Thực tế cách mình dùng Gemini 3 Pro, Nano Babana Pro và Antigravity

Gemini 3 Pro, Nano Banana Pro & Antigravity: Practical Applications & Workflow

Key Concepts:

• Gemini 3 Pro: Google’s latest AI model, excelling in reasoning and problem-solving, particularly useful for complex tasks like math and data analysis.
• Nano Banana Pro: An AI image generation model capable of creating visuals based on prompts.
• Antigravity: An AI-powered coding assistant designed to aid in software development, debugging, and code refactoring.
• Differential Privacy: A concept in data privacy ensuring individual data remains confidential while allowing for statistical analysis.
• Isomorphic & Neon Style: Specific aesthetic styles used in image generation with Nano Banana Pro.
• Prompt Engineering: The art of crafting effective instructions for AI models to achieve desired results. (Though the speaker emphasizes a conversational approach).

I. Gemini 3 Pro: Reasoning, Problem Solving & Data Analysis

The speaker primarily utilizes Gemini 3 Pro and Nano Banana Pro through the Gemini interface, both on mobile apps and web browsers. A key application highlighted is in tackling complex mathematical problems. Specifically, the speaker used Gemini 3 Pro to explore the Euclidean algorithm for finding the greatest common divisor.

Key Point: Gemini 3 Pro doesn’t just provide answers; it outlines the thought process behind the solution. While acknowledging the model isn’t infallible, the speaker emphasizes the value of understanding the reasoning, enabling better problem-solving skills in the future. As the speaker states, “cái cách mà nó suy nghĩ mình hướng dẫn á nó làm cho mình à đây là cái hướng giải quyết khi mà mình gặp những cái bài dạng như thế này” (the way it thinks and guides me provides a solution approach when I encounter similar problems).

Another example involves verifying the properties of a mathematical function (single, onto, etc.). Gemini 3 Pro provides the question, answer, and counter-arguments if the initial response is incorrect. The speaker stresses that the most valuable aspect of AI tools is their ability to guide rather than simply provide fixed answers. “Theo mình cái điểm đó nó quan trọng hơn rất nhiều so với việc mà công cụ AI nó đưa ra một cái kết quả cố định” (I think that point is much more important than the AI tool giving a fixed result).

Gemini 3 Pro was also used to create a comparative table of data warehouse organization methods for internal training materials. While the initial output included unwanted images that disrupted the layout, iterative prompting (removing images, requesting a revised table) eventually yielded a usable result. The speaker acknowledges the table required manual editing for accuracy but highlights the time saved in initial drafting and idea generation.

II. Nano Banana Pro: Image Generation & Style Transfer

Nano Banana Pro is used for generating images based on text prompts. A specific example involves creating an icon for a Python desktop application. The speaker successfully generated an icon in an "Isomorphic and Neon" style, matching their desired aesthetic. The speaker emphasizes the simplicity of prompting – using natural language as if conversing with a friend, rather than complex structured commands. “mấy câu prom của mình á, mình không có prom cái gì dài và phức tạp hết á. Mình chỉ nói với nó như là mình đang trao đổi với lại một người bạn thôi” (my prompts aren’t long or complicated, I just talk to it like I’m talking to a friend).

The speaker also demonstrates style transfer, using an existing image as a reference for the style while changing the content to be relevant to e-commerce data (orders instead of artwork).

III. Antigravity: AI-Assisted Coding & Development

Antigravity is presented as a powerful coding assistant, particularly for Python development. The speaker notes a significant improvement in Gemini’s coding capabilities with the release of Gemini 3 Pro, bringing it to parity with ChatGPT for coding tasks. “Từ lúc mà Gemini 3 Pro ra mắt á thì mình xài con Gemini nhiều hơn hẳn cho việc code” (Since Gemini 3 Pro launched, I’ve used Gemini much more for coding).

Workflow & Features:

• Code Cleanup: Antigravity was used to identify and remove obsolete code from a project directory.
• Feature Implementation: The speaker tasked Antigravity with adding a loading indicator to a camera panel in their application. The process involved:
  1. Describing the desired functionality.
  2. Antigravity analyzing the code and suggesting implementation points.
  3. Antigravity generating the code.
  4. Manual testing and refinement.
• Reasoning & Action Plan: Antigravity initially provides a detailed "planning" mode with step-by-step instructions, but the speaker found the "fast" mode more efficient after initial setup.
• Prompting Style: Similar to Nano Banana Pro, the speaker uses a conversational prompting style.

The speaker highlights that Antigravity doesn’t replace the developer but accelerates the process. Code generated by AI often requires manual review and refactoring. “Nhưng mà code lúc nào á nó cũng dài thiệt dài hết…kể cả khi mình đã kêu nó chia tách thành những cái file khác nhau rồi nhưng mà không thể nào bằng so với lại việc mà người tự refactor được” (But the code is always really long…even after I’ve asked it to split it into different files, it can’t compare to someone refactoring it themselves).

IV. Information Retrieval & Knowledge Enhancement

The speaker demonstrates using Gemini to explain the concept of "differential privacy." They emphasize the importance of cross-referencing AI-generated information with traditional search engines (Google) to verify accuracy. The increasing prevalence of Google’s AI overview feature is noted as reducing the need for direct Gemini searches for general information.

Gemini was also used to summarize a lengthy slide deck on fairness in AI, allowing the speaker to quickly identify key takeaways and formulate specific questions for further clarification. The speaker leveraged Gemini’s ability to extract information directly from the slide content.

V. Conclusion & Takeaways

The video showcases the practical application of Gemini 3 Pro, Nano Banana Pro, and Antigravity in a real-world workflow. The speaker consistently emphasizes that these tools are assistants, not replacements, for human expertise. The key benefits highlighted are:

• Accelerated Workflow: AI tools save time on repetitive tasks and idea generation.
• Enhanced Reasoning: Gemini 3 Pro’s ability to explain its thought process fosters deeper understanding.
• Simplified Coding: Antigravity streamlines code development and refactoring.
• Improved Information Access: AI tools facilitate quick summaries and explanations of complex concepts.

The speaker advocates for a conversational prompting style and continuous verification of AI-generated outputs. The overall message is that these AI tools, when used effectively, can significantly enhance productivity and learning. “Đây là cái cách mà những cái công cụ AI hiện nay nó có thể giúp cho các bạn được trong cái việc code. Nó không thay thế cho các bạn đâu” (This is how current AI tools can help you with coding. It doesn’t replace you).