(click for more)
Comments (0)

A few years ago I started specing out a project to build my own customizable video camera rig. After writing up some code to simulate various sensor types I decided to get my hands dirty with this project, which was a great success. In the following year or so I got busy with a new job and didn't have the same resources as a certain eyewear mogul and didn't follow up. While it was a fun project for me to work on, in the back of my mind I was thinking about the possibilities of commercializing my idea. Since then, RED was launched and is a smashing success.

Maybe I should dust off the old calculator project and get cracking before Dov Charney decides that it is a hot opportunity.

Comments (0)

Los Angeles was the Australia of the art world.

Comments (0)

The web is full of bullet-point lists of handy tips for job applicants. Some of them are OK, some are crap. I'm not sure which direction my post is going to go, but it's certainly what I am looking for in applicants.

1. Read the job ad. If the ad asks you for a particular piece of information, give it. If you don't you will end up looking like one of the RAND_MAX people who apply for everything that appears on Craigslist.
2. Write a cover letter. Especially if the ad asks for good communication skills. Don't send your cover letter as an attachment, just treat the email body as your cover letter. If you aren't feeling particularly creative a standard 3 paragraph letter works well:
   1. Who are you / what attracts you to the job
   2. What skills do you have? A small anecdote, e.g., "In my last job I used my awesome C# skills to fix the blah blah blah system..."
   3. Request to be contacted
3. You must custom write the cover letter to the job you are applying to. You want to spend the next few years working with me, then spend the next few minutes putting some effort in to getting the job.
4. This probably doesn't apply if you are dealing with some company with a dedicated team of HR drones who stick your resume in a database, but I'm not a drone.. Therefore, send your resume as a PDF or as plain text. I like people who pay attention to detail and MS Word resumes always look crap on my computer because it is not your computer. All of your funny spelling and odd grammar shows up with squiggly lines. A PDF lets you format the resume exactly how you want me to see it. And you get bonus points if your resume was generated in LaTeX.
5. Keep your resume short. I don't care what you were doing 10 jobs ago, especially if it is pretty much identical to what you are doing now. I don't care about the details of the various internal systems you have worked on. Give me your last 2 or 3 jobs, and a short summary of previous work if you want.
6. Spell correctly and use a close approximation to correct grammar. And don't misspell a computer related term. Have someone read your resume before you send it out.
7. God save me from those who include a list of 100+ technologies they know. If you are truly proficient in 35 languages, then say "Proficient in 35 languages, ranging from Ada to Fortran." And unless the job is specifically asking for expert knowledge in some dead technology like dBaseII, don't bother listing it. I don't care.
8. This probably means that you should tweak your resume to each job you are applying to.
9. Please include a little bit on your resume showing something about who you are rather than simply what you do. I'd like to know if you ballroom dance, play the cello or play Go.
10. Most resume tips warn you against anything that might result in your application being ignored. I don't ignore applications - while I may not read them all completely, I do want to find the right employee so everything gets at least a glance. And if the application is egregiously bad I'll certainly pay attention; the laughs alone are worth it.

OK. That's 10 and while I could go on, I feel better already.
Comments (4)

I've tweaked my blog 'system' a little bit this afternoon. Instead of being a pile of bash scripts, now it is a pile of bash and perl scripts. Unfortunately, all my blog feeds will be reposted, but on the plus side I now have a copy of all my old posts from various other platorms. Oh, and RSS feeds should now work again. They have been broken since the last time I messed around with the system. I spent a little bit of time writing some javascript to scrape my old blog posts from inside Google Reader, which had the only copy of my old posts from wordpress and blogger.com. There is probably an easier way to extract the posts, but I can always do with some practice on the reverse engineering front.

Ok. That's enough housekeeping. Back to my regular blogging schedule

Comments (5)

It has to be lower than McCains, which as far as I know is infinite.

Comments (0)

Let me just reiterate how much I like Amazon's recommendation service. Collaborative filtering at it's finest, if you ask me. Collaborative filtering is the process of using what a system knows about others to make recommendations for an individual. If most people who like A and B also like C, then if you like A and B, you will probably like C. It is a simple concept and it works well in many applications. Netflix implements a similar system, which also looks at people's ratings of the movies that they have rented. If it turns out that I like A and B, but really hated C, the algorithm will find other people like me and perhaps recommend E instead of D.

Most of this is widely known. And then the human interest stories started popping up, the most famous of which covering the 'My Tivo thinks I'm gay' phenomenon:

Mr. Iwanyk, 32 years old, first suspected that his TiVo thought he was gay, since it inexplicably kept recording programs with gay themes. A film studio executive in Los Angeles and the self-described "straightest guy on earth," he tried to tame TiVo's gay fixation by recording war movies and other "guy stuff."

"The problem was, I overcompensated," he says. "It started giving me documentaries on Joseph Goebbels and Adolf Eichmann. It stopped thinking I was gay and decided I was a crazy guy reminiscing about the Third Reich."

While these stories are cute, they highlight the pitfalls of making recommendations that just don't fit. I've been noticing more news stories and blogs including related links at the bottom of their posts. They do this to increase the stickiness of their site, by redirecting traffic back to the original web site. It is a nice idea, but with the prevalence of good recommendation technologies, the slightest hint of bad is enough to train users to ignore such related links sections.

Computer generated related links tend to suck because, unlike collaborative filtering, the technology tries to use high-power semantic analysis to understand the text and come up with related articles. And text analysis technology just isn't that good right now. At my firm we use some fairly run of the mill text processing to identify similar companies based on SEC filing text. This tends to work well as we have a large corpus of highly idiosyncratic text to work from. Short blog posts wouldn't work well with simple techniques, so more complex techniques are used. And they suck. Even worse, none of these technologies attempt to use any user feedback to help train the algorithms. I don't want to see explicit, click-here-if-you-think-this-story-is-related buttons, but at least track which links are clicked on to help optimize your algorithms.

My gut feel is that for the developers to prove the value of their technology they are compelled to include related links in as many stories as possible. This is a wrong-headed move. They should only include links when relevance is well established; otherwise users (like myself) will quickly learn to ignore that junk text that appears at the bottom of blog posts and articles. It took me a few days to subconsciously ignore the RSS feed 'flair' at the bottom of most posts. And now that my ignoring neurons are well trained, it took me all of nano-seconds to realize that 'Related Links' are junk.

Related Links:

• Price of tea in china
• Northenmost point in Europe
• List of Star Trek episodes (Polish)

Email this   Digg this   Stumble it!   Share on Facebook   Share on MySpace   Link to me on LinkedIn  

Comments (0)

Comments (0)

Comments (0)