4 ms·

if computers taught me anything.. it's if you can do 1 thing really really well perfectly, it's not a ton of work to do it 10x or 100x in parallel. We don't ha

if computers taught me anything.. it's if you can do 1 thing really really well perfectly, it's not a ton of work to do it 10x or 100x in parallel. We don't have enough data to know reliability of say Falcon long term. Is it 99% successful? 99.9%? 99.999999%? Doing a ton of launches, you really want to work on the 9s, so 0-1 of your launches blow up, not 100+. But once you do, you just copy paste it baby and go to town.

vanattab 5y ago
I am not convinced the scalability of cyberspace applies to actual space.
- AnimalMuppet 5y ago
  I don't think it's a good example either. But the scalability of mass production may apply, and that might be good enough...
snowwrestler 5y ago
There is a literally a book called “The Mythical Man Month” about the limits of parallelism in computer engineering. It’s arguably one of the fundamental texts of the discipline.
- brianwawok 5y ago
  Right. I didn't say we hire 100 developers to code the project 100x faster. I said let's build 100 parallel rocket pads with 100 parallel rockets. It will indeed launch rockets 100x faster. The term you are looking for is embarrassingly parallel.
  - snowwrestler 5y ago
    Only parts of the system are embarrassingly parallel. But you need the whole system to be available for every launch. Poorly parallel parts of the system include regulatory clearance and accident investigations. There is only so much clear air space available each day. And if a rocket blows up (which they often do), they’ll all sit on the ground until there is understanding as to why. And then potentially sit even longer if they each require work to correct or mitigate flaws that are found. Starships are still hand-built prototypes. And Musk has a history of underestimating the difficulty in scaling and automating physical manufacturing. It’s way too early to be confident this can embarrassingly parallel. Parallelism is hard in general. It’s hard because often constraints aren’t obvious until you’re actually doing it.
- micropresident 5y ago
  And that book has literally nothing to do with what is being discussed.

vanattab 5y ago

I am not convinced the scalability of cyberspace applies to actual space.

AnimalMuppet 5y ago
I don't think it's a good example either. But the scalability of mass production may apply, and that might be good enough...

snowwrestler 5y ago

There is a literally a book called “The Mythical Man Month” about the limits of parallelism in computer engineering. It’s arguably one of the fundamental texts of the discipline.

brianwawok 5y ago
Right. I didn't say we hire 100 developers to code the project 100x faster. I said let's build 100 parallel rocket pads with 100 parallel rockets. It will indeed launch rockets 100x faster. The term you are looking for is embarrassingly parallel.
- snowwrestler 5y ago
  Only parts of the system are embarrassingly parallel. But you need the whole system to be available for every launch. Poorly parallel parts of the system include regulatory clearance and accident investigations. There is only so much clear air space available each day. And if a rocket blows up (which they often do), they’ll all sit on the ground until there is understanding as to why. And then potentially sit even longer if they each require work to correct or mitigate flaws that are found. Starships are still hand-built prototypes. And Musk has a history of underestimating the difficulty in scaling and automating physical manufacturing. It’s way too early to be confident this can embarrassingly parallel. Parallelism is hard in general. It’s hard because often constraints aren’t obvious until you’re actually doing it.
micropresident 5y ago
And that book has literally nothing to do with what is being discussed.