MtotheThird

This is an excellent book, but as other comments have pointed out, the review misrepresents its argument.

I wouldn't say Desmond is exactly sympathetic to the landlords he profiles, but he does provide a well-rounded picture of them. The main thrust of his analysis is that eviction is a major cause of an inescapable cycle of poverty, and that preventing eviction would go further to reduce poverty and inequality than nearly any other government policy.

IIRC his recommendations involve a huge bump in the federal housing voucher program, decoupling it from work requirements (since it's difficult to get a job without a permanent home), and tying federal aid to code reform in cities (simplifying and loosening the housing code so that landlords would be less likely to ignore it).

I wouldn't be so sure of that. Try reading the comments on any Boston globe article about this affair. Ortiz has strong support from law-and-order voters and many people in Massachusetts outside of tech; in addition to Swartz, she's been very aggressive in her prosecution of corruption and misconduct in the state government. She has a compelling back story and (at least before this) was blessed by Deval Patrick.

I think she could definitely replace Martha Coakley as state Attorney General, at minimum (Coakley also did many odious things as a prosecutor and has never, to my knowledge, paid a political price for it). And a decade from now, long after Aaron has disappeared from the press cycle, she'll make a run for Senate or Governor and his prosecution will likely be nothing more than a subsection of her Wikipedia bio. These are the sad realities of politics in Massachusetts.

Most people here have had good advice. The only thing I would add is that there's a difference between vision and delusion. Visionaries have a strong and clear focus, but are grounded enough to recognize the merits of their competition and work within the ecosystem that customers / users already enjoy. Delusionaries pour scorn on their competition (or even claim to be so revolutionary that they HAVE no competition), and try to create a walled garden or vendor lock-in with zero install base. Even Apple made iTunes for Windows.

My first job out of college was at a startup. The founders were reasonable people who were heavily influenced and invested in by a delusionary, who saw their company as a vehicle to realize his "vision" of how technology should work. As a result we never really managed to grow beyond a couple of million in revenue, and that mostly from what amounted to R&D outsourcing from large firms with technically naive execs. Said delusionary repeatedly vetoed our attempts to make our product more relevant to a wider range of customers since it would compromise his vision. I bailed out of frustration, and a year later one of our customers just outright bought the firm to acquire the founders. I heard the delusionary was very satisfied with his exit. :p

I struggled to make LD work well, then someone smarter than me suggested the two character method. :)

Oh, I'm under no illusions that it's a universally applicable string matcher. What it seems to be useful for is normalizing human input in situations where someone can pick the best match. It handles typical normalization transforms well (word fragments, transpositions) and is very easy to index.

In my example, DPRK is the objective best answer .. There's pretty clearly a correct match in the value domain, the challenge is to help the user find it as quickly as possible.

I've had to do this a few times when building simple input normalization features. The trouble with scoring based on Levenshtein distance in that case is that it improperly penalizes phrases that are significantly different in length but similar in content.

For example, let's say I was searching a database of countries for "North Korea". In my list I have:

South Africa (LD: 6) Congo (LD: 9) Republic of Korea (LD: 11) ... Democratic People's Republic of Korea (LD: 28)

The actual answer (DPRK) will be pushed far to the bottom based on a naive ranking that uses LD.

The hack that's worked best for me? Rank based on the number of common two-character substrings between the source and target. It's simple, easy to build an index for, and has surprisingly great results. Its ideal use case is if you don't need to return a single absolute best result and can, say, present the three best to a human being and let them pick the match.

For the above search I'd get the following results using the two-character method:

Republic of Korea: 4 DPRK: 4 South Africa: 1 Congo: 0

Yeah, I can't really remember the last time I watched a Starz movie on Netflix. It's been nothing but TV series for me for at least a year now. Occasionally I'll watch a movie that shows up in new releases, but they've been signing a ton of premium shows lately, far more than I can watch.

I think it's a perfectly viable strategy for them to focus on series - more content per deal and a perfect match for a long-tail streaming strategy.

Disclaimer: I work for Google, this is my personal opinion, yadda yadda yadda.

I've used (and continue to use) Fastmail for many years behind a personal domain.

To call Fastmail reliable is a bit of a laugh. They've had at least three severe outages in my time there (we're talking 24+ hours without mail access). To their credit, nothing was lost in the end, however I was fuming by the end of the last one and determined to switch to Gmail. However inertia has kept me from doing so -- last time I checked it was tricky in Gmail to import a complicated set of folders like I have and retain the structure, so it's always been something I plan to get around to once I have time. Yeah right.

FWIW, I don't think there's been a significant outage since the Opera purchase. And I'm quite pleased with what I've seen of FM's beta interface.