DEV Community: scorder

I made an LLM visibility tool that doesn't suck

scorder — Wed, 18 Feb 2026 22:53:43 +0000

LLM Eagle is an AI visibility optimization tool for SaaS, built for developers and indie hackers.

Why This Problem

I took on this project not just because:

The GEO (Generative Engine Optimization) industry is booming
VCs are funding similar startups
The idea is validated
Product-market fit is proven

But it's also because I am genuinely excited for this under-explored opportunity.

The Tech Stack

I used:

React with Remix/React Router V7
TypeScript
Tailwind with DaisyUI (I discovered DaisyUI and fell in love)
PostgreSQL with Neon

From Simple to Unavoidably Complete

It started with a simple plan: a rudimentary, minimally viable program that provided a basic GEO report once you entered your domain.
Nothing too shiny, just something that works.

However, as is often the case with MVPs, I kept building.

I fell into the "feature creep" trap - continuously iterating without feedback, stuck in an endless cycle of building and repeating.
Multiple buronouts, sleepless nights, headaches, mindlessly coding and adding features.

This time however, I didn't regret the cycle as much as I usually would. Because that extra work resulted in a more complete product that is actually useful. Since I knew the idea was already validated, I was comfortable putting more effort into the MVP before hitting the market.

The Product

So what features did I actually build?
You can add multiple websites to optimize them for GEO.
There are 4 main sections:

Overview - Visibility and sentiment of your website
Prompts - What prompts are being asked by your target audience
Content - Recommended article titles to maximize reach (you can generate articles as well)
Competitors - Top performers in your industry and their strength

I am iterating on this structure as I get feedback.

The Marketing

I’m still trying to figure out what works. I am currently finding my niche and dialing in my outreach methods by experimenting across X and LinkedIn.

If you are checking the product out, do comment here or message me at scorder96@gmail.com with feedback. I'd really appreciate it!

Product Link: llmeagle.com

Thank you for reading.

I made a Blackjack game with ChatGPT

scorder — Sun, 08 Jun 2025 19:16:02 +0000

My friend recently worked on an HTML Blackjack game. The Javascipt code for the game was 335 lines long and full of bugs. Fast-forward a few weeks, I really wanted to play Blackjack online but couldn't find a website with a good UI, ads not bombarding my screen and no casinos with hot girls trying to take my money.

So I decided to do it myself.

I gave Chatgpt a very simple prompt to begin with. I wanted to see what it comes up with without me having any specific prefereneces.

Prompt -
Make a blackjack game using react, typescript and tailwind.

I set up a new project in VSCode and pasted the code in it.
I now had a very simple, playable Blackjack game.

First thing I wanted to change was the display of your cards. It does the job, but actual cards would look much better.

Prompt -
Instead of numbers and icons, make a card component and use it.

This will do the job.
Finally, I asked chatGPT to add animations with motion.

At this point, I had a game that worked perfectly. I can play it just fine, and it simply works while looking nice and minimal. But there was something missing. It just didn't have that juice.

I knew I had to make some manual changes.

First thing I did was center the game vertically.
Added a navbar and moved the heading "Blackjack" away from the game.
Added some icons to the buttons to make things clearer.
Made it so startGame() was being called when opening the site or refreshing.
Added a How to Play section that was written by ChatGPT.
Changed the title, description, and logo.
Added confetti when you win.

Now it looks like this -

Conclusion

As you can see, I added quite a few things to the game manually. They are not too complex, if I asked chatGPT, or better yet an AI based IDE to do it I'm pretty sure it'd do them without much problems. But here's the thing -
I'm a developer, not a manager. I don't want to explain to an AI exactly what I want done, provide minute details, repeat the phrase "YOU ARE A SENIOR DEVELOPER" etc. I'd rather save the time and headache and do it myself. It'd take 5 minutes and I'd actually enjoy building.

However, I was just too tired to implement the Blackjack logic myself, and that's where AI really shines. The mundane, boring stuff.

A hot take,
AI can help a lot to build much bigger projects a lot faster. But it cannot yet be a replacement. A lot of people advertise building a complete app with AI and getting rich, but that dream is yet to be realized. If you want software that people will love using, you will need to step in at some point. Think of AI art vs human art. AI art isn't that moving. Similarly, a visitor to a website will notice the difference.

Finally,
There are a lot more improvements that can be done to the website. Custom backgrounds, total card value, betting with fake money and more. But I've left those to you guys. Find this project on GitHub and send some pull requests. Let's make this the best Blackjack game to exist!

https://github.com/scorder96/blackjack

I built a scheduling platform for Bluesky

scorder — Fri, 28 Feb 2025 23:15:31 +0000

With Bluesky opening up to the public in February 2024, and emerging as a stronger alternative to Twitter, a gold rush soon began to develop third-party tooling around the paltform.

Scrappy software and AI-generated slop soon began flooding this market, along with bigger social media analytics and planning softwares adding Bluesky to their list of supported platforms.

The Why

Going through some of these tools, I deduced that a clean, fully functional, built for Bluesky platform would have potential to take up a good chunk of market share.

Even if it didn't, my time spent built building the software would at least help me learn a few more things about SaaS development.

The What

I sat and wondered what I could build. What did Bluesky users really need? Or at least that is what I wish I had done. Instead, first thing I did was open VSCode and started coding away.

I had an idea in the back of my mind though, analytics. That was what I was focusing on. But as I built feature upon feature, a consistency tracker here, a growth graph there. I realised a sickening truth. 😰

Nobody is gonna pay for pointless analytics.

You don't really need fancy graphs and calendars to show you what's doing well. The notifications bar, likes and replies inside Bluesky are enough to give you an idea. In simpler words, I wasn't solving a real problem.

That's when I realised I needed something more. And the only thing I could build that would solve an actual problem was post scheduling.

The How

I knew I needed a severless function running day and night uploading the posts at their scheduled times. I had been avoiding Cloudfare Workers since months but it was finally time.
I explained what I wanted to ChatGPT in ExpressJS terms, and it spit out Workers code. Turns out it isn't really that difficult. In fact it's amazing.
I set up a cron job running every minute that checks if there's any posts that needs to be scheduled at that minute, and uploads the post using the Bluesky API.

And just like that, I had a scheduler up and running. 👌

But that wasn't enough. What app nowadays is complete without some AI feature? My tool can now write engaging posts for you. But wait, simple scheduling isn't enough. I needed smart scheduling. Abracadabra 🪄. My tool now tells you what the best times to post are.

The Feature Trap

This needs some fixing.. oh wait this can be better.. hmm, maybe I can add this.. and on, and on, and on.

I have no idea how many times I have iterated over the code, adding features and improvements instead of getting my website out there.
It's embarassing - since I have been through all the advice on the internet about resisting the temptation to include additional, often unnecessary bells and whistles to your product.
I believe the feature trap is a result of the fear of not putting out a good enough product. "Sure, I can share it on Reddit now but if I just add this feature it'll look much better" and the cycle continues. 🔁

What I learned

Define the problem and the solution, exactly what you're going to build, what features you're going to implement, what tech stack you're going to use, and how you're going to market the product, in a single document.
This would be not only be something your future self would thank you for, but would lead to a lot less confusion, a clearer roadmap, and faster shipping times. 🚀

If you are wondering what the tool is it's called Bluestride. Check it out here.

I hope this article and my experience helped you build your understanding on the journey of SaaS development.