I don't try to find bugs in software; they just seem to find me. For example, RssDistiller, for all its sweetness, doesn't seem to like the URLs in http://www.cs.colorado.edu/serl/seworld/archive.html ; the RSS file forgets to drop the archive.html before tacking on database/####.html . (Message thread here.)

11:27:28 PM    comment []  

(Inspired by Powazek's post.)

http://www.apple.com/macosx/newversion/images/jaguar_ichatwindow.jpg

http://research.microsoft.com/scg/images/comicchat.gif

They both seem too cute... I just switched to Trillian because I can turn off all the formatting, add dates, and log everything.

9:19:42 PM    comment []  

I submitted a "bug" on the clustering software I'm using for my thesis. Unfortunately, it has to do with "round-off errors and FP arithmetic." "As a temp fix, I suggest you scale all the values so that the max is 1.0." Wee.

8:46:44 PM    comment []  

Slashdot comment worth reading. Basically, James Doohan was hospitalized and it made the front page of Slashdot. Robert X. Cringely's son is a victim of SIDS and he asks for help from the community. No mention on Slashdot.

7:37:48 PM    comment []  

If I burn a CD of MP3's, the artist doesn't make any money. If I buy a used CD, the artist doesn't make any [more] money. Right?

There's got to be a better solution... I don't know what it is, though.

7:26:11 PM    comment []  

I am so done with finals this semester. Hooray! Celebrate! And stuff! [Stef]

Me too. Except for that whole thesis debacle.

2:54:50 PM    comment []  

So, my thesis works with even more data: agglomerative with cosine distance and log scaling with 10 clusters. The point is that the "FAIL" executions are in their own clusters, separated from the "okay" executions.

It took like 3 hours to do all the data generation for 292 executions (compared to 1 hour with 171 executions) and like 2 hours to do all the cluster analysis (compared to about 45 minutes).

I have an AI final in about 7 hours; it's my last final ever. Sweet.

5:30:07 AM    comment []  

As if to prove how lame I really am, I decided to collect some statistics on #geeks activity. [Joe Bork]

Hot.

I had a markov-chain-based diary generator that worked pretty well; that might be interesting on the source data, too.

4:54:31 AM    comment []  

I wish Doritos would either put ingredient lists on their website or, through the miracles of modern science, figure out some way to make "ZESTY SOUR CREAM AND CHEDDAR" Doritos without "beef stock".

Oh well, the next person that gets something from the vending machine will have a $0.65 non-vegetarian surprise.

2:22:27 AM    comment []