Ask HN: Idea Sunday

Basically: a reddit/HN clone, but at any given time only one article can be commented on. That article is replaced every day or every hour (whatever interval makes sense) with the highest-upvoted submission that doesn't have comments yet.

The goal would be to encourage deeper discussion of matters, rather than fleeting posts. I have no idea if tree-style comments would be better than just a simple forum thread. A major part of the site would also be scrolling through all of the past articles, and being able to read all of the previous discussions.

http://techcrunch.com/2013/11/04/dropbox-snatches-up-sold-th...

It's an interesting system to build, though, and there is money to be made. The very first version of my startup was something that was pretty much the same system, but we focused on offloading liquidation inventory. Some things that you will likely discover along the path

1. If you use eBay as your dataset, you are pretty much SOL. Their lack of rigorous moderation in terms of their image style guide makes computer vision really, really difficult to carry out. Using SIFT image features from the same exact images, I only got about a ~75% accuracy because eBay users tend to slap all types of random extra things onto their images. I can't imagine you'd have much success trying to use CV for identifying (with sufficient accuracy) consumer electronics in consumer supplied photos. I can't speak for Mechanical Turk... We wound up modifying our tech strategy such that we didn't need high accuracy in identifying other listings.

2. Product versions dramatically alter product value, but are just about impossible to figure out without being explicitly provided that data by the person holding the product.

3. Unless you own your own sales channel, you will be beholden to the marketplaces to offload the inventory in. In my case, we automated the process of selling on both eBay and Amazon. We were reading in data from multiple wholesalers (dropshippers), which had catalogs full of an arbitrary assortment of products... Which means we wound up listing weapons, sexual objects, etc. We got dinged pretty bad by eBay for these things (though, turns out batons sell really well on eBay). Eventually we pretty much got banned from both marketplaces for listing violations + bad fulfillment speed. We couldn't control the shipping speed of our dropshippers (we had no bargaining power), which I attribute to killing that section of the business.

As somebody who actually built the same system that you're describing (sans being an open marketplace, which actually makes it more difficult), I can say that you're in for a rough fight. Pretty epic watching an ebay store run itself, though...

Transforming yourself has never been more accessible than it is today. We have access to so much information and so many resources that there's rarely any valid excuse not to become what you want to be. The problem is that the whole process can be overwhelming, and finding what works for you requires discipline and dedication. Most results don't happen overnight, and the only way to get through months of imperceptible progress is to have a clear plan, track everything, and learn from others.

I've seen all kinds of people in all kinds of context attempting to share some regimen with others. You will find that on Oprah, in books, at the gym, at your doctors, on forums, etc. That's all fine, but why do they still have to write down the name of the product, the brand, the posology, the side-effects, their interactions, everything by hand? Wouldn't it be easier for them to write them down with a tool that understands what the regimen means, and easier for us to add them to our own regimen in a single click?

Once the system understands what I (and others) want to achieve, how I progress and exactly what I do to reach it, only good things can come out of it. It can learn (machine learning, correlation finding), it can recommend tweaks, it can help me acquire products, it can reward me, etc.

How hard is it to set-up a database of all drugs/supplements/vitamins, and let people semantically fill the why, what, when and how?

I don't think there's a lack of niches either:

- Bodybuilding

- Cognitive enhancement (nootropics)

- Weight loss

- Hair loss

- Skin care

- Allergies

- Acne

- Long distance running

- Diets (vegan, paleo)

- Chronic disease

- Life extension

Another useful feature would be suggest me the best food list for the city I should eat.