Industry Science

[ meta  foss  prose  methodology  rant  ]

See the great thing about science and technology jobs is our transparency and focus on the problem, not on the finance. You could say that the tech sector is very scientific and rational. You see, we rationalize taking advantage of liberally licensed software.

We value open source software because... it means we didn't have to write it. We express our value by not reciprocating and leveraging outside help/software to shrink our R&D budget so operations doesn't yell at us for making them money. After taking all the credit of saving money and time from our use of free and open source software (FOSS), we're not given those savings back in order to augment, publish, and reciprocate. Don't get us wrong... we love transparency. The kind where other organizations are transparent to us but we don't have to invest in the principles.

The only problem with FOSS is that sometimes the UI is ugly or has long documentation, so we do the smart thing and pay other companies to dumb down the interface or calculation for the users. See we're scientists and are really smart. Too smart to waste time with rtfm or articulate what features we must have or why the FOSS stuff didn't meet all of our criteria. We shop around until something feels right rather than use objective metrics about application usability. If FOSS was so good, it wouldn't be free right? It just feels safer when we can pay a company to tell us what the answers are.

We're also very scientific too. We experiment with decreasing the number of replicates we use during screening. It's actually very methodical. We make sure to use concepts that make sense to our operations overlords like 'budget', 'amortized', and 'asset' to describe what it is we do for them. And sometimes we give them back some of the money from R&D whenever we can. God knows they can invent drugs faster than we can, so we should let them claim the credit for tightening the belt.

See the thing about screening is that you don't need all the data you generate, right? Only 1% of the ideas work. So logically if you reduce the number of replicates by 50%, you take away most of the expense from stuff you didn't care about anyway, the 99%. The 99% is all junk that doesn't provide useful information or serve a purpose of controls, so we don't need to be that confident about the majority (99%) of our measurements. Or the 1% either. In the end, we'll just keep screening like robots until we find one observation above our magic number cutoff and then it either works or doesn't. It's very boolean so we don't worry about optimizing metrics like cost per true positive or the reproducibility of the enrichment screen. We serve at the pleasure of our investors, not the patients.