Geir Freysson

From Iceland. Live in London. Run a startup.

Dear Dropbox: Hire some web engineers

Yes, that is a giant Panda. #dropbox

A photo posted by Malcolm (@ayeitsmalcolm) on

Business insider had a story recently on how Dropbox is cutting costs. It involves a $100,000 giant chrome panda. With all the cash they could save by stop investing in chrome, maybe they can hire some web engineers?

I’m hooked on Dropbox. And I’m a paying customer. I collaborate with others in the Datasmoothie team via Dropbox and I have Word on my iPad which connects directly to my Dropbox folder. Beautiful. And until recently, all my photos were backed up to Dropbox as well.

But Dropbox has stagnated on the web. Log into the Dropbox web interface and your folders will look exactly the same as they did years ago. It feels like a flashback to 2010. One recent attempt to innovate on behalf of the user, the photo service called Carousel, was a usability and performance disaster (the mobile app was better). Dropbox pulled the plug on Carousel earlier this year (see the Dropbox announcement).

Contrast this to Facebook. Facebook seem to attract the biggest brains in web technology and always seem to be racing forwards with regards to speed and usability. And their team have become thought leaders in the technology that drives the web, like CSS and Javascript. They release much of the technology they build as open source software and it simply rocks. Datasmoothie’s front-end is built on one such piece of technology, called React, and I can’t imagine life without it. I can’t see anyone using a Dropbox open source library for the web any time soon.

Dropbox are awesome on the back-end. Syncing your folders seamlessly across different operating systems is no small task and they do it well. But they need to replicate that on the web if they want to remain competitive. So as they morph into a more lean operation, spending less money on statues, they should consider spending cash on hiring some leading lights in web technology.

In the meantime, my photos are backed up by Google. Maybe in a few years time, my documents will be too. Unless Dropbox hires some web engineers.

The “rags to political office” pitch and dirty campaigning


The results from the London Mayoral election are sill be announced later today. A few days ago the above pamphlet dropped in through my letterbox. “The council estate boy who’ll fix the Tory housing crisis.” I’m always a bit wary of campaigning that goes into class warfare terrotory, but this particular pamphlet manages to stay on the right side of that line. It says “I’m one of you” without saying “and we hate them”.

I didn’t get a pamphlet from Zack Goldsmith. If he loses the election I’ll remember him as the guy who went nuclear and declared that Khan was unfit to:

protect Britain’s capital from a terrorist attack because of questions about his links to Islamist extremists

Khan is a muslim so he can’t protect London from terrorists. Sheesh. Whichever way the election goes, I’m rooting for Khan if only so that dirty campaigning doesn’t pay off.

Analysing data about the world’s billionaires with Datasmoothie

At Datasmoothie we love data sets from interesting places about interesting topics. So when we came across a detailed list of the world’s billionares since 1996 we were intruiged. We decided to upload the the data into Datasmoothie and play around with it.

The result can be seen here: Billionaire, reveal thyself.

I’m pretty pleased with what Datasmoothie is turning into. You can take a dataset that looks like this:

Screenshot 2016-04-20 16.27.45

And turn it into something like this in a matter of minutes (or seconds if you’re really quick):

Screenshot 2016-04-20 16.26.10

Click on the map above to see it live in the report. You can filter the super-rich down to industries and gender.

It’s exciting to see Datasmoothie take shape and become such a powerful tool. We’ll be demonstrating its power with more reports over the coming weeks, so stay tuned!

Four ways HipChat can catch up with Slack

Our team has been using HipChat forever. Since before Slack was cool. Yesterday I got an email from Atlassian (who run HipChat).

New: Request your next Uber in HipChat!

I use HipChat every day. I use Uber at least once a month. I am never going to order an Uber car through Hipchat. I’ll probably stick to doing that on … Uber.

My reaction to the email? Moan on Twitter of course. That’s what Twitter is for. It’s the moaning app.

Screenshot 2016-01-28 11.47.12

Hah. Clever me. Sometimes you forget that Twitter is public, though. And that the product manager for HipChat might see your tweet since they’re probably monitoring social media. And indeed he did.

Screenshot 2016-01-28 11.50.06

All of the sudden I had become that annoying semi-anonymous guy who randonly trashes a product because he isn’t into ordering Ubers from chatrooms. The feature is the result of a Hackathon and an apparently very cool API. That’s great.

So to make up for my moaning I thought I’d be constructive and put together a  list of how HipChat can actually prevent our team from eventually leaving to use Slack instead.

1. Match Slack’s most basic features like starring (bookmarking) a particular line. When someone explains in our dev channel how they did some magic in the statistical package Pandas, other devs need to be able to look that up quickly.

2. Keep tweaking the design to keep it current. Slack looks better than HipChat. Try inserting a quote or code snippet into HipChat vs. Slack. HipChat looks a little 2008. And why not stick a thumbnail of users next to what they say. That’s cute.

3. HipChat should be a platform. I’m not just talking about an API but awesome integrations built by Atlassian or the community that do more than just post from the internet to HipChat channels. Lots of applications are pushing notifications into our channels, but we can’t pull anything. If I type (stock:amzn) I want to know what Amazon’s stock price is. If I type (datasmoothie:reports) I want a reply with how many reports our users have generated.

4. Concentrate on the chat experience. HipChat should be obsessed about the chat experience. Aside from look and feel and sci-fi dreams of pulling data into HipChat, what can it do to improve chat? When I post a link to a Dropbox screenshot, don’t just show me a broken image link. Poll a few times and update it once it’s available. Play a different sound according to what user mentioned me in a room. Allow custom mention icons that could also play a sound. Typing @alarm could play an emergency sound and ping everyone. “@alarm We have a live show-stopping bug!”

If the new API that was the backbone behind the Uber feature will start fostering more innovation on HipChat as a platform, maybe they’ll come up with something compelling enough for us to stay. Good luck to them. And hell, maybe everyone will start ordering Ubers on HipChat. What do I know?

Screenshot 2016-01-27 09.14.23

MIT PhD student explains why self-driving cars (probably) won’t lead to the robot apocalypse

Machine learning is all the rage these days. Last year’s FT & McKinsey Business Book of the Year was “The Rise of the Robots”, a bleak prediction of how machine automation will take more and more jobs. In July a letter signed by 1,000 tech experts, among them Steven Hawking, Elon Musk and Steve Wozniak, warned of an impending “AI arms race” and in December Musk and the startup accelerator Y-Combinator announced OpenAI, an AI research institution meant to counter the threat of Robots taking over.

Apparently, if World War III ever comes around it will be between humanity and Google’s self driving cars.

These worries are not entirely unfounded. Especially the part where machines will take certain jobs (I sure hope you don’t drive a car for a living). But most of the recent progress that has been made has been in using neural networks and deep learning for pattern recognition. A machine might understand what you are saying, but not what you mean. There’s a lot more to human intelligence than pattern recognition.

In The Unreasonable Reputation of Neural Networks Luke Hewitt, a PhD student in in the MIT Department of Brain and Cognitive Sciences, writes

Deep learning has brought us one branch higher up the tree towards machine intelligence and a wealth of different fruit is now hanging within our grasp. While the ability [for machines] to learn … is both new and exciting, we should not fall into the trap of thinking that most of the problems an intelligent agent faces can be solved in this way.

Recent advances in neural networks mean that a lot of jobs are under threat. But they don’t mean that we are within reach of creating the next Skynet.


The democratisation of branding

I’ve discovered two restaurants recently via the food delivery service Deliveroo. Giving the founder of Deliveroo a hug is on my bucket list but that is another story. The two restaurants are achieving great heights in the, erm, culinary space. One is Homeburger, the other is Bird and both are relatively close to each other on Holloway road in Islington.

Once upon a time you could look at these kinds of establishments and instantly see whether they were big chains or run independently. The logo, signage, uniforms of the staff, etc. were classic giveaways. Both Homeburger and Bird look great from a branding perspective. So it came as a surprise when I discovered that Homeburger only has one outlet and Bird has three.

Sites such as, where you can crowd-source your branding material, have made it affordable to look good without paying a fortune. Remember when the 2012 Olympics folks paid £400,000 for their branding? You can get away with less these days.

But it’s not just the availability of 99Designs. It also seems like restauranteurs are becoming more brand savvy, more of them are getting their branding right. At Homeburger for example, unwrapping a burger is almost like unwrapping a new Apple product. Very slick packaging.

Better branding, along with having Deliveroo deliver to customers doors without having to employ a guy or gal on a scooter, hopefully means that the indies are will do better and better. And if this means I have more access to such wonders as Bird’s chicken burger with bacon, I’m all for it.

Football statistics visualised (and how data is giving Arsenal an edge)

In sports the term “moneyball” refers to using statistics rather than gut instinct when it comes to recruitment and strategy. Recently, football is being moneyballed to a greater and greater extent. My London neighbours Arsenal are ahead of the game when it comes to moneyballing, which shouldn’t come as a big surprise: Manager Arsene Wenger has a degree in economics. Just over a year ago they acquired football statistics company StatDNA and are reportedly both using them for recruitment and strategy (buying Gabriel Paulista and playing Aaron Ramsey in central midfield).

A lot of data is available in football. I recently came across an excellent dashboard with all goals and assists in the current season visualised on the pitch. It can be used to nicely highlight the quality of the Premier League’s top performing player, Mezut Özil. Özil is well on his way on breaking the record for most assists (a pass that leads to a goal) in a season. The season is only half way through and he has 16 assists with the record standing at 20 assists in a season.

Below Özil’s assists can be seen as compared to last season’s champions Chelsea, who haven’t had a worse start to a season since 1978 (see’s excellent post on the topic).

Arsenal’s Mezut Özil has more assists this season than Chelsea’s entire team.

Other interesting visualisations include all shots on target this season (left) and all of Jamie Vardie’s goals (right). And yes, Bournmouth scored straight from a corner against Manchester United.


Because of the StatDNA acquisition Arsenal are the only club in the Premier League that gather their own data. Whether Wenger will be Football’s Billy Beane, the real-life protagonist of Moneyball who revolutionised baseball with statistics, remains to be seen. But the club certainly has an edge to their rivals thanks to Wenger’s open mind and forward thinking.

Machine learning in the real world

A long time ago in a galaxy far far away I studied neural networks, the methodology that underpins much of machine learning, as part of my master’s degree in computer science. At the time it was a technology with a lot of promise and many flaws. In the following years I was always very sceptical when someone created hype around intelligent machines (I’m looking at you Siri).

Over the last 3-4 years however the technology has leaped forwards at an incredible pace. Deep learning (a rebranding of neural networks?) is especially hot these days, as the below Google Trends graph shows:

If I was an investor I’d be giving startups that are using this technology a very close look. Which is why a recent blog post by just such an investor caught my eye recently (via the excellent Data Elixir).

In Machine Intelligence In The Real World investor Shivon Zillis discusses the various types of machine learning companies that are making waves in the tech world from IBM’s Watson to The Grid. Zillis makes an attempt to categorise machine learning startups into groups such as those that “solve industry-specific problems with laser-like focus” to others he calls “alchemists” (no pun intended) that promise to turn your data into gold.

If someone would have told me ten years ago we’d have self driving cars in 2015 I would have laughed in their face. It will be exciting, and slightly scary, to see what the world of machine learning will give us in the next decade. I look forward to it.

Illustration from the Economist

The Economist on the problem with Clayton Christensen

The Economist’s Schumpter column published an article on Clayton Christensen last week called Disrupting Mr Disrupter. The author isn’t quite as harsh as I was the other day but is sceptical nonetheless:

The problem is more that the definition of disruption he seeks to impose is too narrow. He rules out Uber because, from the start, it offered a better level of service than existing taxi firms, rather than something cheap but inferior. But ask any cabbie if it threatens to disrupt his business, and you will be left in no doubt of the answer. As Isaiah Berlin, a philosopher, would have put it, Mr Christensen is a hedgehog (someone who knows one big thing) rather than a fox (who knows lots of little things): his hedgehog mind leads him to ignore or belittle companies or market forces that do not fit his template.

I couldn’t agree more, but I would go further. What makes Christensen’s work so seductive (he was my favourite business author for many years) is that it is presented as emperical research, based on data, instead of the self-help anecdotes that dominate so much of busines literature.

But the data doesn’t confirm the theory. The MIT Sloan Management Review recently published an article, How Useful Is the Theory of Disruptive Innovation? that concludes that

few quantitative tests have been performed … [and] the ones that have been published fail to provide confirmatory evidence for the theory …

Christensen counters that “the lack of numerical support is the result of the blunt measures used in statistical analysis” where the “more nuanced case analysis” works much better.

It’s starting to sound like “case-based research” is a fancy word for “anecdotal evidence” and theories based on it should not be taken too seriously, no matter how useful they may sometimes be as general guidelines or war stories of what has worked and what hasn’t for others in the world of business.


Board games are getting better (confessions of a former role player)

At the time I would probably not have admitted to it in certain circles, but in my teens I was an enthusiastic player of role playing games such as Dungeons and Dragons and Cyberpunk. The rule book for the latter is pictured above. Somewhat alarmingly, it was meant to take place in the “dark future of 2020”. My circle of friends (think The Big Bang Theory where everyone is Icelandic and there is no babe across the hall) also enjoyed a break from the role playing every now and then and played board games instead. Among them was Space Crusade, pictured below.

Screenshot 2015-10-13 10.31.39

According to the board games rating site BoardGameGeek, Space Crusade was just above average in quality when it was released, in 1990, with an average score of 6.8. But these days, board games are getting better. Nate Silver’s FiveThirtyEight recently analyzed the ratings on BoardGameGeek and reached the conclusion that over the last decade, the average user rating for board games has been on the rise (see Designing The Best Board Game On The Planet).

Golden age of board games

And it seems to be a virtuous cycle. Board games are getting better and they are also getting more popular. The industry is in fact absolutely booming. In a recent article, Not twilight, but sunrise, the Economist reported that in America and Canada alone the board games industry is estimated to be worth $880 million, showing a double digit annual growth for the past half-decade. This year’s Spiel Tage in Germany, one of the biggest board gaming festivals in the world, expects more than 160,000 visitors over the course of four days. While I wasn’t looking, board games seem to have gone mainstream. Or as the FT reports from Spiel Tage:

Beyond the sheer number of enthusiasts, the striking thing is that they look, well, normal.

I’m quite looking forward to the time when I can introduce my daughter to board games, not only because I’m sure it will make me super popular (ehem) but also because it’s the perfect excuse to revisit the concept after a few decades break. It seems that games that I had never heard of until now, like The Settlers of Catan, could be a place to start. But since games are steadily getting better, we might have a wider range of choices when she’s old enough to start appreciating them.

Page 1 of 15

Powered by WordPress & Theme by Anders Norén