The weird thing about working for an open-source company is that any random person can come by and look at what I've been doing: I resolved a couple issues brought up by Vault users. Today I've started

OK, that's a mouthful. I wrote an answer to How can I show that a function is not computable over at Quora, which brought up the Busy Beaver function. BB(n) is a typical example of a non-computable function.

When I decided to look for a new job, I had in mind the sort of thing I most wanted to do. I wanted to work on software security and correctness tooling, or more generally on infrastructure for building

I haven't kept on my blogging here in the past couple months, and there's a significant reason why. After nearly 11 years at Tintri, I've decided it's time for me to start something new. This wasn't an

I've wanted some onigiri for a while so I decided try making some for lunch. My first attempt was edible (it's hard to screw them up that badly) but not very pretty: I followed the instructions from

A mathematician I follow on Twitter recently referenced On Determining the Irrationality of the Mean of a Random Variable by Thomas M. Cover (The Annals of Statistics 1973, Vol. 1, No. 5, 862-871). Testing

The Bell numbers count how many ways there are to partition N elements into sets. What if we want to partition the elements into ordered groups instead, as this Quora question asks? For example, {a,b}

On Quora, I answered this question: How long does it take for your money to grow 10 times its original value if the rate of interest is 5% per annum? In my answer, I pointed out that there's a missing

Via Twitter: Recently my co-founder left Tintri. Now his email address has been aliased to me. Along with our interim CEO's email address. So now I get about 2.5x as much founder/CTO/executive-directed

An acquaintance verified their email with Steemit on November 29th and finally got the "One last step to set up your account" email yesterday, January 8th. That's a 40-day wait for a new user,

Day 13 is a simulation of carts on tracks. The track layout is given by an ASCII diagram, like this one: /->-\ | | /----\ | /-+--+-\ | | | | | v | \-+-/ \-+--/ \------/ I decided it would be better

Day 12 asks us to implement a cellular automaton. There's a lot of flavor text, but that's basically it. The rules are part of the puzzle; I don't know if they're actually different for everyone or not.

Haskell has a maximum function and it has lazy evaluations of lists. I come from Python that has a max function and list generators. But there turns out to be a crucial difference. Day 11 asks us to find

Day 8's puzzle was basically just about parsing a long list of integers correctly, and writing tree-recursive functions correctly. Both are pretty easy in Haskell. I defined the tree object like this,

I completely gave up on solving the Advent of Code problems as they went up, and I'm backfilling the days I missed, which is most of them. Day 5 was the first time I really felt happy with my Haskell solution,

In a data-science context, clustering refers to organizing data into categories by using some sort of distance metric. "K-means clustering" is a common technique for doing so, but other clustering

AC^0 is a circuit complexity class. It represents the set of decision problems that are solvable with a family of constant-depth unlimited-fanin polynomial-size circuits. Photo by Yung Chang on Unsplash.

During the first Internet boom of the late 1990’s, Sun was everywhere. Startup companies bought Sun hardware for their data centers, back when building your own data center was still a thing startups did.

The Day 7 puzzle is about parallel builds. So, just convert your input into a Makefile and solve it with GNU Make! OK, maybe not. I probably could have gotten that working in less time than it took me

(Inspiration: