Comment by pnathan | Hacker News Reader

Comment by pnathan | Hacker News Reader

Parent Full thread

There is a fairly deep difference between a software startup and a 'hardware' startup. The fundamental difference - IMO - is the 'get off the ground' capital.

Non-labor initial outlay for some hackers to build a software system is well under 100K.

To put together a chemistry lab takes well over 100K (or so I have been informed). Imagine having to spend 100K just to have a spot where people can work.

Those costs of course start to spiral heavily upwards as you get more and more high-tech in your hardware. Then, if you're setting up startups for the hardware world, what's the ExpectedValue of your ROI. Batting average ain't going to be great. So the incentives are skew compared to the software world.

That said, I'd love to be involved in the software side of nuclear energy. Among other things, you can't rely on your memory as a given, since it is more prone to corruption.

Boulderdash. It completely depends on the kind of software system you are building. Do you need formal methods? Do you need complicated mathematical tests? Do you perhaps need to acquire patents or 3rd party software? Extensive domain analysis can take months, even years. Complicated systems can take years to design, implement and deploy.

It all depends on what you are making. So far, the startup world seems to be in a goldrush with devestating results (both for investors, as well as participants). The article is right: why are these developers not making better simulation software? Why are they not creating the next iteration of mission critical control systems? Two words: gold rush.

So, if you talk about 'a software system', you should specify the bounds. Failing to do this, I have no choice but to diagnose you with yet another case of Silicon Valley gold fever.

Who, me? I don't even live within a day's hard drive of California.

you can't rely on your memory as a given

Oh, it gets even more exciting than that. You can't always even rely on your computations, let alone your memory!

http://en.wikipedia.org/wiki/Radiation_hardening#Radiation-h...

I think most computers in nuclear energy applications are going to be installed well away from places with significant radiation fields. Most things that are close to the reactor are going to have a multiplexing or demultiplexing role and can probably be better implement with programmable logic controllers.

Reply on news.ycombinator.com