Using cursor to develop production worthy codes in company is fantastic. Converting user prompt to ready to ship code is so easy and I just didn't code VS code for a long time. Ready to switch to new coding IDE. For example, implementing a trie tree structure take time and find production ready code from Google or Bing it's not easy. But cursor actually produces a clean version of trie tree implementation. With only a few naming changing, you got your production worthy code.

Another more complicated task I tested on Cursor Code Editor is to write a python code to implement the Transformer function as in paper "Attention is all you need". The code I am writing include function name "def transformer(x):" and some of the beginning lines like "q=torch.linear(". And with the suggested code, I wrap up the transformer function in less than one minutes. So it can really help me increase the productivity of writing code. The only drawback is when I want to modify the code and use backspace to delete some lines, it seems like Cursor doesn't know what to do and give no hints for a while. Unless I start writing something else. And the newly suggested code will usually be quite different from the previously suggested ones. ``` ## Machine Learning ### Input import torch def transformer(x): ndim = 16 q=torch.linear(ndim, ndim)(x) k=torch.linear(ndim, ndim)(x) v=torch.linear(ndim, ndim)(x) attn = torch.matmul(q, k.transpose(-2, -1)) attn = attn / math.sqrt(ndim) attn = torch.softmax(attn, dim=-1) ### Cursor Suggested Code: out = torch.matmul(attn, v) out = torch.linear(ndim, ndim)(out) return out ```

I just downloaded Cursor on Mac and tried some coding task. The first task is to write a fabonacci function in python. And once I typed the name "fabonacci(n)" and end the function name with the colon ":", there is a code hint tab poping-up suggesting the remaining code. The suggested codes are correct and are very succinct. Very Helpful AI coding tools. ``` def fabonacci(n): if n == 0: return 0 elif n == 1: return 1 else: return fabonacci(n-1) + fabonacci(n-2) ```

Overall, Gemini did a good job though not optimal in the pytorch model coding test. I asked the same question on ChatGPT vs Gemini vs Claude. Among all these three platforms, Gemini‘s python code is correct but less concise. It also uses some unusual functions of pytorch package, such as """ context.permutate(0, 2, 1, 3).view(...) """. This is just a weighted average tensor operation, so this implementation is very uncommon and not necessary. And ChatGPT and Claude use the correct function for this operation, which is """ torch.einsum().reshape() """. Even though it's not a big deal, but generating common python code in Gemini is not optimal.

Gemini's results of data analysis and the feature of drawing plot really surprise me. It gives exactly the bar chart comparing the market capitalization of several companies. And I even double checked the July 2024 financial data, which is correct. It also gives me the results of the python code which draws the chart so that I can save it and modify the chart for later use. Really amazing!!!

I asked Gemini to give an introduction to Maxwell equations and explain each terms in the equation. Overall it gives reasonable answers to explain Maxwell equations are a set of four equations. But It will be nice if the equation itself is provided before explaining the term. So Maybe it lacks of generation of the latex code of maxwell equation?