Skip to main content

Jeffrey Byun on Startups

I’ve started listening to the Startup School Radio and one of my favorite episodes is this interview with Jeffrey Byun:


Lesson 1: Build things that initially don’t scale

(and don’t begin premature optimization):

They start the show talking about, an earlier startup that Jeffrey and his partner Henry were building:

(Starting at 7:20):

Aaron: So, how’d that go?

Jeffrey: Absolutely terribly!  We made every possible mistake that you could think of.  It took us 4 or 5 months to build a product and then when we were ready to ship Henry had a thought: “well what if we get 100,000 users after we launch we’d need to make sure that our servers can handle that.  So we spent another month making sure that we had the right capacity so that our servers wouldn’t explode.

Aaron: So here’s a good lesson for everyone: theres really no need to build for scale right at the start. You kind of have to build just to attract the first person.  We have a mantra [at Y Combinator] about building things that don’t scale, that includes your code for servers.


Lesson 2: First, get sales:

Fast forward to where Jeffrey begins to talk about the nucleus of OrderAhead, his current company that allows ordering food ahead of time – skipping the line.

(Starting at 18:00):

Aaron: How do you start convincing restaurant owners that this is a thing that they want.  Do they pay for it?

Jeffrey: <<edited for brevity>> …I just wrote up a contract in Microsoft Word and I basically drove to University Avenue downtown Palo Alto and spent the next 2 weeks walking up and down University for 8 hours a day trying to signup every local business.  It was a pretty interesting experience.

Aaron: How many of the restaurants did you sign up?

Jeffrey: About 30.

To put that into context, Jeffrey estimates he talked to 100 restaurants during that period, giving him a 30% close rate.  Keep in mind he wasn’t just signing restaurants up, they were committing to pay a $100 setup fee and pay OrderAhead a commission on every order that came through their system.

He did all of this without a single line of code, wireframes, or product.

His trick?  He cites persistence — saying he visited some restaurant owners 8 times throughout the 2 week period.


Read More

Goldman in ventureland

Goldman in Ventureland

Bloomberg ran a great article that summarized Goldman’s history in Venture Capital (you can read the article here: Goldman in Ventureland).

Read More

The Bank of Montreal’s Passwords are a bit weak

The Bank of Montreal should be ashamed of themselves by forcing users to have exactly 6 digits in their online banking password.  They just provided my mom with the following instructions while resetting her password:

Screen Shot 2015-10-13 at 12.15.18 AM

6 digit passwords are not even remotely safe.

Even Google recommends under their web tips that passwords are “long to help keep your information safe”:

Use a long password made up of numbers, letters and symbols

The longer your password is, the harder it is to guess. So make your password long to help keep your information safe. Adding numbers, symbols and mixed-case letters makes it harder for would-be snoops or others to guess or crack your password. Please don’t use ‘123456’ or ‘password,’ and avoid using publicly available information like your phone number in your passwords. It’s not very original, and it isn’t very safe!

Read More

Could a computer have saved my mom from 10+ years of chronic pain?

As someone who has worked in both technology and healthcare – in particular Radiology – for the majority of my professional life this video makes me stop and pause.



IBM’s announcement marries their super computer Watson with a large database of medical images that the computer could soon make predictive sense of:

…Big Blue announced a $1 billion acquisition of Merge Healthcare, and thanks to a marriage of Watson’s existing image analytics and cognitive capabilities and Merge’s data and imagery, the computer should soon be able to recognize many types of medical imagery. In essence, Watson will be able to “see,” IBM said…and eventually save a lot of M.D. eye strain.

Imagine the possibilities:

  • The ability to apply new learnings to previous data.  This one is huge for me personally – as a very long story made short; my mother had a terrible back injury for more than 10 years that her physicians said was never going to get better.  After extensive travel around North America to visit some of the marquee clinics and doctors she was told her condition would never resolve and she had all but lost hope.  Thankfully, through one of my dearest friends Dr. Deepak Kaura (a brilliant human being and physician in his own right), we were connected with the head of Radiology at a hospital in Toronto and shortly thereafter told she was a candidate for surgery.  We flew to Toronto in January of 2012 to have a relatively simple 60-minute out patient procedure performed on her spine, where a cyst was drained and injected with the equivalent of glue to prevent it from filling back up.  Put back into context: for 10 years she had lived with debilitating pain as a result of her condition that all but one Radiologist correctly identified and treated.  Imagine if Watson (or similar technology) was able to go back through a database of medical imagery and find other patients with similar scans and conditions, they too could benefit from such a procedure or treatment as our medical knowledge evolved.
  • Radiologists can spend more time focusing on the images that are of concern, and less on normal images.  This could create real economic advantages within health regions, allowing Radiologists to work smarter instead of harder.
  • The turn-around time of reports could be near-instantaneous.  In some circumstances the time between a Diagnostic Image being acquired (getting an x-ray) and the time that a Radiologist issues a report (looks at the image) could be days or weeks at the extreme, impacting the care and treatment a patient could receive.  A computer, however, would be able to process the data as soon as it was acquired, leading to faster diagnosis and improved timeliness of care.
  • And more!  Computers are far better at consuming massive amounts of data and finding the correlations and connections amongst disparate data points.  I’m confident many more benefits will come of predictive systems like this.

There are companies other than the IBM/Merge combination outlined in this article.  All I hope is that these types of systems will be in place to prevent a similar decade of pain and suffering for patients similar to my mom in the very near future.

You can read the fastcompany article here:

Read More