๐ Join our WhatsApp Community
Get the latest AI updates, tips, and insights straight to your inbox:
https://dub.sh/ai-updates-vs
Google dropped Gemini 3 Pro about 3 weeks ago and it topped every benchmark. LM Arena leaderboard? Number one. OpenAI panicked, called a Code Red internally, and rushed out GPT-5.2 just yesterday โ claiming it beats Gemini on multiple benchmarks. 100% on AIME 2025 (first model ever). Better at spreadsheets. Better at presentations. Better at code. 70% on their GDPval benchmark โ “human expert level” they say.
But benchmarks are one thing. Real-world tasks? That’s where it matters.
So I tested both models on 10 actual startup tasks that founders deal with every day. Same prompts. Same criteria. And to avoid any bias (I’ve been using ChatGPT daily for years), I used Grok and Claude as judges to evaluate every single output.
The results surprised me. One model caught a trademark conflict that could’ve cost lakhs in legal fees. One opened emails with “I hope you’re doing well” (instant delete). One literally gave up on a handwriting test. And one built a playable 3D Mario game with enemy AI and win conditions from a single prompt.
Everything is in front of you. No bias. Just results.
0:00 – Intro
1:09 – Test 1: Business Model Canvas
3:36 – Test 2: Pitch Deck
6:25 – Test 3: Revenue Spreadsheet
7:28 – Test 4: Landing Page
8:46 – Test 5: Cold Email
10:13 – Test 6: Content Calendar
11:38 – Test 7: Resume Analysis
12:44 – Test 8: Mario Game
14:32 – Test 9: Handwriting Recognition
15:44 – Test 10: Floor Plan Analysis
17:11 – Final Verdict
——–
To Know More,
Follow Vaibhav Sisinty On โคต๏ธ
Instagram @VaibhavSisinty
https://www.instagram.com/vaibhavsisinty
Twitter @VaibhavSisinty
Facebook @VaibhavSisinty
https://www.facebook.com/vaibhavsisinty/
LinkedIn – Vaibhav Sisinty
https://www.linkedin.com/in/vaibhavsisinty
source
Leave a Reply