Arvind Narayanan, a computer scientist, wrote on this LiveJournal about data, privacy, web traffic quality, caffeine and startup topics.
Commenters link it for a power-law-style data fit that kept working on new datasets, and for his analysis of low-quality StumbleUpon traffic.
If this sort of thing is fascinating to you, and you consume caffeine, you absolutely must spend the short time to learn about antagonistic and reinforcing caffeine usage. I've read a few papers on it, but I'm not finding any of the best ones by googling. This link has a short overview though:
http://arvindn.livejournal.com/57651.html
> What is common to these tasks is that progress happens in spurts, due to the fact that they involve frequent cognitive bottlenecks. A cognitive bottleneck can only be overcome when attention level exceeds a task dependent, typically very high threshold. Clearly, then, antagonistic caffeine consumption results in worse-than-normal productivity, because it flattens the attention level curve and decreases the fraction of time spent at peak attention level. Instead, reinforcing consumption helps maximize productivity (the red line). According to this strategy, the best time to drink coffee is when you are already very alert.
Basically, caffeine gives you a boost, then withdrawal. If you're doing creative work, time your caffeine consumption to your high energy periods, then rest during the withdrawal. If you're doing rote, repetitious work, take caffeine during your low periods to smooth your energy out.
This is incredible for your productivity, especially if you're doing creative work that you can't brute force yourself to finish. Highly recommended, almost should be mandatory for any anyone who consumes caffeine and does creative work.
Honestly, this data should be free, and available to researchers who work on social tagging.
You should contribute your bookmarks to a public data set. http://bulk.archive.org has offered to host such data sets.
Someone should build a webservice that allows you to easily share your delicious tags. If no one steps up, you can simply email me your export (turian at gmail), and I'll package up whatever I receive.
Here is a delicious data set, for those who are interested:
http://arvindn.livejournal.com/116137.html
I have some other social tagging datasets in my file. Let me know if you'd like me to dig them up.
An interesting and very different take on caffeine consumption strategy was posted here a while ago:
http://arvindn.livejournal.com/57651.html
The idea is that you should use caffeine when you're at your peak, because you want to maximize your productivity for doing really hard stuff that you normally would only accomplish when you're having a creative spurt anyway.
But what sort of traffic?
See, e.g., Arvind's comments on StumbleUpon visitors:
http://arvindn.livejournal.com/133249.html
To quote the takeaway, "most of the traffic generated by StumbleUpon users to any given site is going to be low quality because the dopamine junkies make 100x more clicks".
Man, this never gets old. Every time I see it work on a different dataset, it's as surprising as the first time I saw it. Recently I plotted supermodel incomes. It was an amazingly good fit. http://arvindn.livejournal.com/98510.html
Via coffee. Scratch that; via well-planned caffeine consumption: http://arvindn.livejournal.com/57651.html