Hacker News new | ask | show | jobs
by _mad_eye_ 504 days ago
The tool helps you compare different voices by the topics on the same topic(s) efficiently.

I use it to study how different companies approach problems like hiring, product strategy, org design etc. But as of now I'm not sure what are the user segments I should directly talk to.

Backstory of the idea first (why), followed by how to use it.

Backstory: I am a reader of Lenny's Newsletter, and I was reading the "How {company} builds product" series. But reading alone is not enough, the ideas are easy to read and easy to forget.

Learning these things for me, means comparing how different companies make decisions under the same topic, like hiring, building culture, org structure, etc. I want to compare their answer to the solution, and their reasoning if possible. So I have to deconstruct each article into different single theme chunk, and then sort it by themes (hiring, product strategy,..), then I ask questions about each theme ("what worked for ..., why?") to get useful results ("YC mentality + senior talent"), and then finally, I get to "organize" the information to better build my understanding on the topic.

Real learning comes from 'reading' and 'organizing', the rests (deconstructing, sorting, finding themes) are steps that prepare the information that's helping you to 'organize'. And I think the value of this tool (Reconstruct) is to help you skip the manual preparing part (it takes a shit ton of time and energy).

How to use it: (if it breaks, just hit refresh) 1. Upload some documents of the same topic, ideally newsletter, high quality articles, podcast transcripts etc. Personally I also think this works well with YC startup library too. This only accepts .txt and .md now 2. Click 'deconstruct' to break it down into multiple chunks. It should take less than 10 seconds. 3. Click 'identify themes' to learn which topic that chunk is talking about. It's good when you can find repeated themes (colors) in the articles. This should take around 30 seconds, depending on how many chunks there are. 4. Click 'sort' to view the chunks by their theme 5. Select some cards you are interested in, and click 'Extract by'. Enter what you want to extract from the cards, and you will get the results next to it. 6. To understand the results better, you can go through step 5 multiple times, until you get all the useful information you want. You can then switch mode to connect mode (cmd +2) and connect the cards together to form a learning.

If you think this can help you, also feel free to contact me at ian.xiao@hotmail.com