com/posts/db-chat

SEO is hard. I spent 30 days learning it from scratch. Here are 12 quick wins to boost your site's SEO today. 👇

The stock market right now

I can chat with any database now! Input: Natural Language Output: Result from connected Database It's hard to explain but this idea 'feels' right.

Domain Rating update: +0.3 after 1 week What I did: 1. Found top blogs in my niche 2. Used free Ahrefs to check their backlinks 3. Posted my site on the 6-7 most relevant Aiming for +0,4 next week.

Let's see how difficult it is to build backlinks. I'm starting at 1.1. Will share update in a few weeks.

It's done! Today I ship. ⛴️ I made a small tool to host JSON & PDF files on the web. It's still super early - but would love some feedback! snip.host

I'm slowly coming back from my social media break. 🥳 The content across social media was unbearable during the elections in Germany, especially on X - had to switch off. Anyway, here's a summary of recent learnings. Shipping continues ⛴️. maxrohowsky.com/blog/solving...

Don't spend all your time on X. There are better platforms. Doubling down. Aiming for 100k.

Great example of why I started sharing stuff online. Post often, show your work, build in public. Do that regularly, and good stuff will happen.

My blog's evolution: 2022 - WordPress 2023 - Headless WordPress + NextJS + GraphQL 2024 - NextJS + MDX Experimenting with my blog taught me more about coding than grad school.

Virality on X relies on early post engagement. To improve the odds, cite your post below related posts: ➡️ Copy the link of your post ➡️ Go to a popular post on a related topic ➡️ Paste the link and explain how it adds value This has helped me go viral a few times.

I created this little helper to debug my CSS a while ago. Now that I use only Tailwind, it has lost its job 💔.

POV: me splitting attention between X and Bluesky

Reddit blocked me many times. But my last 5 posts have all taken off 🚀 Here are 3 thing I changed🧵 (including examples)

Added some moving gradients to my Next.Js blog 🥶

Innovation. There are different interpretations.

One of my awkward habits is stacking tailwind classes vertically.

Making a flythrough was easier than expected. Didn't even need a ComfyUI workflow.

The person who inspired me to start posting has followed me. Awesome full-circle moment!

I recently crossed 1,000 followers on X. It's easy to do the same - here's how 🧵

Decided to optimize the image models for my next SaaS. Retexture functionality will redesign the room but preserve the structure of the room. Managed to do this with a simple controlnet workflow.

You judge yourself by what you feel you can do. Other's judge you based on what you've done. Best way to bridge that gap is to build a portfolio.

Stared playing around with Flux dev in ComfyUI. Images take about 120 seconds to generate on an NVIDIA Tesla GPU with 16GB of VRAM running on Google Cloud. Running the models is easy, but generating custom pipelines tailored to specific use cases in ComfyUI isn't.

Pushed my first test model to @replicate.com from a Google Cloud instance!

Today's key learning: I need to understand the ComfyUI nodes. Randomly mixing together nodes hoping the output looks good won't work.

It works! I present: Stable Diffusion pics of a surfer being computed within seconds on a Tesla Nvidia T4 in a datacenter somewhere in Wyoming 😂 The 600 $ free-credits on GCP are more than enough to tweak the image models I need for my next SaaS. Best thing is I can even run this from my mobile.

Running ComfyUi on my machine is terribly slow. 🥲 Now switching to a Nvidia Tesla T4 on GCP. Quite close to getting it to work.

My weekend plans: 1. Setup fast comfyui environment on GCP 2. Create a basic workflow (e.g., upscaler) and push it to replicate 3. Run the workflow from a Next JS frontend via the API My head already hurts from all the bugs I'll need to fix.

Me: Validate before you build Also me: Stuck 10ft deep in a broken ComfyUI model

Late night session setting up ComfyUI

I used to live by the Olympic Stadium in Berlin. It has a capacity of about 75,000 people. The idea that a random post I wrote at the gym can reach 5 million impressions — equivalent to filling 66 Olympic stadiums — blows my mind.

To whoever is still up late coding: www.youtube.com/watch?v=n--S...

I'm from Germany. Everyone talks about how our economy is failing. But why do people say that? Well, let me explain: 🧵

POV: In the kitchen, cooking.

Since yesterday, ~20 people have tried my indie hacker quiz. Currently the best score is 6/12. Definitely beatable 😅 Feel free to try your luck: indie-quiz.maxrohowsky.com P.S. Do you have a viral post you want me to include? Link it in the comments below!

When things don't go as planned, I remind myself: To hit a home run, you either need to get better or swing more often. Here's to many more swings, a lot of improvement, and accepting that you can't choose the pitcher.

Comparison of my old and new blog on maxrohowsky.com Old: Next JS, Modular CSS, Headless WordPress New: Next JS, Tailwind, ShadCN, MDX, Supabase I think my design sense has (at least slightly) improved over the years.

On GitHub, I have a list of past projects. I keep it there for whenever something does do well. For those moments, I have something to point to where I can say: "Sure, that one did well, but look at all the ones that didn't." 😅

POV: In the kitchen, cooking.

My top 10 favorite initial commit messages for new projects. Any other good ones? 😅

The 'Starter Pack' hype is dead, but it left one big problem. Many users now follow hundreds of accounts they don't care about. That hurts engagement, as users engage less with stuff they're not interested in, which makes the platform stale. 🧵[1/3]

Yesterday was a crazy day: - 100 PH upvotes on my Jan 1st launch - 1.7 mil. impressions on an X post - 30 upvotes on Reddit, first time without a block Reaching millions of people by just pressing 'post' is humbling.

This is exactly why I love being an engineer. There's something rewarding about going a from a rough sketch to published app in just one weekend. No other discipline compares.

Called it! Now it's at 500k. 🚀