Our pick of the week from around the web, including notable science failures, the French way to treat cancer and how relationships evolve on social media.

The case for blunder
Freeman Dyson | New York Review Of Books | 18 February 2014

Review of Brilliant Blunders, by Mario Livio, telling how five great scientists – Darwin, Kelvin, Pauling, Hoyle, Einstein – proposed five wrong theories. “The inventor of a brilliant idea cannot tell whether it is right or wrong. Humans need beliefs in order to live, and great scientists are no exception. Great scientists produce right theories and wrong theories, and believe in them with equal conviction.”

Is tech money good for San Francisco’s middle class?
Gregory Ferenstein | Tech Crunch | 15 February 2014

On balance, yes, say local economists. “Every tech job creates five in other industries, as compared to just two from a manufacturing job.” The tech sector is responsible for “the vast majority of the economic growth in San Francisco since 2010″. There is a but, of course, “and it’s big enough to need two plane seats”. Rising rents have forced many locals out of the city. “Those who get hit get hit hard.”

The formation of love on Facebook
Carlos Diuk | Facebook Data Science | 14 February 2014

What Facebook sees when we fall in love. “Relationships start with a period of courtship. Messages are exchanged, profiles are visited, posts are shared. During the 100 days before the relationship starts, we observe a slow but steady increase in the number of timeline posts shared between the future couple. When the relationship starts (‘day 0’), posts begin to decrease. We observe a peak of 1.67 posts per day 12 days before the relationship begins.”

The smart-pill oversell
Katherine Sharpe | Nature | 12 February 2014

Pills for attention-deficit disorders – usually methylphenidate or amphetamine – do calm people down and increase the ability to concentrate. These behavioural changes “make the drugs useful”. But “a growing body of evidence suggests that the benefits mainly stop there”. Medicated children don’t perform measurably better in later life. “Much beyond a year the benefits either vanish or shrink to clinically meaningless proportions.”

Construction of a Twitter aesthetic
Jason Fagone | New Yorker | 12 February 2014

Profile of Eric Jarosinski, assistant professor at the University of Pennsylvania, who tweets as @NeinQuarterly. His ambition: “I want to see myself as an aphorist – and not even a Twitter aphorist. I think we need to re-establish that as a profession”. Sample tweet: “At Starbucks I order under the name Godot. Then leave.” Style tip: “Tweets work best in dialogue form, because dialogue helps readers imagine a scene.”

The French way of cancer treatment
Anya Schiffrin | Reuters | 12 February 2014

It’s better than the American way. “When my dad began to get worse, the home visits started. Nurses came three times a day to give him insulin and check his blood. The doctor made house calls several times a week. French healthcare was not just first rate – it was humane. We didn’t have to worry about navigating a complicated maze of insurance and co-payments and doing battle with billing departments.”

Extreme medicine
NPR | 11 February 2014

Interview with Dr Kevin Fong, specialist in extreme-environment medicine. Topics include: doctors in space, freezing to death, drowning. “As you dive into the water, the water starts to compress the tissues of your body so that you become more dense. After you’ve gone maybe seven, eight metres from the surface you will no longer float. You become negatively buoyant, which is to say, you sink.”

A review of Her
Ray Kurzweil | Accelerating Intelligence | 10 February 2014

Perhaps the first film review to conclude with a list of the author’s relevant patents. “I would place some of the elements in [film director Spike] Jonze’s depiction at around 2020, give or take a couple of years, such as the diffident and insulting videogame character he interacts with, and the pin-sized cameras that one can place like a freckle on one’s face. Samantha herself I would place at 2029, when the leap to human-level AI would be reasonably believable.”

For more articles worth reading, visit The Browser. If you would like to comment on this, or anything else you have seen on Future, head over to our Facebook or Google+ page, or message us on Twitter.