WeChat 'lite' app


WeChat official accounts function is out for only 3 years, the kinda influence it had over marketing industry is incredible. This time, WeChat is back with something seemingly even more ambitious. This time, rumour has it, WeChat wants to replace all other applications. They call it 'WeChat small apps', or 'lite' apps if you will, because you can only use the apps through WeChat and the apps are gone after you use it. Zhang Xiaolong made a keynote speech at the end of 2016 and talked all about it. Here is a collection of what he said and my thoughts on it.

WeChat 'lite' apps:

do not require installation
are within your hands' reach
enable you to use it and close it
do not need to uninstall

One analogy to summarise what WeChat 'lite' app is: the next generation websites on your phone. Xiaolong's speech was sincere and he showed us the engineering side of him. It is not hard to feel his excitement when he talked about the time (20 years ago) he read a book written by Bill Gates. He admires Bill Gates' foresight of how the world could be connected by the Internet and with his engineer's humility he visioned something similar for WeChat 'lite' app. To be honest, I am not entirely sold on the idea, but I have to admit it poses a big question mark for me, that I respect.

Last time I went to Shenzhen Airport, there is something interesting to many signboards. 80% of them have QR code, the QR code that links you to a particular WeChat official account (a lite content publisher). In the PC era, every ad comes with the website. I am very happy nowadays WeChat QR code replaced websites.
One thing my western audience might not understand is that consumers in China behave in a very different way. QR code is an excellent example. In China, you can buy your breakfast just by scanning the QR code. You might say that we have apple pay, android pay. Yes, it is true, but that requires a POS machine. This QR code thing, is huge in China and it is a critical component for 'lite' app's success. Imagine the use case where you go to a restaurant and are waiting at a queue, with no one serving you, you scan the QR code and order all your food on the 'lite' app. I can image the 'lite' app bringing value in this particular case.

"Where should we find 'lite' app in Wechat ?"

There is no entry point for WeChat 'lite' app. This conforms with the decentralised vision for WeChat. if you do not actively seek for the 'lite' app, it would not appear in WeChat.

'Will there be an 'lite' appstore?'

No. There will be neither app ranking nor app distributing stores. WeChat has considered AI recommendation before, but since WeChat is a social application. We will leave this job to your friends.

"Do you subscribe to a 'lite' app?'

No. Going back to the Website analogy, you do not subscribe to a website.

"Will there be push notifications?"


"Can these 'lite' apps be shared?"

Yes, they can be shared to chats and chat groups but not to Moments.

"Can 'lite' apps be searched?"

Yes. But through a very restricted sense, avoid people taking advantage of the first mover advantage.

What is the relationship of 'lite' app and official accounts?

Not related, they are independent of each other however, through one, you would be able to see the other.

'When is 'lite' apps available?'

Tentatively, 9 Jan 2017.


It is not hard to see how the future is planned out for WeChat. However, track record has proven once and once over that social apps like WeChat and Facebook have natural life cycles. Back to the topic of 'lite' apps, I think because the name 'lite apps', people have some unrealistic hopes for it. It is certain (from Xiaolong's speech) that WeChat apps will not replace the everyday apps you use. However, on the battlefield of fighting for people's attention, WeChat might have just won again on the strategic level.


Deep Learning

deep learning red.jpg

AI was one of the most hotly debated technology topic in 2016, to the extent that almost all of the start-ups wanted something to do with it. It seems that no matter what a particular start-up's expertise is, they want to claim that they use some sort of AI
 technology or "algorithm" behind the scene. While as someone who worked at a VC before, I know that AI is definitely a turn-on for investors in 2016, but aside from investment reasons, can AI really do what those start-ups claimed?

Just a month ago, I was at a hackathon event organised by IKEA. They want the teams to come up with ideas for an AI chatbot, I listened to all of the teams' pitches. Most of the teams consist of university students. After the pitches, I realised a big problem -- majority of us do not understand what AI is really about. That is why I decided to write this piece to give it some thoughts and share with my audiences what AI is about and what AI can do. But wait, Why is the title Deep Learning then? You might ask. Well,  this is exactly my point! AI has existed for more than half a century already, it has also been revisited many times throughout its history, however this time it's all about deep learning! Thus the topic.

Deep Learning - a really really brief primer
The big break of deep learning started with the famous object recognition of course. In 2012, the Google Brain team led by Andrew Ng and Jeff Dean created a neural network that learned to recognize higher-level concepts, such as cats, only from watching unlabeled images taken from YouTube videos.

The above image captures the gist of deep learning. Using layers of data representation to make sense of seemingly unrelated data. As shown above, the input is a set of human faces. With those faces, DL is able to represent brightness contrast in the first hidden layer. From that, it matches the contrast with known objects like eyes or nose. In the hidden layer 3, DL could compare with the facial template to predict if the picture is a face.

Big Data & GPU
Two big enablers of DL are the increase in the sheer amount of data as well as the improvement in GPUs. The reason why GPUs are best suited for DL is because they are designed to work in parallel, and since DL need to run thousands of pictures to know what faces look like, GPUs are very good for DL. Furthermore, when you are training your AI using deep learning, you also need a lot of data to make the prediction more accurate, this is where Big Data comes into play, luckily we have been generating more data than ever.

Unsupervised Learning
Many people confuse DL with unsupervised learning, it is not true. Although unsupervised learning is a very important feature of DL. Many deep learning algorithms are applied to unsupervised learning tasks. This is an important benefit because unlabeled data are usually more abundant than labeled data. Examples of deep structures that can be trained in an unsupervised manner are neural history compressors and deep belief networks.

Deep Learning & its implications
Coming back to the event, all teams had incredible ideas for IKEA, but one interesting fact I noticed about the pitches is that people tend to under or overestimate the implications of AI and DL. Some groups seem to focus a lot on some form of virtual assistants,  the other groups on the contrary are visioning some kind of Terminator/ Westworld-ish AI that basically is like human. While it is certain that we are still far away from achieving Strong AIs that are able to exhibit human emotions, we definitely should not be so pessimistic either. The internal logic of DL does show us something that we have not been able to do before, and with the increase in the amount of data available, DL backed AIs would unveil a new way of collaboration, a form of collaboration we have never seen before!

Layers of representations in three iconic DL applications:
Image recognition: Pixel → edge → texton → motif → part →object
Text: Character → word → word group → clause → sentence
Speech: Sample → spectral band → sound →…→ phone → phoneme → word


We consume content everyday, in many ways and forms. They come from different channels, explicitly or not. It might be that Rolex ads right beside your YouTube video, it might also be the google ads on top of your search results, it can even be the legitimate content that you clicked on CNN that directed you to some other publishers websites. One big company that does content marketing is the one in the title, Outbrain. With infinite passion in marketing, advertising and PR and my prior experience with Outbrain engineers when I was in Israel, I researched on this company and here is what I found out.

Outbrain is doing well

Outbrain is the leader of the content recommendation market, with 25% market share. Currently the company generates annual revenue of 200 million dollars with a CAGR of 80%, pretty impressive. In May 2016, it raised 45 million dollars in private equity, there are rumours saying Outbrain may be going to IPO soon, but they seem to have backed off that route. On the left side  is an infographic summary of Outbrain's services.

Making money?

Brands and businesses

For brands and businesses, "Outbrain Amplify" promises to help you obtain the "attention your content deserves". As Outbrain partners with globally reputable companies like CNN, Fortune and so on, it is able to drive traffic to your website and of course in return, you would also need to pay-per-click for every audience that is brought to you by Outbrain. So you would be able to set up daily campaign budget (minimum $10) and the cost-per-click, you pay for every visit you receive until you reach your daily budget. You can also optimize reach by yourself or use Outbrain's account strategist. However, to qualify, you content needs to be informational or have entertainment value. However, on this note a reported 25 percent of those links are arbitrage links, meaning some website try to direct the traffic somewhere else to make money again.

Media Companies

The promise to media companies is that Outbrain makes the integration truly authentic. Regardless of the devices, whether it is an iPad or an iPhone or maybe a laptop. The modules that are built into the website would look just like a seamless part of it. They are customizable too.


The most obvious competitor is Taboola that is also established in Tel Aviv, they are almost doing the same thing - content recommendation. The main claimed difference is that Outbrain have a filter system to filter out the irrelevant or junk content while for Taboola, it has an interactive part that allows users to choose what kind of contents are not relevant to them


Currently, the most notable content recommendation company is baidu tuijian, it is a part of Baidu of course. Baidu made a strategic investment in Taboola in 2015 to make this happen. According to Baidu, they have the biggest market share in China with 7000 partner websites, 1 billion page views as well as 3 billion views to their native ads.

Ctrip acquires Skyscanner, what's its play?

Just a few days ago, the biggest chinese travel booking platform acquired the british Skyscanner for 1.74 billion dollars according to major media outlet Bloomberg and the Guardian, After Ctrip's merger with Qunar last year, they have collectively about 70 percent of the Chinese online booking market. With its acquisition of Yilong in 2015, it has seemingly become the biggest travel sites of China. However, it has been facing very strong competition from Tuniu. According to the latest Yiguan analysis, Tuniu for the first time, has surpassed Ctrip to become the biggest Chinese online travel booking agency.

Credits going to Yiguan. For the first time in many years, online travel industry is not dominated by Ctrp, its market share dropped from 28.2% in 2015 to 22.4% this year. Hence the question comes. Why is it still acquiring Skyscanner ? Should it worry about its competition instead of continuing to expand?

Tuniu and Crip has widely different business models.

To understand this, one need to understand that these two companies have very different business model. For Tuniu, they buy flight tickets and hotel stays wholesale, packaging those tickets and sell it out in the form of trips. However, Ctrip has a much simpler commision model, it helps hotels and aviation companies sell their rooms or flights, simple as that. So the competition between them are hard to resolve.

Skyscanner acquisition reflects the market's prediction

Ctrip's acquisition of Skyscanner reflects the company's vision and belief that there is a behavior shift among Chinese travelers. From domestic and regional to global. With the growing middle class in China, more and more evidence has pointed out that Chinese travelers are not satisfied with regional trips anymore. They want to experience what the rest of the world has to offer. Skyscanner helps them with that.

Does Ctrips has global ambitions?

Well the simple answer is Yes. But not quite. Ever since the news of Skyscanner acquisition came out, western sentiment is that Chinese company is expanding and buying again. But what they fail to understand is that, China has a huge untapped market with 13 million skilled workers coming from villages to the cities to be plugged into the global machine. This means that Chinese companies like Ctrip would never need to worry about expanding globally, they would be hugely successful as long as they provide a good service in the domestic market.


任何一个认真的创业人,都希望自己都能走在时代的最前端。不说能大致理解新鲜的一些概念比如区块链,以太币,Deep Learning之类的,至少不能别人说到一个非常火的概念的时候却一无所知吧。我第一次听到IP的时候就是这个感觉,被吓得魂飞魄散。对,这里的IP就是 intellectual property的缩写,意思是知识产权。但是这里说的IP可和传统意义上的IP有着天壤之别。

IP在Google上的解释是 intangible property that is the result of creativity, such as patents, copyrights, etc.意思就是说,IP是通过创新取得的无形的的资产。就比如说一部华为的手机从材料到生产到组装最后的销售,本身的价值也许只有600块钱,但是由于手机里面某个芯片的创新,这部手机的价格就大大的增加了。这里的知识产权是为了保护开发这个芯片的华为的利益的。同样,Google在自己的产品上申请了很多的IP,比如击败了传奇围棋选手李在石的AlphaGo就拥有很多知识产权。AlphaGo背后的公司DeepMind,如果没有知识产权的保护,恐怕那些尖端的算法早都被别人抄去了。

可是我们今天要讲的不是这个。正如我上面这张图看到的。这里我们所说的大IP,广泛意义上来讲是那些被广大受众所熟知的、可开发潜力巨大的文学和艺术作品。IP不仅仅只包括简单的影视形象而已,从宏观意义上来看,IP可以理解成一个关于“创造力”的法律术语。它可能是一个承载各种新奇想法的工具(比如乐高玩具),也可能是一种传统(比如泼水节),更可能是一个组织或传统(比如门萨俱乐部和美国童子军BSA)。据罗振宇在他的时间的朋友里说,随便一集IP写手的电视剧稿费就能有80万之高。像琅琊榜花千骨这样的网络小说也能卖出上千万。这就是网络世界里IP的威力。IP已经从以前的产品技术产权摇身变成了知识人格产权。从根本上来讲是现代社会人们需求转变的体现,也是产品升级的体现。人们现在已经不能满足于买卖产品了,因为这个市场上产品太多了,好的产品也太多了。多到以至于产品差异(differentiation) 变得越来越小。就拿手机举个例子,苹果为什么能以15%的市场份额获得80%的市场利润,原因就是如此。像我们这样的普通人根本就分不清高档手机的区别在哪里。本质上也许就没有什么区别,可是苹果产品和产品背后所代表的“人格产权"是其他品牌代替不了的。直到现在我们想到苹果,我们还是要刮目相看的,原因是它对产品完美的追求。所以苹果的IP就不仅是技术产权,还有它的"文化产权”。




最后我来附一张Casey Neistat的(我的偶像)照片,并不是我偏心。他确实是现在最炙手可热的YouTube IP了。