We are YC rejects (I made it to the interview once!) but have benefited greatly from everything YC has put out to help founders like us. Since YCombinator has shared so much helpful content over the years in all sorts of areas, we decided to put some order around it for everyone’s benefit. The process looked like:
- We collected everything YC has put out scattered across different channels (and developed a system to keep it updated)
- Categorized it using a hybrid NLP-human approach to uncover key macro and micro categories along with identifying the people behind the advice. The goal was to 1. Aid exploration 2. Be able to quickly find the most relevant advice in any given situation.
- Our company (Polymer Search) specializes in converting a spreadsheet (or any structured data set) into a search and insights engine automatically. That was a huge part of why we decided to tackle this. We used our own beta stack to create YCAdvice.com. The spreadsheet is linked from the site in case anyone wants access to the underlying data (free to use).
Would love any feedback and happy to answer any questions.
You should post about your underlying technology, when you're ready. It sounds like something HN would find interesting. Feel free to email us at hn@ycombinator.com and we might be able to help with how you present your work to the community. This offer goes for anybody—just realize that sometimes there are long delays before we can get back to you.
Are you saying that all the YCAdvice data is sitting in a spreadsheet? Other than the effort to categorize all the content, aren't there many tools that let you use a spreadsheet as a database? What am I missing here?
You're right, sorry if my other comment didn't make sense. I was talking about both the data itself and also the YCAdvice discovery and search interaction layer on top of that data.
We did first construct the spreadsheet (where lot of categorization was automatic) but then the real value IMO is being able to understand and make use of that multi-dimensional data really easily. The goal being to explore efficiently and develop rapid intuition for anything. The actual spreadsheet is here so you can see the difference: https://docs.google.com/spreadsheets/d/1xTMF_t_EDG34IjnXo-ho...
We are improving stuff so that Polymer's stack can auto-convert spreadsheets with numerical data and do great visualization etc so it can improve human intelligence beyond just categorical data which is the case here.
Let me know if I can clarify anything. Thank you for checking it out.
So the visual/interaction layer is generated somehow based on the underlying spreadsheet? How are those interface decisions made and how does the UX/UI change from one spreadsheet to another?
Hmmm nice questions. So when I was at Google before, I saw a lot of different kinds of datasets. But I also saw commonalities in terms of not just data types, but also data abstractions or the kind of stuff people like to do with that data. The goal (a difficult one for sure) is that can we automate a lot of that for any dataset so it's still unique and powerful enough for that particular dataset but at the same time can be done by anyone without writing code or ton of manual customization ?
In summary, there is a fairly complex process towards making those decisions. Feel free to pm me at ash [at] polymersearch.com if you want to dig deeper.
Great tool to generate slick wikis quickly. So, if I connect a sheet to polymer then what's the output? the whole website is generated by Polymer which is then connected to a custom domain? A video demo of your tool/dashboard would clear some doubts.
Also, in the no-code tools area, people use Airtable with integromat/zapier, Webflow to make websites. So, in case you do generate the website, then I think it would be make sense to market/target this as a no-code tool too, it would gather interest from the no-code community quickly .
I didn't know YC had so many resources and topics in general. This looks like a pretty well laid out interface for that. If you don't mind, what did the categorization process look like?
Yeah there is a lot of breadth in what YC puts out as we found out ourselves. Categorization for specific domains is hard since we don't have labelled data. It's a mix of unsupervised token extraction, clustering to detect significant concepts along with a human layer to tune false positives and also curate few topics manually since the process will never get everything right on its own.
We are in the process if scaling it further so we can help create many more resources for users and their own data.
This is dope! Slick design and tool! How many dimensions of data does it support? I visit the website with nothing particular in my mind and think it would be cool to have a sort by popularity type of thing to help me sort things through. This way I know which resource is most valuable to people and most liken by people and encourage me to check it out.
There are a lot of talks on product and fit in general by YC and they are very informative. What was surprising to me was how many talks/resources they had around this and that can lost. I hope you can find some more around that topic.
Thanks. Currently, we have been focusing on all the official resources from YC and not so much on Hacker News threads. But if there's demand for better discovery for Hacker news itself, it'd be very interesting to look into.
We are YC rejects (I made it to the interview once!) but have benefited greatly from everything YC has put out to help founders like us. Since YCombinator has shared so much helpful content over the years in all sorts of areas, we decided to put some order around it for everyone’s benefit. The process looked like:
- We collected everything YC has put out scattered across different channels (and developed a system to keep it updated)
- Categorized it using a hybrid NLP-human approach to uncover key macro and micro categories along with identifying the people behind the advice. The goal was to 1. Aid exploration 2. Be able to quickly find the most relevant advice in any given situation.
- Our company (Polymer Search) specializes in converting a spreadsheet (or any structured data set) into a search and insights engine automatically. That was a huge part of why we decided to tackle this. We used our own beta stack to create YCAdvice.com. The spreadsheet is linked from the site in case anyone wants access to the underlying data (free to use).
Would love any feedback and happy to answer any questions.