Tuesday, April 30, 2013

How we failed our YC interview

We had our YC interview on Friday and were super confused after it. We sort of expected that. We talked to over a dozen YC founders and most of them said “You’ll leave confused”. However, it was an unexpected kind of confusion. We weren’t confused because we didn’t know what they thought about us. We were confused because the interview was nothing like what we prepared for. The preparation we needed couldn’t be done two weeks before a YC interview (well.. it shouldn’t). It should be done at the idea stage of your company.

As an entrepreneur, you are building at least one of:

  1. A solution (something that solves a problem)
  2. A something (something that you think should exist in this world)

A solution is a something. But a something isn’t always a solution.

When you start building “somethings”, you are stepping into dangerous territory. 

  1. It’s easy to confuse what you think should exist in this world and what should actually exist. 
  2. It’s easy to pretend there’s a problem you’re solving. 
  3. If it’s not a real problem, it makes it harder to explain and get empathy from whoever listens. The best you’ll get from them is “that’s cool”. 
  4. If anyone thinks deeply about your business, they’ll think you’re crazy. Especially because you can’t justify anything you’re building without some sort of traction.

It’s not bad to build somethings. Writely (Google Docs) was once a something. Sam Schillace built it to test the powers of AJAX.

If Sam wanted to tell you about his something, he would say:
“I’m building an online word processor.”
And you’d probably say “That’s cool".

However, if he wanted to tell you about his solution, he would say:
“I’m building an online word processor because we can’t collaborate on Word.”
Then you’d say, “Holy shit, tell me when it’s ready.”

So now, where did we go wrong? We never realized we were pitching a something when we actually had a solution.

We pitched a cloud Matlab. We never backed it up with a terrifying experience using Matlab (we actually like Matlab). We just said we want to build a cheap alternative because Matlab is inaccessible now (the license costs $2k). We pitched something that we thought should exist in the world. However, we couldn’t back up that it was actually needed and still can’t. The sooner you realize that you are building (or pitching) a something, the better. That’s not to stop you from working on your idea, but just to find quick and honest validation.

We know there are a lot of people who have trouble learning Matlab and they can benefit from our platform as a learning environment. However, when we pitched our “big vision”, we focused on replacing Matlab, without clearly defining the problems. We just sounded completely nuts. The only reason we got the interview was because they wanted to know what our secret sauce would be.

During our demo, I remember Paul Graham and Trevor Blackwell standing behind me and Trevor saying “Yep, that’s Matlab”. I saw the disappointment in Paul’s face as soon as they sat back down. He wanted to find out what made us different from Matlab. He wanted to know what problems we had with it that we were solving. He wanted the damn secret sauce. Most importantly, he wanted to invest in us. We just looked a little too on the delusional side for that to happen.

If you are reading this and building a product (YC interview or not), stop and think about whether you’re building a solution or a something. If it’s a solution then pitch it like one. If not, then get that validation asap.


  1. The thing is ... there's already a 'cloud Matlab' - several of them, in fact. Most prominent IMHO is Sage (http://www.sagemath.org/), but there are starting to be R-based solutions like RStudio Server and Shiny Server.

    1. This article is not about a cloud-based Matlab. It's incidental.

  2. http://www.templeos.org

    "MATLAB operating system" that is a sure thing because it's God's temple.

    1. I have seen you everywhere this week. Spooky.

  3. Nonetheless, quite interesting read! Good luck. Sometimes, when somethings are compelling, industry creates problems (that are big enough) so that using the somethings as solutions gets immediately justified. But I agree with your learning -- get your assumptions validated asap.

  4. Irrespective of whether this particular something/solution exists already, the point is well taken. Thanks for sharing.

  5. Thanks so much for sharing this. I really enjoyed reading about this experience, especially because this is exactly what I tried to communicate to my students this morning. I teach a first-year entrepreneurship course at a university in Chile, and the students' assignment was to analyze a problem, one that they will later build a solution for. But many of the students simply presented a solution, without truly analyzing the problem. I look forward to seeing what they come up with next.

    Best of luck to you!

  6. Thank you for your article.

    Incidentally, if you could make Octave look pretty and put it in the cloud, I'd pay for it.

  7. It's good I stumbled on this blog. I'm a newie in development, and I thing this piece will help a long way.

  8. We are actually building a "cloud Matlab" as well, except the vision is far broader than that. We chose Python and the SciPy stack as our starting point. http://wakari.io

    You should email me - pwang@continuum.io.

  9. I would buy a cloud based version of Matlab. Keep at it!

  10. Thank you for this excellent post! You bring forth a great distinction between a solution and a something in need of a problem. Thanks!