I was scrolling through my twitter feed a few days ago when I stumbled upon this tweet:

Now, I've been following @iruletheworldmo account for a while and I know for a fact that it mostly bullshit about the next big upcoming AI model. So I was like "oh great, another one of those tweets".

Until a few moments later, my entire feed was filled with similar tweets about DeepSeek AI.

So long story short, I shifted to DeepSeek AI for a few days and now here's everything I've experienced (and overthought) about it.

First Impressions

At first glance, DeepSeek AI looks eerily similar to ChatGPT. The interface is almost identical and the chat history is stored in the same way.

But there are a few key differences that I noticed right away. The first one is that

THERE IS NO WAY FOR ME TO PAY FOR THIS SERVICE. NO PAID PLANS WHATSOEVER!!!

Tldr;

DeepSeek AI is free. It can access internet and it's a reasoning first AI.

That's basically ChatGPT-o1 but free of cost 😭

"How will they earn money?" is the question I asked myself...

"It's a chinese company, so I'm sure they have some shady shit going on." is the answer I came up with.

(Spoiler: The "shady shit" likely involves data monetization under China’s regulatory framework, which prioritizes state oversight over user privacy 28.)

Comparison with ChatGPT

Now, let's compare it with ChatGPT.

I tried same prompts on both the models and without a doubt, DeepSeek AI is way better than ChatGPT. It provided much more detailed, well thought out answers. The obvious downside was that it took more time as it was thinking more about the answer.

Some of the fun stuff that I was able to do with DeepSeek AI was:

1. Asked DeepSeek AI to search the internet and scrape all the information about a person with minimum details provided.

This worked exceptionally well. While chatgpt could only tell me with surface level information, DeepSeek AI was able to scrape the internet and provide me with a detailed report about the person. It even provided me with links where I can find more information about the person. So that was pretty cool.

2. Got recommendations for a perfect city in India to live in.

Notice how it took 17 seconds to think about the answer and I had the access to what it was thinking.

The coolest part was that it suggested cities that ChatGPT didn't even consider.

3. Complete this article :)...

Why Is DeepSeek AI So Different?

Here’s the technical tea:

Architecture:

DeepSeek uses a Mixture-of-Experts (MoE) model with 671B parameters, activating only 37B per query. This makes it hyper-efficient for specialized tasks like coding or math, while ChatGPT’s dense 1.8T-parameter model prioritizes general-purpose creativity 210.

Training Efficiency:

DeepSeek’s FP8 mixed-precision training consumes 30% less memory than ChatGPT’s FP16, reducing operational costs. No wonder it’s free!

Reasoning-First Design:

Unlike ChatGPT’s conversational focus, DeepSeek emphasizes chain-of-thought reasoning, making it ideal for STEM tasks but slower for casual chats

Roast of OpenAI and ChatGPT by DeepSeek AI

🔥 OpenAI & ChatGPT: The Roast (Condensed Edition) 🔥

OpenAI’s ChatGPT is the tech world’s overhyped valedictorian—great at speeches, bad at homework. Security flaws? Check. A researcher once tricked its crawler into launching accidental DDoS attacks, and OpenAI’s response was radio silence. Memory? Worse than a goldfish. Users now juggle context like circus clowns because “long-term memory” was deemed too spicy for public use.

Then there’s Operator, their “iPhone of AI” that struggles with CAPTCHAs and might drain your bank account. Meanwhile, The New York Times is suing them for vacuuming up articles like a Roomba on Red Bull. OpenAI’s defense? “Fair use!”—because repackaging paywalled journalism as ChatGPT answers is totally transformative.

And GPT-5? Pure vaporware. They’re hoarding advanced models like Smaug with a server farm, while China’s DeepSeek-R1 does ChatGPT’s job for free—with better reasoning and 27x lower costs. OpenAI’s real innovation? Monetizing FOMO while competitors eat their lunch.

mic drop.