4 ms·

This is an awful analogy because silicon valley has a huge failure rate for companies. Even an established company can fail quite rapidly. Imagine launching the

by anonjon 17y ago

This is an awful analogy because silicon valley has a huge failure rate for companies. Even an established company can fail quite rapidly. Imagine launching the shuttle and 90% of the time it blows up spectacularly. That would not do anything good for the space program.

stcredzero 17y ago
That's not the author's analogy, only a straw man resulting from your misplaced referent. A shuttle launch is not analogous to a company launch. A shuttle launch is analogous to a version or a feature deployment of one particular company. A company that iterates a lot, and deploys a lot, gets good at iterating and deploying. An organization that preps, launches, and retrieves spaceships a lot gets good at that too. More effort on proving the null hypothesis is suggested.
- idlewords 17y ago
  The parent's point remains completely valid. Rapid iteration only works if you are allowed to fail. In the case of manned space flight, failure means you lose the crew.
  - stcredzero 17y ago
    What straw-man organization is unable to send up unmanned prototypes? Oh, yeah. Must be NASA. http://www-pao.ksc.nasa.gov/history/mercury/mercury-unmanned.htm http://www-pao.ksc.nasa.gov/history/mercury/mercury-unmanned...
    - idlewords 17y ago
      If you're sending up unmanned prototypes, why go on to man the missions at all? Save yourself the cost of all that life support plumbing and call it a brilliant hack.
      - stcredzero 17y ago
        Fine, your side can explore space and establish enterprises on Mars with remote-controlled machines. My side will have all of that technology, plus immediate human presence. We can see who innovates faster.
- anonjon 17y ago
  The way I read the article was that we would cut NASA up into a number of rapidly iterating teams, each analogous to a small company. If you take this model and comparison to heart, and realize that there are actually /two/ levels of 'looping' that make silicon valley effective. One is competitive attrition, the other is rapid iteration. The attrition factor should be obvious, you have a large number of companies vying for success, and only the top few are even moderately successful. There would have to be a large enough margin that it doesn't cost NASA much to fail horribly on quite a few of its projects. Iteration is good because if you iterate rapidly enough, you are allowed a larger margin of error, because any errors in your previous version will be fixed relatively quickly in the next version. But here's the unfortunately reality: Sending anything into space is horrendously expensive. It is not like programming where releasing and distributing a patch is a tiny part of the overall cost of the program. When you screw up with a space ship, bad things happen (shit falls out of the sky, there is pollution to clean up), it costs money into the millions, and you have to fill out a TON of paperwork. Its not like a major error in a rocket you are launching can just be 'pulled from the website' after the rocket has exploded. Even if you /do/ go with cheap unmanned probes, you still have the massive pr disincentive of potentially failing every day for a month, and losing tons of money. NASA would become a joke.
- eoyola 17y ago
  I don't think it's a straw man. Just maybe the closest to what you're thinking is a slippery slope, but it didn't seem like the parent was trying to refute the author's argument. The author does deal with the concern: "I recognize that NASA cannot push a system to launch more frequently than it is capable of, because this could mean overrunning the budget or, worse, cutting corners on safety. Instead, future systems should be designed so that they can be rapidly prepared for launching by small teams. This would not only increase NASA’s ability to send up innovative payloads but also make launching systems more reliable. After all, the more a rocket is flown, the better it can be understood and the safer it becomes. Frequent launchings would also reduce costs per flight in the long run."

by anonjon 17y ago

stcredzero 17y ago

That's not the author's analogy, only a straw man resulting from your misplaced referent. A shuttle launch is not analogous to a company launch. A shuttle launch is analogous to a version or a feature deployment of one particular company. A company that iterates a lot, and deploys a lot, gets good at iterating and deploying. An organization that preps, launches, and retrieves spaceships a lot gets good at that too. More effort on proving the null hypothesis is suggested.

idlewords 17y ago
The parent's point remains completely valid. Rapid iteration only works if you are allowed to fail. In the case of manned space flight, failure means you lose the crew.
- stcredzero 17y ago
  What straw-man organization is unable to send up unmanned prototypes? Oh, yeah. Must be NASA. http://www-pao.ksc.nasa.gov/history/mercury/mercury-unmanned.htm http://www-pao.ksc.nasa.gov/history/mercury/mercury-unmanned...
  - idlewords 17y ago
    If you're sending up unmanned prototypes, why go on to man the missions at all? Save yourself the cost of all that life support plumbing and call it a brilliant hack.
    - stcredzero 17y ago
      Fine, your side can explore space and establish enterprises on Mars with remote-controlled machines. My side will have all of that technology, plus immediate human presence. We can see who innovates faster.
anonjon 17y ago
The way I read the article was that we would cut NASA up into a number of rapidly iterating teams, each analogous to a small company. If you take this model and comparison to heart, and realize that there are actually /two/ levels of 'looping' that make silicon valley effective. One is competitive attrition, the other is rapid iteration. The attrition factor should be obvious, you have a large number of companies vying for success, and only the top few are even moderately successful. There would have to be a large enough margin that it doesn't cost NASA much to fail horribly on quite a few of its projects. Iteration is good because if you iterate rapidly enough, you are allowed a larger margin of error, because any errors in your previous version will be fixed relatively quickly in the next version. But here's the unfortunately reality: Sending anything into space is horrendously expensive. It is not like programming where releasing and distributing a patch is a tiny part of the overall cost of the program. When you screw up with a space ship, bad things happen (shit falls out of the sky, there is pollution to clean up), it costs money into the millions, and you have to fill out a TON of paperwork. Its not like a major error in a rocket you are launching can just be 'pulled from the website' after the rocket has exploded. Even if you /do/ go with cheap unmanned probes, you still have the massive pr disincentive of potentially failing every day for a month, and losing tons of money. NASA would become a joke.
eoyola 17y ago
I don't think it's a straw man. Just maybe the closest to what you're thinking is a slippery slope, but it didn't seem like the parent was trying to refute the author's argument. The author does deal with the concern: "I recognize that NASA cannot push a system to launch more frequently than it is capable of, because this could mean overrunning the budget or, worse, cutting corners on safety. Instead, future systems should be designed so that they can be rapidly prepared for launching by small teams. This would not only increase NASA’s ability to send up innovative payloads but also make launching systems more reliable. After all, the more a rocket is flown, the better it can be understood and the safer it becomes. Frequent launchings would also reduce costs per flight in the long run."