Hacker News Reader

Comment by amelius | Hacker News Reader

At least Pied Piper has some real technology to bring to the table (a compression algorithm).

Most real SV startups are, it seems, unfortunately just building a business around a CRUD application (see AirBnB, Uber, to name a few, but there are many many more).

View on HN

  > Most real SV startups are, it seems, unfortunately just
  > building a business around a CRUD application (see
  > AirBnB, Uber, to name a few, but there are many many more).

I think you vastly underestimate the technology that goes into the parts of Uber we don't see -- the logistics of dispatching and monitoring a couple hundred thousand drivers (or so?) in real time.

Yes, so advanced that new versions by competitors that do the exact same thing seem to pop up weekly

Scale.

End of discussion. What works for 100 customers/drivers does not work for 100,000

In the grander scheme of things, that's scale and concurrency. Nothing about the code is particularly novel or difficult.

>In the grander scheme of things, that's scale and concurrency.

I actually laughed out loud. I'm just the Postgres team would love your insight into how they can implement master-master replication because "it's not novel or difficult."

Right, but those are problems solved by database technologies. Uber can follow pretty well-defined patterns at this point to scale up. Especially since they have a natural sharding point built into their product (each city).

Pretty much this. Like Uber's unique problem is master-master replication?

True, but there's something to be said for companies that execute well. Providing a smoother service or more intuitive/sleeker product can be the distinguishing factor that makes a company stand out, and you want cycles of that mixed in with cycles of innovation.

Not that I'm making an argument that Uber itself executes particularly well. I have no idea about that.

How is AirBnB or Uber just CRUD applications? Seems to me there's a lot more than that involved in launching a nationwide rental or taxi service with a two-sided market. You have to solve regulatory issues (or ignore them strategically), routing and pricing algorithms, user trust issues, etc.

These two businesses in particular highlight that it's a lot harder than you think, otherwise the market would be saturated with clones.

That's after it's huge. Uber 0.1 was probably just WhatsApp to a couple of drivers.

And early Google was 'just' a search engine, something that already existed in some forms, that looked totally unlike the Google of today. You are correct, but what are you trying to say? All companies are going to be 'lame' early on by definition.

No, Google 0.1 did something technically nontrivial. These other co's only got technically interesting after scale, which is generically true. Since you don't know whether the startup will succeed, you'd better enjoy the work, eg scraper (FB 0.1) and forum software (0.5).

There are any # of co's not lame early on: Google, Adobe, Zee Aero, Joby, Viv/Siri, Cruise, Magic Leap, DeepMind, Vicarious…

> Google 0.1 did something technically nontrivia

Indexing and rating web pages based on links back to pages. Basically a massive "outer join" on the web. Basically a software implementation of how academic papers are rated, but technically a massive data processing exercise - not just a good idea in theory but required a great deal of engineering skill to implement.

Worldwide.

You realize that Uber has to solve a huge multifactor optimization problem?

How so?

Spoilers follow.

One fairly recent episode involves pretty much throwing away all of the hard technology that the engineers want to work on in order to make an appliance that's easier for the sales team to sell. They came up with the idea after one of the engineers says something sarcastically like "yeah we could just ignore all the tech and make a dumb box."

cough Facebook cough

Facebook is not a CRUD application.

Facebook is a graph theory application.

Reply on news.ycombinator.com