Baiting Tech Twitter (AI Sells)

*note: i have adhd and i’m not a good at keeping my thoughts in order/flow.

This social experiment led me to understand that Humans are flawed and they are very easy to fool even if they are experienced beings, people who are funded by one of the biggest venture capitalists in the world.

Don’t ever feel insecure if you’re not getting funded maybe it’s about your soft skills and mindset more than just your technical skills.

Just because you’re not being validated by one single successful person doesn’t mean you’re not valuable. You’re supposed to be valuable infront of masses because they are your target audience not a single person.

Idolizing and seeking attention of one specific person is the worst thing you can do to be successful as per the societal norms. At the end of the day you’re meant to compete with your previous generation and not with your peers who had a totally different circumstances.

Well enough yapping about the society, let’s just dive more into this thing I pulled off.

Raghav Sharma did a tweet mentioning that he joined openai and he has chatgpt in localhost.

And well he got reach on that tweet and well I can get that reach too. Well i would like to call this systematic tweeting as bandwagon reach on twitter (i just made that word up, because whenever there’s a specific theme trending you can just join the bandwagon and get extra reach)

Then I was like let’s do something similar maybe with Claude, I manually copied the frontend of claude and then ran it on localhost,

Some days back when i was running Gumroad’s open source code locally and it was not using localhost instead it was mapped to gumroad.dev so hmmm my goofy ahh brain thought I can do something with this mechanism.

On a random morning I was like why not build a reverse proxy backend which maps a website to localhost, Then i went on to build a reverse proxy golang backend and Next.js Frontend to bait people.

Before we start this i would like to give special thanks to Raghav Sharma, Ayush Singh(shydev) and Salfiya for helping me with reach and emotional support.

At first people believed that this is so real then someone said it’s downloaded html or i’m using wget or using /etc/hosts to map all the websites to localhost.

I had to delete the tweet after 1.3M impressions and 1.8k followers because it got reported.

Example of one of the videos used in the tweet (i made a thread of 20 videos in that tweet tagging those websites).

*ps: i use screen studio to record videos.

Why did this go viral?

Whenever we write a tweet we are supposed to consider that this tweet is going to be seen by a lot of people and not everyone has the same knowledge base inside their smoll brain or attention span to understand the tweet.

After considering all that info we are supposed to simplify it so much that everyone can comprehend it.

But again complex stuff gets attention if your target audience is geeks or wanna be geeks.

Reasons for Virality:

• Simplicity of the tweet
• Media Inclusion(crazy boost when you use image + video)
• First 1 Hour(5 Reposts from big accounts and 30 likes are enough to get the ball rolling if your tweet is goated)
• Lovable, V0 & Bolt are not enough for them. (no one will ever trust a cursor alternative as it’s too good already)
• People need this product which can replicate beautiful components (they hate the tag of vibe coded shadcn ui).
• It was just very unique + very unrealistic easy to trigger quotes & replies.
• Tagging accounts help the tweet reach to the followers of the tagged account (personal experience)
• Big accounts liking your tweet gets the tweet recommended to their followers.

Personally, I think there’s a huge space in generating animated components with perfect color theory and ux laws using gsap and framer motion.

The Public Composition

Public composition is almost similar everywhere and if you can milk them then you can easily make it.

ps: Haters also include people who maybe need a piece of your virality.

People move on really fast but it takes time, fabricating and changing your public persona isn’t that hard even if you’re the worst person on this planet.

Who Noticed

• CTOs dm’d me to help them to improve their product.
• People offered me money to sell access to the tool.
• Two Founders tried to hire me.
• Designers & Web Agencies.

Considering I want to stay independent, I had to ditch all the offers.

I should mention that there’s still room to serve customers with AI wrappers. AI products sell well, and despite high churn rates, you’re veryyy likely to generate at least thousand dollars in revenue.

ALSO WHY TF IS THERE NO GOOD CUSTOMER SERVICE BY AI WRAPPERS EXCEPT CURSOR?

Technical Implementation

API Integration

The frontend makes HTTP requests to http://localhost:8080/api/create-proxy?url={target} which returns JSON containing the assigned port and localhost URL.

The Go server handles these requests by spawning new proxy instances and returning the connection details for immediate use.

Golang Backend

The backend uses httputil.NewSingleHostReverseProxy(targetURL) to create dedicated reverse proxies for each target site. Each proxy runs in its own goroutine on auto-allocated ports starting from 3001. It selectively rewrites URLs based on Content-Type headers - only HTML content (text/html) undergoes URL transformation while other content types (CSS, JS, images, JSON) pass through unchanged.

The proxy performs bidirectional header manipulation by setting r.Host = targetURL.Host for outbound requests and rewriting Location headers in responses for proper redirect handling. All original headers are preserved by default, including Content-Type, Cache-Control, and Set-Cookie headers. The response modification pipeline reads HTML content, transforms external links to localhost addresses, and updates Content-Length accordingly.

CORS headers are automatically enabled to allow frontend communication.

Data Flow

User submits URL → API call creates proxy → Go server spawns reverse proxy on available port → Returns localhost URL → User accesses proxied content through localhost while proxy fetches from original site behind the scenes.

The architecture maintains thread-safety through mutex locks and provides real-time proxy creation with instant feedback.

ps: this won’t bypass the auth because the origin is not localhost.

I’ll open source the code if you guys are interested, but for that help this tweet reached 200 likes first.

The End Results

Got shadowbanned after a report after getting 1.8k Followers & 1.3M impressions.

Well I got to know that you can just deactivate your account for 72 hours and then reactivate it and you’re good to go but most likely you’ll have to delete your tweet

• request for a review you’ll most likely get a notification for that.

Feel free to check my other Viral Tweets too.

Also you can always reach out to me on X or at meow@nermalcat69.dev.

ps: if i made any mistake you can @ArjunShips me on X.

ps: images with drawings are from pinterest not sure who’s the artist but you can always reverse image search using google image search, yandex image or even tineye (tineye is mid tho).

Also just a random tip whenever you’re trying to find info don’t forget to check exif data but nowadays a lot of platforms just delete that data.