2014/11/03 第90期

写给学生的 startup 指南

Paul Graham 的文章。 文章里提到年轻人创业的一个优势是:没钱。没钱之所以成为一个优势是因为,1)软件行业主要成本是人,人的生活成本低,就意味着省钱;2)因为没钱,你思考的东西也是便宜,你做的东西的定价也会很低,这样普罗大众才有可能用 -- 革命性的东西都是比既有的东西便宜10倍100倍的。 Robert Morris and I were 29 and 30 respectively when we started Viaweb, but fortunately we still lived like 23 year olds. We both had roughly zero assets. I would have loved to have a mortgage, since that would have meant I had a house. But in retrospect having nothing turned out to be convenient. I wasn't tied down and I was used to living cheaply.

如何在任何城市营造 startup 的氛围

文中举了洛杉矶、伦敦和斯德哥尔摩 作为样本。 There are four things that need to be in place in order to build a startup scene.
  • The first is a venue that is cheap and central, where meetups can take place.
  • The second is a monthly event where all of the startups gather.
  • The third is an established hashtag everyone in the community can use to share photos and event info.
  • And finally, a coworking space that is open 24/7 so that when an outsider lands in the city, they have a place to go and meet tons of people in the scene.

全球仍有43亿人还没连上互联网

连上互联网的人比还没连上互联网的人少得多了。
In essence, this is a “demand side” problem, i.e. getting the unconnected to start demanding the Internet through building incentives (i.e. the how) and mental models (i.e. the why). Improving Internet access and affordability, on the other hand, are “supply side” initiatives. Both demand and supply sides need to be tackled in tandem so that — in simple Econ 100 terms — the demand and supply curves intersect at a non-trivial price, leading to the generation of a non-trivial economy around connecting the unconnected.

Facebook 如何提高 iOS App 的 Newsfeed 的性能 50%

(原链接可能被墙了,打印成 pdf 放到这里 https://nfil.es/w/OWKCXl/) 这是今年 @Scale 上,Facebook 的这个 talk 的文字版。 Core Data 是瓶颈。 Core Data is at heart an object-relational mapper (ORM). It provides features like full normalization and synchronous consistency across multiple isolated contexts. But since the Facebook app is essentially a cache for data that lives on the server, a completely normalized representation of data wasn't needed. All of those fetch-or-creates while parsing JSON objects were resource-intensive and unnecessary. We developed our own bare-bones form of model objects guided by three principles:
  • Immutability. In this new data layer, models are completely immutable after creation.
  • Denormalized Storage. To serialize these models to disk, we chose to use NSCoding.
  • Asynchronous, Opt-In Consistency. By default, there are no consistency guarantees.