I'll echo rproffitt in that anything is feasible with the right tools, skillset, money, and time. However, I think the question you really mean to be asking is how feasible this is for you?
I will start off by saying that, as long as you're only thinking of very simplistic complexity, I think the existing LLMs out there are capable of looking at two pieces of code and determining if they do something similar. So the really hard work is done.
Now how good at you at prompting LLMs? What would you even prompt the LLM to look for to consider the matches similar? I asked ChatGPT:
What is pseudocode for an algorithm to rank 1000 articles in terms of similarity to a particular article?
Here’s a pseudocode for an algorithm that ranks 1000 articles in terms of similarity to a particular article. We'll assume the algorithm uses cosine similarity or another similarity measure to compare the articles.
1. Initialize a list of 1000 articles
articles = [article1, article2, ..., article1000]
2. Define the target article
target_article = articleX // The article to compare against
3. Initialize an empty list to store similarity scores
similarity_scores = []
4. For each article in the articles list:
a. Compute the similarity between the target_article and the current article
similarity = compute_similarity(target_article, current_article)
b. Append the similarity score to the similarity_scores list
similarity_scores.append((current_article, similarity))
5. Sort the similarity_scores list in descending order by similarity score
sorted_similarity_scores = sort(similarity_scores, key=lambda x: …