- OpenAILearning
- Posts
- Gemini 2.5 is Insane
Gemini 2.5 is Insane
5 Unexpected AI Side Hustles You Can Start Today

Gemini 2.5 Pro is a game-changing AI model from Google, primarily known for its unprecedented 1 million token context window. This massive memory allows it to process enormous amounts of information like eight novels or 50,000 lines of code in one go.
Gemini 2.5 Pro Beats ChatGPT
5 Unexpected AI Side Hustles You Can Start Today
Gemini 2.5 Pro can handle entire lengthy documents, codebases, or even a feature-length film without losing context.

Gemini understands and works with text, images, audio, and video seamlessly. For example, it can watch a tutorial video and answer questions or analyze an image and generate related text.
It thinks through problems step by step internally before responding, leading to more accurate and coherent answers, especially for complex tasks.
Gemini can not only generate code but also run it to check its work, significantly improving accuracy for math and data problems. It's also tuned for complex coding prompts and can even generate interactive simulations from simple requests.
It demonstrates state-of-the-art performance across various benchmarks and supports multiple languages.
This technology has the potential to revolutionize fields like law, medicine, and software engineering by efficiently handling vast amounts of data. It also paves the way for more natural and continuous AI interactions and more capable AI assistants.
Google sees Gemini as a foundation for AI agents that can observe, plan, and act, moving towards more proactive and autonomous AI.
While powerful, this model is computationally intensive. Overall, Gemini 2.5 Pro represents a significant leap in AI capabilities with its massive context window and multimodal understanding.
Real Things People Built This Week with Gemini 2.5
A 3D car simulator โ built using nothing but Three.js and one prompt.
Turn images into 3D-printable objects
Itโs not just code โ Gemini Pro 2.5 understands visuals too. See it in action.Smashes performance benchmarks
๐ง Reasoning: 84.6% on GPQA Diamond
๐ป Coding: 92.0% on AIME 2024
๐งฌ General Knowledge: 18.8% on Humanityโs Last Exam
Outperforming GPT-4 and Claude 3.5 Sonnet in multiple categories.
Generated better results than GPT-4 โ when tested side-by-side using the same prompt.
Built a Smart TV UI with one prompt - No UI design background. Just vibes and Gemini 2.5.
Outwriting human writers โ some say it's crafting better blog posts, email copy, and story hooks than most pros.
Built a Minecraft clone from scratch โ without writing a line of code manually.
Created an animated 3D scene โ again, using Three.js and pure prompt magic.
Rubikโs cube solver & generator โ ready for 3D and interactive use.
Snake game revival โ written in modern JavaScript, playable online.
Freeform drawing with SwiftUI โ like a full creative canvas in one command.
Simulated a Galton board โ a physics experiment gone digital.
Built โvibe codingโ games โ ambient, stylish, and 100% generated with AI.
A working flight simulator โ controls, physics, scenery and all.
A smart timestamp generator โ perfect for developers and video creators.
๐ง What Gemini 2.5 Pro Really Unlocks
You can now build:
โ Full SaaS apps in minutes
โ Fully functional games with simple prompts
โ Beautiful UIs โ calculators, tools, dashboards
โ AI agents that automate your entire workflow
โ Email systems that actually work by themselves
It excels at:
๐ Creating visually compelling web apps
๐ Building advanced agentic programming systems
๐ Code transformation, optimization, and editing
๐ In Case You're Wondering...
Gemini 2.5 Pro topped the latest LiveBench tests with an 82.35 global average, dominating in both reasoning and coding.
Gemini 2.5 Pro is such a strong coding model. I was able to use it to convert an image into a 3d-printable object, and bring it to live. More in ๐งต๐
โ Fei Xia (@xf1280)
5:34 PM โข Mar 25, 2025
PRESENTED BY SUPERHUMAN AI
Find out why 1M+ professionals read Superhuman AI daily.
In 2 years you will be working for AI
Or an AI will be working for you
Here's how you can future-proof yourself:
Join the Superhuman AI newsletter โ read by 1M+ people at top companies
Master AI tools, tutorials, and news in just 3 minutes a day
Become 10X more productive using AI
Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.
5 Unexpected AI Side Hustles You Can Start Today
The EASIEST AI Side Hustles Nobody Talks About ($700 Per Week)", here is a mind map outlining the AI side hustles discussed:
#1. AI Voice Cloning
How it works: Record at least 30 minutes of your voice and upload it to a platform. Earn when people use your voice for AI-generated voiceovers.
Platform: 11 labs.
Earning Potential: $2 to $200 per week per voice, with higher quality voices earning more. One Reddit user earns $500 to $700 per week.
Ideal for: Individuals with a good voice or voice acting experience, but not necessarily required.
Effort: Relatively easy, requires initial recording time.