philgoetz

The site isn't down. It's censored. Look at the link address.

You don't RENT in that area. You have to buy. Then you can sell it later, and get back all that extra money you were paid for living in a place with high housing costs.

Google and Amazon have, instead of going remote, established branch offices all over the US, in every cool city where hip young non-Republicans live.

If you try to balance speakers racially, you'll run into the problem that we often deliberately discriminate against certain minorities (usually the Chinese) in order to raise the technical value of talks by making sure the speaker can speak English clearly. In nearly all colleges, teachers and TAs in college must prove proficiency in speaking English before they can be hired to teach a class. This deliberately discriminates against foreigners who don't speak English well enough, or whose accents are too thick, for the students to understand.

Discrimination is not always bad. Discrimination is the morally-neutral act of comparing things on some criteria. Additional context is required to determine whether it is a bad use of discrimination. We should not strive to be inclusive of people who speak the conference language poorly, or, say, people who are crazy. Kevin's blog expressed concern over how to be inclusive of the mentally ill when choosing speakers. Was he joking? I don't know anymore.

Given that this was whole-genome sequencing, is 1700 mutations higher than other tissues? The paper doesn't appear to say that.

If they were claiming that this were an unusually high mutation rate, I'd wonder why everyone doesn't die of brain cancer. If other tissue had N mutations per genome, and cancer typically takes 10 mutations to create, in the absence of compensatory mechanisms, the ratio of brain cancer to other cancers should be about (1700 choose 10) / 50 * (N choose 10). If N were, say, 100, that number would be so large that not one person would ever had developed any cancer other than brain cancer.

Are you sure you're not counting mitochondrial DNA mutations? A single neuron has about 1000 mitochondria, and would easily have tens of thousands of unique mitochondrial mutations.

It is extremely unlikely that the brain encodes information by introducing controlled mutations into DNA. Random mutations would add noise, not information.

It is likely that some mutations affect the efficiency of signal propagation, and encoding might adapt to this, so that equivalent inputs would create more or fewer or "stronger" or "weaker" synapses to generate the same output from the second neuron. But this could be detected functionally, if you could observe the signals and responses of neuron pairs.

It says "~1700 mutations per neuron", not "up to 1000 mutations".

Probably. On the other hand, cells which have a shorter lifespan might have more mutations, since some mutations are introduced during replication--several per genome per replication, I think? On the other other hand, if they're produced from stem cells, that would keep the mutations down.

It probably also is related to the fact that neurons use a lot of energy and have a lot of mitochondria. Energy requires oxidative phosphorylation in the mitochondria, which produces a lot of free radicals, which damage DNA.

Yet, 1700 (the number they reported is 1700, not 1000) is about typical. Many cancer cells have been sequenced; typical findings are that about 100 genes in cancer cells have acquired mutations, about 10 of which contribute to the cancer. Genes comprise about 1% of DNA; this suggests that the typical somatic cell has 1000 to 10,000 acquired mutations. (1000 assuming that all 90 mutations past the 10 that caused the cancer happened after the tissue went cancerous; 10,000 assuming they all happened earlier. Though that linear interpolation is a bad estimate, because genes and intergenic DNA mutate at different rates, owing to transcription and chromatin.)

However, if they sequenced just the exome (the expressed genes) rather than the entire genome, then 1000 is 10x typical. I can't tell, since their paper is paywall-protected.

First, you don't score points with me for saying not to use Hadoop when you don't need to use Hadoop.

Second, you don't get to pretend you invented shell scripting because you came up with a new name for it.

Third, there are very few cases if any where writing a shell script is better than writing a Perl script.

Better get some more servers. newsblur.com and dev.newsblur.com are not responding. "502 · NewsBlur is down"

I never found Google Code Search useful. What I wanted to be able to do was to search for code expressions with Google -- to be able, for instance, to search for the words "C++" or "call-next-method", exactly as I type them, punctuation and casing and all. No synonyms, no "corrections", no singular/plural conversions, no punctuation-stripping. Google never provided a way to search for exactly what I asked it to search for. Still the main thing missing from google search.

Home buyers are systematically better at planning for the long term than bankers, because most bankers don't plan to be working at the same bank 30 years from now - or, if they do, they don't expect to be held accountable for the loans they approve today in 20 or 30 years. Otherwise, what banker in the nation would offer 30-year fixed-rate mortgages at 4.25%?