The Ultimate Star Wars t-shirt

After toiling for eons in the my idea-factory, I finally present to you the Ultimate Stars War's t-shirt:

star wars t shirt ultimate

You will turn heads as you move through the crowd wearing this instant classic.

Whether it is the thrill of seeing their favorite character, the delight of the perfectly on-brand typography, or the joy of such precision alignment, any TRUE FAN will immediately grab you by the shoulders and demand "WHERE CAN I GET ONE!?"

The answer is: right here my friend:

Ultimate Starwar's t-shirt

Nerdy discussion that ruins the fun...

Now that you've finished ironically buying one, I want to talk about product ideas for a moment.

I tweeted about that shirt last week, and my mentions immediately filled up. It was an obvious hit within minutes.

This doesn't happen to me very often, so I like to take notes when it does. And this blog is where I put the first draft of those notes.

Fortune of course favors the prepared Boyscout or something like that. It only took a few minutes from having the idea to unleashing the product; it was ridiculously quick. But there were hours of behind the scenes preparation that put me in this position: last year I hosted the inaugural Stupid Ideas Powershell Slackathon, during which I created a command-line tool that let's you turn error logs into t-shirts: Out-TShirt. Developing that also meant I was already signed up and approved to sell at teemill. And it meant that once I'd completed my work in MSPaint, I could turn it into an online shirt (and put the resulting url in my clipboard) with a single command:

dir "bad_shirt_idea.png" | out-tshirt | clip

When you're trying out product ideas you're looking for a demand signal. If you don't get a demand signal, then you ditch the idea or make a big change. Once you get the demand signal you only make small changes. You don't keep trying to sell a product that won't sell, you propose little ideas, continually, and see what the market tells you. If you've done well, you'll hear about it. If you've done badly, you'll hear nothing.

In the long periods between seeing 'demand signals' it's easy to forget that these signals exist. So, as a nerd and an archivist, I want to record some of the forms that the demand signals took in this instance. This is the kind of detail I gather up and try to understand, as research for my forthcoming book on making products.

Some of the things I saw happening in my twitter mentions were:

The signs may be more subtle than they were in this case. But the only point I'm making is: look for the signs. If you see the signs, then take advantage of them. If you don't see the signs, move on.

Now go sign up to be notified in case I ever finish the god damn book. (p.s. I *will* finish it.)

p.s. the shirt is available in black and yellow now too:

black and yellow 50


pushpop, a mental stack manager

You're working on one problem, when another problem arises... or perhaps an interruption occurs... now you need to push the new problem on top of your mental stack, and the old problem moves down deeper into your soul. You start to solve the new problem, but remain ready to come back and solve the previous problem.

What do you do?

everyday problems

Turn to pushpop.

pushpop is a mental stack manager.

It lets you see the top of your mental stack.

push pop


When you've completed a task, click pop. (it makes a nice popping sound) and that task is gone.

The next item rises back to the top of the stack.


But wait... you are interrupted in your quest to complete the item on top of the stack.

What do you do? You enter the new item into the textbox, and push it. Now that item is on top.

push pop pushed


There's one more feature... peek...

push pop

Peek lets you inspect the entire mental stack. You can drag items around to reorder them. You can search for items (and can move the item you find to the top of the stack).

Finished peeking? Click peeking... to hide the stack away.

OK, OK, nerdy note... in proper 'stack' parlance, 'peek' is the operation that lets you see the item on top of the stack without popping it. Looking at the entire stack is called 'snooping' (well that's the slang term we used at university when i was kid...) But I'm using Peek here because I like the aliteration. The CLI would call it 'snoop' however.


While peeking, you can also view previously completed items, if you click history....

Current status

As you can see from the balsamiq mockups above, this is just in the idea phase. It's based on Next Action


There could be a command-line interface.

pushpop --peek                 # shows you what's on top
pushpop --push "debug the sql" # puts a new item on top of your mental stack
pushpop --pop                  # returns the top most item, but also removes it from the stack (and archives it?)
pushpop --snoop                # shows all items currently in the stack, from top to bottom.
pushpop --history              # shows all completed (i.e. previously popped items) including datetime of completion.
pushpop --help                 # provides help with pushpop and life in general


I've created a repo for the project, but written no actual code as yet.

pushpop on github

Implementation is left as exercise for the reader. I suggest electron. Send me a pull request when you're done.

If the stack was stored in the cloud then you could also have iOS/android version for managing your mental stack while on the move.

Last thing

I didn't mention Yak Shaving. Why didn't I mention yak shaving? Here's a yak in need of a shave:

Yak needing a shave

Yak image from wikimedia commons


Productivity Flowchart: 13 years in the making

TL/DR: Go here:


The very very first article I tried to write for this blog (before it was even a blog) was called 'GeekBlock' waaay back in September 2003. The article was about a process for getting un-stuck. It has always amused & annoyed me that the article sits there unfinished.

Eventually I rewrote it as an article at with the humble title: 'A 3 Minute Guide to Fixing All Your Productivity Problems Forever.' But its non-linear structure was simply too tricky to follow. People who read it said they got lost about half way through.

Finally, I rewrote it as a tiny application; an interactive flowchart, a choose your own adventure, a tiny state machine. Now I'm happy with it. You can play with it here:

Productivity Flowchart

Some people have suggested I re-write it as an infographic (What is this, 2015?) but I like it as it is. It'll be due for a rewrite in another 13 years, around 2030, I guess.



Last year I spent a few weeks completely away from all computers. In this primitive state of savagery (that some refer to as a 'holiday') I took to playing a little Sudoku each day, using old-fashioned materials known as pen and paper. Let me tell you, pen and paper are severely over-rated. Devoid of any computing facilities, I found that — much like Porky Pig whom, whenever he's forced to diet, begins to fantasize about food — I began to fantasize about programming. And Sudoku in particular got me really excited.


I decided there were three interesting challenges here:

  1. How to start with nothing and create a fully solved board.
  2. How to remove numbers from a fully solved board until an interesting game is created.
  3. How to solve a partially completed board: i.e. how to play sudoku.

Start with nothing and create a fully solved board.

In my techno-isolation I came up with an algorithm that I thought would have no trouble generating solved boards.

Here it is in pseudo code.

Do this for each number from 1 to 9, using variable "i".
    For each row:
      Keep doing this until you've placed the number i somewhere on that row:
          Pick a random cell.
          Check that the cell is empty.
            And that "i" does not appear in the current column.
            And that "i" does not appear in the current 3*3 subgrid.
          if all three criteria are met, then place "i" in the current cell.          

I thought to myself... will this process get stuck? Of course it won't, I thought. How could it!

This is because, as it turns out, when I am away from a computer, my brain turns into useless jelly.

When I was finally sat in front of a computer and wrote a version of the code (using LinqPad) I found that very often it did break down.

Here's an example:

_   4   _   |  _   _   3   |  2   1   _  
_   3   _   |  2   _   1   |  4   _   _  
1   _   2   |  _   4   _   |  _   _   3  
_   2   _   |  4   _   _   |  1   3   _  
3   _   1   |  _   2   _   |  _   4   _  
4   _   _   |  1   3   _   |  _   _   2  
_   _   _   |  _   1   4   |  3   2   _  
2   1   4   |  3   _   _   |  _   _   _  
_   _   3   |  _   _   2   |  _   _   1   ← Try and place a "4" on this row!

Sometimes, the numbers that have already been placed will put so many constraints on the next number, that there is no solution.

I ran the program a few times and found that sometimes it stopped at 3 numbers, sometimes it made it as far as 6 numbers.

I thought I would need to write some clever algorithm that tries to shuffle the problem numbers around, to break the deadlock.

In the example above, we could shuffle some other '4's around, to come up with a workable solution... for example....

_   4   _   |  _   _   3   |  2   1   _  
_   3   _   |  2   _   1   |  _   _   4 → move this 4 from first to third position
1   _   2   |  _   4   _   |  _   _   3  
_   2   _   |  4   _   _   |  1   3   _  
3   _   1   |  _   2   _   |  _   4   _  
4   _   _   |  1   3   _   |  _   _   2  
_   _   _   |  _   1   4   |  3   2   _  
2   1   4   |  3   _   _   |  _   _   _  
_   _   3   |  _   _   2   |  4   _   1  ← Can now place this 4.

And if shuffling 4's above isn't enough, we could expand the range of our shuffling.

But then I had one of those Engineering Breakthroughs you read about in clever articles on Hacker News. (Mathematicians are never as clever as this.) "Sometimes it hits a deadlock by the number '3'... sometimes it gets as far as '6'... what if I just re-run it a bunch of times, and see if it ever completes?" aka, The Retry Pattern.

So I put a loop around it and said, "keep going until you're stuck, then restart from scratch, over and over."

And thus, within a second, it would always succeed at creating a sudoku board. Sometimes it would have to start over only 3 times, sometimes it would be over a thousand times. But I always had a complete new sudoku board in much less than a second.

That was good enough, and no need to perform any tricksy mathematics to "satisfice" the problem.

I came up with similarly simple approaches for the other two problems. I've put some c# sudoku code online.

I can now type "linq sudoku" (from powershell) and see a fresh sudoku to solve, any time I feel like it. Which it turns out is not very often.


2016 by the numbers:

After reading patio11's voluminous rundown of 2016, I decided to run some numbers of my own. Here's what I came up with as far as creative output outside of work and family time:

No wonder I feel soooooo tired.


Television Bankruptcy

I am declaring myself bankrupt on consumption of new media.

In particular I've declared television bankruptcy. Heard of some amazing new show? I don't care anymore. My ears are shut.

It also means that I can't ask anyone else to avoid "spoilers" on anything. I just assume up front that if I haven't watched it yet I may never watch it: even if it's a new seasion of show I love. Spoil away. I'm a TV bankrupt, I can't take any new loans against future viewings.

On the plus side, this frees me up to read about stuff on wikipedia in advance of seeing it though... some things I think I enjoy reading about more than actually watching, for example, critically acclaimed films from before the 1980s... often boring to watch, but interesting to learn about. (Putney Swope in particular springs to mind)

Same for books. I have a short amount of time each night (Min: 0, Max: 20, Mean: 3, Unit: minutes) to do any reading. I need to find really choice material that will hold my attention for possibly months on end. I'm still working on deciding what to put into this gap.

Mostly for the last year or two I've been reading a steady diet of Sci-Fi. But it's hard to find anything good enough when the diet is so limited. I'm now in the category of "If you only read 1 book this year... make sure it's...${this_book}"

Book suggestions welcome. (Spoilers about television shows are also permitted.)


The Joy of Making Simple Edits to Microsoft's Docs

When working on the previous blog post, I spent some time at microsoft's documentation site, where I noticed a small error: two similar articles had their titles swapped with each other. As in: Foo had Bar's title and Bar had Foo's title.

tiny bug

Usually you see an error like this and just move on.

edit button

But the edit button caught my eye, and I decided to try fixing the mistake. I was pretty apprehensive because these kind of things so often lead to yak shaving, and my current schedule does not permit the shaving of yaks.

So I clicked 'edit'. This took me to a github repo containing the source file, written in markdown. Hey, I write everything in markdown these days, so this looked promising. I backed up and soon found a page describing how to contribute.

contribute link

Reading the description there, a tiny bug fix like this doesn't require too much work.

I "forked" the repository:

Fork the repository

...and then cloned the fork of the repository onto my local machine, with:

git clone

UPDATE: There is a much simpler way to do this, that avoids the need to press 'fork', or to clone onto your local machine. It is detailed at the end of this article in a P.S.

I found the files that need to be corrected, made the correction, and then committed the changes locally:

git add *
git commit . -m "title of and iis-with-msdeploy were back-to-front."

Double checked that the changes I'd made were the right ones, by inspecting them again...

diffs ok

git diff --cached

Then pushed the changes back to github, so that my online forked version of the repository was correct.

git push

At github, in my fork of the repository, I clicked on 'New pull request' which led me to a page where I could see how my fork differed from the original repo. I reviewed the changes once more, then clicked the 'Create pull request' button. I wrote a brief description of my change and then sent it into The Ether by clicking the first green button I saw.

create pull request

Off it went. Wheeeeee! My pedantry, embodied in digital form, zipping around the globe.

Back in the original repository I saw a fleet of bots spring to life, checking and analyzing everything that was happening. One of the bots noticed I was a first time contributor, and asked me to digitally sign a contribution license agreement. It even gave this helpful guarantee "I promise there's no faxing" which was a nice touch.

I was then whisked expertly through a DocuSign process where I gave a website enough details about me that they could prepare the paperwork for this contribution license. It arrived in my inbox a minute later. I signed what I needed to sign, digitally, and was amazed that some team of people, somewhere in the world, had designed this well enough for legal hoops to be jumped in mere moments. This sort of baloney would've eaten up a week, easily, not too many years ago.

While I was busy doing the signing, the bots were hard at work verifying everything else, and I noticed from other emails that the bots were now happy, and had decided to forward my pull request onto their human masters.

I felt oddly nervous at this point. Battle scars from years of forum administrators shouting that a question has just been asked in the wrong forum, or from bureaucrats rejecting work because form 27B is not filled out in triplicate... a knot formed in my stomach as I prepared for a tiny little rejection, of this tiny little contrib.

Shortly thereafter there was a kind response from the original author, who was happy with the contribution, and the deal was done.

It was a good feeling, good enough that I felt inspired to write about it, at a time when I do not have time for writing about things.

You know me: I usually just take pot-shots at microsoft from afar, and have plenty of fun doing it. But now, this increasingly open microsoft lets me make positive contributions instead. And having done it once, it will be easier in future.

I hope that this extremely detailed walkthrough will inspire you to consider making a tiny contribution to an open source project for yourself.

(As usual, let me know the bits I did in a silly way, so others can learn.)

(Threw this together, including historically accurate screenshots, in 15 minutes, thanks to TimeSnapper)

P.S. There's also a simpler way that avoids the need to clone onto your local drive...




6 different ways to run an core web application

<blink> Gratuitous self promotion: Joseph Cooney and I will be talking about running core on Linux, at the upcoming DDD Brisbane conference at 4:05 pm, 3rd of December, less than 3 weeks from now. </blink>

Now that you've suffered through the advertisement, here's some content.

PLEASE tell me if I say anything misleading in what follows... if I'm going to stand in front of people and pretend to be worth listening to, I want some rigorous vetting to occur first.

Tell me leon, what are all the ways you can run an core web site?

Well I don't know all the ways, but I do know 6 different ways!

Get your head around this lot (even if it requires extra background reading) and you'll understand a lot about how core sites work.

  1. Visual Studio F5

If you're developing an core website in Visual Studio, then you might run it by pressing F5, for debugging purposes. But that's not the only show in town...

  1. Commandline "dotnet run"

Your website is really a dotnet console app, that self-hosts a website using a tiny webserver called Kestrel. (There's a lot to unpack in that sentence, but just let it wash over you for now)

You can run it, from the console, by calling dotnet run from the folder that contains the project.json file.

The output in the console will say something like:

Now listening on: http://localhost:2000

So if you then browse to http://localhost:2000, you'll see your website (and the console will show logging info about your visit)

  1. dotnet publish → cd bin{...}\publish → dotnet YourProject.dll

On your local machine, you can prepare the application for deployment by running "dotnet publish". This builds the application artifacts, does any minification and so forth.

If you don't specify where the published results go they will end up in YourProject\bin\debug\netcoreapp1.0\publish

If you go into that folder you can run the resulting artifacts by calling:

dotnet YourProject.dll

Note that you don't call "dotnet run YourProject.dll" -- leave out the run for this one!

So the commands in full (starting in the folder that contains the project.json file)

dotnet publish
cd bin\debug\netcoreapp1.0\publish
dotnet YourProject.dll
  1. IIS

You can host it in IIS. I've never done this and don't intend to. Me and IIS are parting ways for now. But it can be run by IIS. More info here: Publishing to IIS and here: Publishing to IIS with Web Deploy using Visual Studio.

  1. Running on Linux, from the console.... "dotnet YourProject.dll"

You can grab the artifacts from your local computer's "publish" folder (created in step 3), and copy them onto a Linux machine (using a technique such as SSH, scp, sftp). Then you can run it in the console, exactly the same as step 3:

dotnet YourProject.dll

(This assumes that you have have .net core installed on that linux machine already, instructions here.)

From a different console attached to the same machine, you can view the website by running, for example:

curl http://localhost:2000

...which isn't the most comfortable way to surf the internet. But since our webapp isn't accessible from the open internet, it's about the best you can do at that point.

Also, as soon as that first console window is closed, the application will stop. So this is not your final production technique. For that....

  1. Running properly on linux, with supervisor + nginx

In Linux you can configure supervisor to run your application (and keep it running). This is analogous to the work that Application Pools do in Windows land.

And nginx is a popular webserver, analogous to using IIS on Windows. The two work together to run your application and deliver webrequests to it. You set up nginx to receive requests from the internet and pass them on to your application (i.e. to "proxy them" through to your application, also know as acting as a 'reverse-proxy')

Details about using supervisor, at

To learn how to configure nginx to proxy requests through to your application, try the article here:

With those in place, you can browse to your site from the internet (assuming you purchased a domain and configured it to point to the webserver, or perhaps you are browsing by IP address, like all hardcore nerds.)

Okay, that's 6 different ways to run your core web app.

(You can swap nginx for some other webserver like Apache, but I'm not counting that as a separate method, just a variation on number 6.)

(And you can user systemd and upstart instead of supervisor: notes here.)

What did I get wrong!?

Update: Some answers to this question have come in already...

I wrote Katana instead of Kestrel -- fixed.

You can of-course also host a .net core app inside an MS Word Macro.

I left out Azure. You can deploy .net core apps to Azure, and if that's something you're interested in doing, I think this article covers it nicely: Deploy ASP.NET Core 1.0 apps to Azure web apps.

Just kidding about the Word Macro.

Further reading

This document brings together documents on each deployment method: core: Publishing and Deployment.


Have you ever seen the International Space Station?

The International Space Station is a gigantic space base, the largest artificial object in space, and there are people on board! Real actual people, all the time. Ever since the year 2000 it has been continuously occupied, with anywhere from 3 to 10 people.

It zips around the earth, 350 kilometers above the ground. It moves so quickly (about 30,000 kilometers per hour) that there is no gravity (well, only a tiny bit of gravity) and it completely orbits the Earth 15 times a day (once every 92 minutes... in the time it takes you to watch a movie, that station goes right around the Earth!)

And one of the best things about it is that when the conditions are right, you can see it very clearly from Earth, without using a telescope or binoculars. The best time to see it is just after sunset, when the sky is getting dark but the ISS is up high enough to still catch the sunlight. It can appear far brighter than any other star. It travels from one side of the sky to the other quite quickly, in about 5 minutes.

My 6 year old daughter told me last night, "When I grow up I'm going to be an artist or a scientist, and if I'm a scientist I might get to go up on the ISS." (My ten year old said "Oh suuuure" because she's currently studying for her masters in sarcasm, which some very talented kids receive by age 13).


If you want to see the ISS for yourself, these are the steps:

  1. Sign up at Spot the Station — and NASA will email you every time viewing conditions are favorable from your location.

  2. Get a smartphone App such as SkyView (SkyView on Apple Appstore, SkyView on Google Play store) which helps you locate astronomical bodies.

  3. When you receive an email from Spot the Station create a reminder to tell you when to go check the sky. For example you might receive an email at 08:00 telling you the ISS will be visible at 19:00 that night. So immediately set a reminder for 18:55, to make sure you don't forget.

  4. At the appropriate time, head outside, and use your app (such as SkyView) to locate the ISS.

  5. Be amazed and filled with awe.

By the way, even if you don't do any of the other steps, you should check out SkyView (SkyView on Apple Appstore, SkyView on Google Play store). It's very handy being able to locate planets, stars etc, and you learn a lot.

And here's an example of the email you get from NASA. It's so succinct that if you don't remember signing up you won't realize what it's talking about:

26 Oct (1 day ago)
to: me 
Time: Wed Oct 26 7:15 PM, Visible: 6 min, Max Height: 88°, Appears: 11° above NW, Disappears: 15° above SE

Teach kids (and adults) to master algebra with DragonBox

Lately I've been playing with an app called DragonBox. This game has expanded my mind.

It presents itself as a simple game, with a level structure and game play reminiscent of Angry Birds. The first few puzzles are very easy. Slowly, as the game progresses, new rules and abilities are introduced. And each new rule or ability is oddly specific, maintaining a pleasing kind of symmetry. Every time you solve a puzzle, or master a new ability, there is a satisfying feeling of victory.

Slowly, over many levels the complexity of the rules are increased, but always in a fun and engaging way. And slowly the style and appearance of the elements in the game transform from boxes, critters and dice, until eventually, ever so gradually, they become letters and numbers, and you see that what you have been manipulating all along are equations! Beautiful wonderful equations!

dragonbox: teaches you to manipulate equations

I found this app because I was playing with Tangle. What is Tangle? Tangle is a tool from Bret Victor for creating 'Explorable Explanations', and somewhere in my reading, someone mentioned DragonBox.

I was playing with Tangle, in order to create this minimum price calculator, as part of that damn book I'm still damn well writing.


Building a minimum price calculator was a lot of fun. I finally got all of the ideas about costs clear in my head, ten years after writing my actual first product. Costs. Boring, but crucial. I've done what I can to make them fun. Have a play.

Or if you would rather learn algebra than build a product, play with dragonBox instead.