The other half of "Artists Ship"
paulgraham.com
The other half of "Artists Ship"
1–10 of 132 posts
Re: The other half of "Artists Ship"
#2I guess so many checks in a big company could be considered the biggest mistake of all?
Re: The other half of "Artists Ship"
#3pg worked in fast food?
Re: The other half of "Artists Ship"
#4One problem is that the cost of checks is only visible in the aggregate. The marginal cost of each new check seems to be pretty low.
I wonder if there's a way to limit the total cost without hard-and-fast silly rules.
Re: The other half of "Artists Ship"
#5One problem is that the cost of checks is only visible in the aggregate. The marginal cost of each new check seems to be pretty low. I wonder if there's a way to limit the total cost without hard-and-fast silly rules.
Is "one in, one out" a silly rule?
e.g. you can only add a new step to a process if a step is removed from another.
Re: The other half of "Artists Ship"
#6pg worked in fast food?
it was pre-artix ;-)
Re: The other half of "Artists Ship"
#7I guess so many checks in a big company could be considered the biggest mistake of all?
Well, as companies get larger they can afford less and less to make mistakes. If I'm running a start-up company, I can get away with changing things abruptly, because I have a smaller base of users and I have fewer people relying on my running at production-level. Also, chances are I'll be able to look at user feedback effectively.
Once you get large, and your company is providing sustenance for all your employees, people rely on your product, and you've got too many users for effective feedback-checking, you have to close up, take fewer risks. Because suddenly, people want you to move slowly. They don't want you constantly skyrocketing ahead with their playing backup. Look at any big company - even Google, which was once famous for moving quickly - and you'll see that part of what gives a big company a good reputation is its being "solid." They have to give things up for an advantage.
It's why newspapers are so relied-upon. Of course, now it's what is hurting newspapers the most. They're being beaten by the flexible Internet. But even there, we're seeing a trade-off. Look at the quality of stories by the top writers online and by the top NY Times writers, and the online writers are much more amateur. They're faster, occasionally they're more interesting, but the Internet is thus far not retaining a high level of professionalism among reporting. Similarly, start-ups are much less reliable on the whole than large companies - look at Twitter and its problems, for instance.
So I think that PG's article is right. You can't restrict people and expect them to do as well. However, too much freedom leads to less stability, so it becomes a trade-off. Everything in moderation.
Re: The other half of "Artists Ship"
#8This past summer I worked as an intern for a big company. On the first day of orientation the head of the legal compliance department said very flatly to us that any individual could do far more harm to the firm than good.
This mentality, reenforced by the company's bureaucratic change management system, really did not sit well with me.
Fortunately for my summer experience, my "buddy"/summer mentor and I found a loophole which we used to "hack" the change management system so that once we got our initial approval, we could propagate changes to prod without running through the whole process again for each change (which otherwise would have been required).
Re: The other half of "Artists Ship"
#9One problem is that the cost of checks is only visible in the aggregate. The marginal cost of each new check seems to be pretty low. I wonder if there's a way to limit the total cost without hard-and-fast silly rules.
Is "one in, one out" a silly rule? e.g. you can only add a new step to a process if a step is removed from another.
Is "one in, one out" a silly rule?
Yes. ;)
When do you institute the rule -- on day one, in which case it is functionally equivalent to "do not have any checks"? On day 13? Day 7386?
What is the quantum of a "check"? Is Sarbanes-Oxley one "check" or a collection of hundreds of "checks"?
Does the limit apply department-by-department or company-wide? Your sysadmins will naturally tend to employ a lot more formal checks (most of which are hopefully enforced by tiny Perl scripts) than your R&D prototyping team. Do you force the sysadmins to compete with the accounting department for a quota of checks? Do you incentivize sysadmins to evade checks by getting tasks reassigned to the R&D prototyping team, which (alas) can't afford to have any checks?
The last thing you want to do is encourage teams of lawyerly meta-checkers to run around enforcing rules about rules. That's costly, squared.
Re: The other half of "Artists Ship"
#10I guess so many checks in a big company could be considered the biggest mistake of all?
Well, as companies get larger they can afford less and less to make mistakes. If I'm running a start-up company, I can get away with changing things abruptly, because I have a smaller base of users and I have fewer people relying on my running at production-level. Also, chances are I'll be able to look at user feedback effectively. Once you get large, and your company is providing sustenance for all your employees, p…
This thinking has always kind of confused me. Why are customers/users at a big company more important than those at a startup? Just because there's more of them, now you can't make mistakes?
If you have the agility to make rapid production changes, you also have the ability to rapidly rollback. So the argument that larger companies require more checks and testing than startups isn't really valid, especially when you consider the costs.