Dec 4, Recap for November | Singularity Gallery

type

status

date

slug

summary

Cursor & WindSurf

Let me start by discussing Cursor & WindSurf in November! While I had heard about Cursor many times before, I didn't start using it heavily until this month. I began by building some demo sites with Python, a programming language I'm familiar with. I was so amazed by its capabilities that I pushed further—using it to build projects with unfamiliar programming languages like NextJS. The experience was incredible.

WindSurf is a newly released AI IDE that shares some similarities with Cursor. Its main distinction is its agent-based function. You simply describe your task in natural language, click a few times, and watch as AI develops your code. The whole process in WindSurf is automatic and seamless.

Initially, I thought WindSurf's agent function would give it an edge over Cursor. However, Cursor's newest version now includes a similar feature. After trying it myself, I was impressed—it not only answers questions but also creates files and checks for bugs during execution. Watching it assist with development sometimes makes me feel like AGI is just around the corner.

With the help of Cursor and WindSurf, I built a demo site called AI Sandbox

The site features several interesting and useful AI tool prototypes. Here's a preview:

During the developing process with AI, I aslo had some insights wanted to share with you:

AI programming is incredibly useful and helpful now. It can help you execute your creative ideas effectively. The knowledge gap between technical people like engineers and non-technical people like product managers is shrinking rapidly.

Does this mean we don't need to learn programming? The answer is definitely no. While working with Cursor, I noticed it made several mistakes, especially as code files grew larger. Since I wasn't familiar with NextJS, I struggled to guide AI in fixing these mistakes. This was frustrating—all I could do was carefully prompt Claude and hope for the best. This experience taught me that while AI is a powerful tool, having a solid programming foundation is crucial. It helps you understand AI's suggestions, spot potential issues, and make informed decisions about the code being generated. The relationship between developers and AI should be collaborative—AI augments our capabilities rather than replacing our expertise.

Learning to code is easier than ever before. When I was a university student, learning to code was challenging. If I couldn't solve a problem, I had to search Google or ask classmates and professors for help. Now, everything has changed. I can ask AI any question and learn directly from it. It's a completely different learning experience.

M4 Max MacBook Pro

In November, I purchased a 16-inch M4 Max MacBook Pro with 128GB of memory. Though expensive, I made this investment specifically to run large LLMs locally using Ollama. The 128GB of memory was essential for this purpose. When I tested qwen2.5-72b-instruct, its performance was incredible—it felt surreal having a GPT-4 caliber LLM running privately on my own laptop.

PTE Examination

I passed the PTE Examination in November with an overall score of 77 out of 90. I'm proud of myself for getting up early to study while continuing to work during my exam preparation. Though it was a challenging period, it ended with success.

See you next month!

<ins/>