DEV Community

Khue Pham
Khue Pham Subscriber

Posted on • Edited on

A Multi-Part Series: Architecting a Giant - 17 Products Enhanced by AI - Day 1

WLH Challenge: Building with Bolt Submission

This is a submission for the World's Largest Hackathon Writing Challenge: Building with Bolt.

Here you go, my friend. Let me retell you the story of my "AI bake-off," just like we're chilling and chatting over a beer.

Day 1 - First Contact: The Initial AI Test / Picking AI (ChatGPT, Gemini, Grok) for a Deep Dive?

So, here's the deal. My brain's been firing off ideas like a broken firework launcher lately. We're talking everything from Open-source projects, Chrome Extensions, Games, Web3, E-commerce, to Health apps, apps for kids... I've even thought about resurrecting old dead projects, making social media videos, or just straight-up cloning extensions and data. I did a quick count, and we're talking about 17 projects with a clear concept. For most of them, I've just dipped my toes in, you know, a little Proof of Concept (PoC) to check the technical feasibility.

So, What's the Mission?

Since my brain is basically a messy browser with too many tabs open, I asked the AIs to pick a starter for me. The main rule? It had to be something that could make money ASAP. Cold, hard cash. The more the merrier (though let's be real, any amount is a good start). Of course, "generating cash flow" is a vague beast and depends on a million things, but I just threw that general request out there to see how the AIs would handle it.

My AI squad is all top-tier stuff: Gemini 2.5 Pro, Grok 3 Pro, and ChatGPT o3-pro. To analyze the feasibility, I made them all use their strongest "Deep Research" features: For Gemini, it was "Deep research" on the Gemini 2.5 Pro model; for ChatGPT, it was o3-pro (the "best at reasoning" one) combined with its Research feature; and for Grok, it was "Deeper Research" with Grok3. To keep things fair, I fed them the exact same prompts in two different ways: first, a series of smaller prompts adding more details each time, and second, a version where I crammed all the questions into a single, massive prompt.

Round 1: The Face-Off

Kicking things off with Gemini, the Google guy, it shot back with a whole research strategy for me to confirm. Meanwhile, ChatGPT wanted more deets, asking a bunch of follow-up questions. And Grok from xAI? Just jumped straight into the analysis. Ballsy.

I gotta say, I liked ChatGPT's approach. It asked some really legit questions, hitting angles I hadn't even thought of. Honestly, this guy made me create new prompts just to feed the other two :)) By the end, even when I combined everything into one monster prompt, it still managed to find more questions to ask @.@ Anyway, I'll admit, this thorough approach is solid. Score: ChatGPT (C): 1, Gemini (Ge): 0, Grok (Gr): 0.



Thinking GIF

But... and it's a big but... OpenAI's golden boy has this super annoying bug where its 'thinking' process just... hangs. Forever. I'd heard a buddy complain about this before, so I wasn't shocked, but man, is it irritating. I had to F5 the page to see the results, which made me wonder: if I hadn't refreshed, would it still be 'thinking' right now? The UI not updating when the job is done is just awful! But hey, all I care about is the result, so I'll give the other two a little bonus point for not giving me a headache. Score: ChatGPT (C): 1, Gemini (Ge): 0.5, Grok (Gr): 0.5.

While the other two dudes were cruising along, showing their progress as they thought, ChatGPT was noticeably slower. What gives? Out of curiosity, I peeked into its thinking process and saw it was totally different. To analyze a project's real-world potential, ChatGPT spent ages searching and analyzing YouTube videos for a project I'd already shelved, then compared it to other similar projects to scrape for info.

Obviously, this deep-dive method paid off in terms of application analysis—the results were super detailed and spot-on. I guess there's a tradeoff here: more research time for more detailed results. I really want to give ChatGPT points for this, but if it goes down the wrong rabbit hole, all that detail is useless, right? I'll put a pin in that for now and see how other factors play out.

Top comments (0)