An intelligent Twitter client?

Do you think we should make yet another twitter client?

My gripes about existing Twitter clients is that they aren’t much helpful in taming down information overload, dampening the noise and removing spam. A lot of these 3rd party clients just repackage the Twitter website or clone the official client. They don’t do much data processing to improve the user’s experience to help them gain insight and engage in meaningful conversations.

The problem makes me feel that we should take the matter into our own hands. That is making a Twitter client that can solve these problems. One use case is for extracting meaningful contents and conversations from a large number of follows. Another key feature is something to help you discover and connect with like-minded people. Something that makes the thinkers stands out above the sea of airhead teenagers and opportunistic trolls.

With these in mind, the application will probably have zero features meant for amassing herds of Twitter followers. There is plenty of stuff out there claiming to do this. Some even allows you to procure Twitter followers. I found that a large number of these followers are useless anyway since they are likely to be spambots and mindless meta-marketers. If you need something to stroke your Twitter ego (tweego?) please move on – nothing to see here.

My re-discovery of Twitter

It’s been about four months since I started seriously using Twitter again. I registered my first Twitter account back in 2009, but I was not sure what to do with it. Thanks to the miliscoble that lead me to think that Twitter was for letting off steam and venting out crap. So at the time my account only served as a sink for announcing my blog posts and various public Internet activities automatically delivered by a clever combination of Yahoo Pipes and

About five months ago at the influence of some podcasters Twitter again made its way into my mind. I began to actively and personally use Twitter, trying to see “what’s all the fuss about this Twitter thingy?” So, I started doing these:

  • Tweeting parts of my personal life
  • Sharing news & interesting articles
  • Tweeting my own blog posts
  • Following a number of tech celebrities and commenting their tweets
  • Got involved in a number of ad-hoc tweet-discussions.

Things that have been working so far

Talking with famous people

I found that Twitter is great for interacting with minor celebrities or thought leaders — people well regarded in their respective niches. Usually I chat with technology and small business activists in Twitter. They tend respond nicely to my queries and comments. This is quite a change from contacting many public figures via e-mail, which often goes unanswered.

In case you’re curious, who are these so-called “celebrities”? These are some of them, in no particular order.

“But I don’t know any of them”, you might think. Yes, that’s why they are called minor celebrities. The individuals listed above are pretty famous in the micro-startup and independent developer scene. But you only know them when you are “in” this field (if you are and you still don’t know them, it’s about time you do). You probably also know some relatively famous people in your own industry.

Content discovery

Another great use of Twitter is discovering relevant content. That is, finding articles, video, etc that are interesting for me. Finding content in Twitter is somewhat different from subscribing to blogs or websites via RSS feeds. When you subscribe to a website’s feed, you get a steady stream of content from only their site. In contrast when you follow a person’s Twitter stream, you get content from other sites that they find interesting along with a short comment of what they think of it.

You might think, how does this differ from Facebook? For one, you don’t need approval to follow people’s stream (in Facebook, you both need to acknowledge each others as “friends” before you both can see the other person’s activity stream). In addition since Twitter streams are public by default, there are less personal information that gets posted which may be not interesting outside the person’s real-life friends and family (i.e. those baby photographs, family gathering pictures, etc).

How do you get relevant content? By following people with similar interests. For example, I first found out about the Lodsys issue in May 2011 on a Saturday morning via Daniel Jalkut’s tweet. Then about patent troll bootstrapping model via Florian Mueller’s tweets and his blog. Of course if you don’t have diversity in your follows, you’ll get caught inside a filter bubble of online groupthink.

What are not working so far

Small business marketing

I didn’t find small business marketing to work in Twitter. Specifically, marketing our niche software products via Twitter doesn’t produce any measurable results. I followed Pluggio’s methodology and signed up with Pluggio for the past two months with the hope that this can help improve recognition and sales of our products. Within these two months there wasn’t any noticeable increment in the sales figures. Similarly click-through rates from Twitter are also low. Yes, the number of our Twitter followers grew, but we don’t have a humongous ego that needs stroking with a large number of Twitter followers and Facebook friends. So in essence this kind of Twitter marketing didn’t work well for our products.

Why it didn’t work? My guess is that people are not in the “ready to buy” mode when browsing Twitter. This is the key to selling niche products on the Internet – you want to target people who are looking for solutions to their problem and thus ready to buy your stuff (or your competitor’s stuff) which solves their problem.

Lazyweb queries

Asking open questions in Twitter – also known as #lazyweb queries — also didn’t work. Some bloggers may argue that this isn’t true since they get good quality replies most of the time. Probably it works for them since all that they do is just talk on the Internet all day and night and their followers are all hungry for attention from him/her. It didn’t work for me that still need to take care of a life-sustaining dayjob in addition to a nightjob bootstrapping a software startup.

What are the challenges for me in Twitter

The biggest problem I face is the signal to noise ratio. Specifically there is lots of irrelevant noise compared to actionable or useful information. People being people have many facets and not every tweet is interesting nor relevant. Like for example I have zero interest to which baseball team is playing this season, and I would like those kind of irrelevant tweets de-emphasized from my view, even if they came from the people I follow. Yes, they do have a right to say whatever they want, but I also need a way to filter out the unnecessary baggage.

Another challenge is in finding interesting people to follow, more importantly interesting to me. That is their tweets are relevant to the things I’m doing right now or I will be doing in the near future. Twitter’s “who to follow” lists isn’t useful for this since they tend to cater for the “mainstream”, the average, whereas I have specific needs.

Proposed solution

These are the features that I believe will solve my problems above. Strangely, there isn’t a twitter client that fits the bill. I’ve yet to see a Twitter client that has all of these features or will solve my problems above by some other way.

Why am I sharing this? Aren’t I afraid that you would steal my idea? Actually I’m thinking out loud and refining my idea with potential input from you. Besides, in the software world, people don’t buy ideas – they buy implementations of ideas (that’s why software patents should be abolished from the world, but let’s stay on-topic and save that for another post). Ideas without implementations are practically worthless.

Word Clouds

Word clouds can be used to summarize your timeline. This way you can get a bird’s eye view (pun intended) of what’s happening in your world (or more precisely what are the people that you follow are talking about). From this view, you can dive in to individual tweets that you find interesting.

Wordle word cloud of this article

Another good use of word clouds is for discovering who to follow. They can be used to summarize what a person generally tweets about. From there you can easily decide whether he/she is interesting enough for you. The idea goes like this. First you enter some keywords of the things that you’re interested in. Then the application will search Twitter for people who said those words. Afterwards you will be presented with word clouds of these people so that you can make decisions whether they are worthy to follow.

Relevance Filters

This is yet another tool to help you see more valuable information from Twitter and tune down much of the noise. Ideally there should be commands like “show more tweets like this” and “hide tweets similar to this”. In response the software will only show tweets that are relevant and important to you. In the real world this may not always be possible and the software could be wrong.

The next best alternative is to de-emphasize tweets that the system sees as less important instead of completely hiding them. That way you can still make corrections and the application can gradually learn from your preferences. One way to do this is to have the application re-order your incoming tweets based on a scale of importance. It could use bayesian filters to achieve this. Then again there are other methods of dynamically prioriting text messages.

Spam Filter

This is pretty similar to e-mail spam filters but applied to tweets instead. Based on my personal experience, the algorithms to detect twitter spams should be far simpler than e-mail spam detections since spam tweets and spam accounts are easy to spot (I’m not really sure why Twitter didn’t do this automatically already). Obviously I wont reveal my idea here since it would easily tip off the twitter spammers and made the spam detection algorithm useless. Another useful feature in this area is to quickly block spammers and report them to Twitter.

Thread support

Public ad-hoc discussions happen in twitter all the time. It’s a shame that a lot of Twitter clients don’t recognize this type of tweets and treat them in a special way. If two or more people that you’re following were having a debate, it would make sense to reorder their tweets in your view so that their conversations are placed near each other. That way you can quickly glance what they are talking about and be more informed should you want to jump in their discussion (since this is a public conversation in the first place, they shouldn’t be bothered about it as long as what you say is relevant).

Coalescing of re-tweets

If more than one person re-tweeted the same message, that’s a tweet that you need to read twice. One way to solve this is to combine re-tweets together and add an indicator how many people re-tweeted this message. But what to do about quote-retweet? That is what if someone adds more information to the retweet? I haven’t figured out how to solve that one yet — please tell me your ideas in the comments section below.


This is a pretty long blog post. It also took about two weeks and many failed drafts for me to gather all the thoughts and come up with a somewhat coherent article that are you reading now.

In bullet points, these are what you’ve learned from reading this post:

  • Twitter are useful for interacting with celebrities and discovering interesting content.
  • Twitter aren’t much useful for small business marketing and asking random questions.
  • The biggest problem in Twitter is the high signal to noise ratio.
  • Existing Twitter clients aren’t much help in distilling information and discovering like-minded individuals.
  • This article proposes a number of features for a next-generation Twitter client, which centers on refining tweets via word clouds and filters.

Do you agree with me? What are your problems in Twitter? What features would you like to see in a Twitter client but you can’t find it anywhere? Please tell us in the comments section below. In addition, we’re developing our own Twitter client and you can register your e-mail address to be notified when we launch.


join our mailing list
* indicates required







Get articles like this sent to your inbox as soon as they are published.

* indicates required

Unsubscribe any time and we won't share your details with third parties.

Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , ,

%d bloggers like this: