Jkroeger123
A few months ago I built a web scraper to analyze reddit posts for side project ideas.
I decided to double down on finding insights from reddit, and build an entire platform for exploring insights from reddit data
I use a combination of LLMs for semantic metadata extraction, topic analysis / clustering, semantic analysis, and custom classifiers to derive insights from.
I hope a few of you may find this useful, would appreciate any feedback
~Justin
I first ran each post title through gpt-3 to classify as either likely or unlikely to be insightful to founders. Then, if it is likely to be insightful I pass the entire post content including up to 100 comments into gpt 4.
This comes down to around 13-14 million tokens used for the 15k posts, which for the latest gpt-4 model is ~$150.
Im sure this could be done cheaper with something like Llama 2 but tbh I am fine with this for now since I was aiming for speed with the initial list to gauge interest :)
Im an indiehacker who loves to solve for niche pain points
I realized I spend a lot of time just scrolling through subreddits looking for user complaints, insights, etc
I decided to have ChatGPT do the heavy lifting for me, and consolidate a more curated list of posts with insights for me to read
Stats:
Posts Processed: ~15k Insights Found: ~1K API Costs: ~$150
I hope a few of you may find this useful, would appreciate any feedback
~Justin