James Tiptree Jr was the pseudonym of Alice Sheldon, a fascinating woman in her own right (I have her biography next up, but I thought I'd read the fiction first). In a thirty year career, she dazzled the field with short stories, including two Hugos and two Nebulas, and kept up a steady correspondence that hid her real identity.

Sheldon has her themes, and she hits theme again and again. Those themes are best summed up in the titles of two of the stories, "The Women Men Don't See" and "Love is the Plan the Plan is Death." Her stories are full of ignored and abused women achieving a kind of tenuous victory in the face of the patriarchy, and immense urges to breed and die that overwhelm futuristic astronauts in orgies of race suicide and nuclear apocalypse. The scope is Ballardian, but where Ballard dreams submerged psychological behemoths rising through the consumer detritus of the Space Age, Sheldon sees the Space Age as a futile masculine gesture against death.

The short fiction is incredible. Sheldon can write 10 pages of high-concept weirdness that'll knock your socks off. Unfortunately, that energy is dissipated rather than amplified over longer stories, which play around with drugged or mentally damaged unreliable narrators and non-linear temporality.

But holy shit, "The Screwfly Solution". If she's written just than, it'd be a career!

One of the odd gaps in my knowledge is that as a data scientist, I don't really know that much about data. Most of my work is is with very high level abstractions, things like Pandas DataFrames and Numpy Arrays or the various SQL tables of the data warehouse. Computers are fast enough that most things I work with are effectively medium data, large enough that I have to consider optimizations for my own sanity, if nothing else, but small enough that I can be confident that local RAM and disk will handle it without problems.


How many levels of SQL are you on? from @largedatabank

Working with data when you have one computer, or perhaps a simple database and an app engine, is pretty easy. You can trust that writes and reads will happen robustly and in a sensible order. But true web-scale big data cannot be done on any single machine. And when data is distributed across many disks and many data centers, things get complex very fast. I'm not a data engineer, I won't have to implement the gritty details of a distributed data warehouse and solve the hard problems of leader election, linearizability, serializability, and data consistency at scale. But knowing a little about how it works is useful.

The last chapter has some interesting nods towards "the modern data stack", and the idea that we can borrow old ideas from Unix, like the pipe operator liking together simple components, to describe dataflows as append-only change logs, allowing us to replicate state by replaying the past.

Blood, Sweat, and Pixels is a celebration of crunch time, and of all the pieces that have to come together to make a video game work. Video games are as complex an entertainment medium as we have, yet making them is a process beset by managerial chaos, incredibly bad tools (and I thought SQL Management Studio was a drag), and long long hours.

It's weird that 35 years on from video games becoming more than just a toy for geeks, the process for making them is so chaotic and poorly understood. Worse, what makes a game good is an emergent property of many different systems, all of which could depend on the tiniest details, so problems don't emerge until the game is almost done. And that doesn't even get into online games, which depend on thousands of real humans to make them work.

Schreier's stories are interesting, but his ten case studies tend to blend together, aside from the indie studios behind Stardew Valley and Shovel Knight. Video games are big business, bigger than the movies these days, but we still don't seem to know to make them, aside from whipping programmers and artists.

Severance is a novel which fully demonstrates Mark Fisher's adage that "It is easier to imagine the end of the world than the end of capitalism."


Ant infected with 'zombie fungus' Ophiocordyceps

Candace Chen experiences the end the world in two timelines. In one, she is the daughter of Chinese immigrants, working as a project manager at a publishing company where she supervises all the miniscule details that go into producing this year's Bible at the Chinese factories which actually make the things, along with all the other things. She has an alienating job that pays just well enough not to starve in a very expensive city, two cultures she doesn't feel a part of, friends she's not cool enough to hang out with, a moderately sketchy boyfriend who freelances and works on his novel, and dead parents.

In the other, Candace wanders through a world hollowed out by Shenzen Fever, a fungal disease that turns people into hollow husks of their former selves, enacting out the static rituals of the lives before biological damage claims them. She links up with a small group of apparently immune survivors heading towards a Facility in Illinois, falling under the domineering personality of rifle wielding Bob, a former IT manager. They scavenge what they need from the post-apocalyptic suburban landscape. Canadace is imprisoned and breaks free, ambiguously alive in the place that was once America.

So this is a perfectly serviceable novel about Millennial ennui in New York City, wedded to a perfectly serviceable post-apocalyptic fantasy. Ma has a genuine talent for those MFA-honed moments of specificity, an image of a starlet in a Shenzen print shop, a Puerto Rican chicken restaurant before a hurricane, the survivors removing their shoes and chanting a benediction to the tune of "New Slang" before starting one of their 'stalks' to get supplies. But the sum is weaker than the parts. This is a fine novel for people who can't handle the heavy stuff, who would find Mark Fisher, J.G. Ballard, and James Tiptree Jr. too hostile and distressing, who need to leaven post-human post-apocalypticism with quirky Millennial pink narrators. But that isn't me. I want my stories to go all the way, and Severance blinks in the finish.

The Shining Path is a fascinating account of Peru's bloody Maoist insurgency and the figures at its head, Comrade Gonzalo, the nom-de-guerre of former professor Abimael Guzman and his first and second wives, Augusta La Torre and Irena Iparraguirre. In the early 1970s, Guzman quit his job as a professor in Ayacucho and went underground, inaugurating a unique brand of People's War that would drench Peru in blood until his arrest in Lima. The narrative also ducks through secondary figures, Peru's great novelist Mario Vargas Llosa, rondero milita fighting against the Shining Path, police officers, reporters, and martyred activist Maria Elena Moyano.

Guzman's Marxist revolutionary leanings were fairly common in the 20th century, with successful insurgencies in an arc across Asia, Africa, and Latin America. While Peru was not strictly post-colonial, having won independence from Spain in 1821, the Criollo Limenos who had all the money and power were very much a case of 'new-boss-same-as-the-old-boss', especially in the harsh altiplano highlands and remote deserts and jungles. There was plenty of injustice left to fight, even if the worst excesses of the hacienda system had been reformed in the 1960s.

Shining Path rapidly liberated areas around Ayacucho from lackadaisical military control, but the actual indios who lived in the highlands proved more conservative and less revolutionary than the idealized People of Marxist-Leninist-Maoist-Gonzalo Thought. Sendero guerillas embarked on a brutal campaign of murder until the people who lived in the liberated zones matched their idealized People.

Meanwhile, Abimael, Augusta, and Irena directed the war from safehouses in luxurious Lima neighborhoods, enduring little more than isolation and security while their supporters froze and died. As the tide of retaliatory warcrimes bore against the guerillas, who disdained external support (Sendero bombed the Soviet and Chinese embassies, denouncing both major Communist powers as revisionist), the focus of the war shifted to Lima, with carbombings and assassinations in the tony Mira Flores neighborhood. The government of Alberto Fujimori countered with increasingly extreme tactics, including death squads and mass sterilizations campaigns, though the leadership was finally brought down by good policework, rather than Fujimori's reign of black sites, torture, and extra-judicial killing.

Starn and La Serna as both distinguished academic experts on Peru. They've deliberately written a popular book, focusing on people rather than theory. And indeed, so much of what made Sendero successful was a matter of it's leadership rather than ideology. But I'm despite the biographical details, I'm still left with major questions about what inspired thousands of Peruvians to join Sendero, to slaughter their countrymen, to die at hands of government death squads. The Shining Path is probably the first book an American interested in this war should read, but I hope it's not the last.

I'm somewhat mixed on this book. The authors are very clear about one particular design pattern, using Domain Driven Design to create aggregates which push events to a message bus, and layers of abstraction between the changeable facts of any specific database or infrastructure on one end, and the confusion of business logic which has to represent the messiness of the real world in a way that computers can understand on the other.

I'm somewhat pleased that I stumbled into their preferred pattern on my own before reading this book. But the meat of what I was interested in, test-driven development in the real world with legacy code, is left underexplained. I should spend more time with the accompanying Github, and in many ways this book is closer to my day to day than Designing Data-Intensive Applications, but as other reviews have mentioned, it's one architecture pattern.

Second Rebel sees the characters from the first book mostly scattered and on the run. First Sister has a name now, Astrid, and she's aiming at becoming the Mother, the spiritual head of the Gaen nation. Lito and Hiro are working with separate groups of spaceborne Asters, striking back against their corrupt leadership. And Lito's sister Luce is a new viewpoint, as her idealistic artist gets wrapped up in the deadly intrigues.

There's lot of drama here, big emotions same as the first novel, and a little better sense of intrigue and setting. I missed the audiolog revelations from Hiro, which provided a unity of focus that this novel lacked. Four points of view were one too many, and the conflict, playing out in the outer belt between Asters, Icarii, and the Synthetics who have locked humanity in the inner system. Second Rebel is stronger than the first book in some ways, with a more assured setting, but weaker in terms of mystery and that satisfying unity of an ending.

Still, I don't read much new speculative fiction these days, and a light novel or terrible revenge is just the thing.

A longer review will follow, but Restricted Data is a close look at the practice of American nuclear secrecy. These days, we take the whole complex of classification as a fact of nature, much like the fissile possibilities of U-235. But of course, classification is whole artificial, a political system born out the Manhattan Project. And while classification may seem simple, a wall between the open ordinary world and the official secrets of the states, the reality is anything but simple.

The first regime of nuclear secrecy was purely voluntary. Leo Szilard recognized the military power of his idea of a self-sustaining nuclear reaction, and assigned the patent to the British royalty. As fascism role in Germany, he embarked on a not very successful campaign to halt publication on fission research. While pre-war activity could be tracked by who stopped publishing, the barriers to the bomb were primarily logistical. Only the United States could afford the expense of isotope separation and then bomb design. Classification in the Manhattan project was a system of compartmentalization, designed mostly to protect against leaks, and penetrated in key places by Soviet spies like Klaus Fuchs.

The category of Top Secret was actually invented for the Manhattan Project. Where the story gets weird is that immediately after the bombings of Hiroshima and Nagasaki, a concerted press effort from popular articles to the technical Smythe report discussed the bomb, create a world where some facts were public and others denied. The category of "Restricted Data" related to anything about science that could be used to make an atom bomb, basic scientific facts rather than the military secrets of defense plans and names of spies.

Since 1945, the story has been one of unwinding of secrecy, with categories of civilian nuclear science around power plants, and activist efforts to reconstruct the Teller-Ulam hydrogen bomb configuration from publicly available data in the 1970s. Modern nuclear non-proliferation is based on control of material: uranium ore and centrifuges, rather than control of knowledge. But even so, the world of nuclear secrecy is vast, a black monolith in a supposedly democratic state. Thinking about nuclear secrecy reveals many paradoxes.

Fire is a collection of short pieces from the 90s by master of narrative non-fiction Sebastian Junger. The work starts out with firefighting crews in the American west, and then moves through the conflicts of the 90s: Kosovo, Sierra Leone, Cyprus, Afghanistan. It's a weird collection, written in Fukuyama's End of History and published when history returned with a vengeance after 9/11. Two decades on, could we have a little less history, please?

Even pedestrian Junger is an entertaining and thoughtful read. My two favorite pieces were on the frozen conflict in Cyprus, which asks the question "If this went on forever, would anyone mind?", and a short piece on those who court danger called "Colter's Way", which gets at the heart of Junger's whole project: finding moments of adventure, of desperate life-or-death survival, beyond the stifling comforts of modernity.

I love Gödel, Escher, Bach, but GEB is over 800 discursive pages. Turing's Vision is a short translate of Turing's key paper "On Computable Numbers, with an Application to the Entscheidungsproblem". Turing paper conclusively proved a key result in mathematics, that some questions cannot be answered "yes" or "no", but will drift in infinite indeterminability. And second, his model of a simple machine with states and an infinite memory provided a conceptual design for the first universal computers as opposed to electro-mechanical calculating machines.

While no one programs pure Turing machines, instead working with friendly abstractions, the Turing machine is the strongest model of computation that we know how to build (stronger ones involve breaking the laws of physics). Bernhardt offers simple and elegantly explained proofs by contradiction to show the powers and limits of this class of machines. While the arguments are fuzzier than pure math demands, this is also a book that is almost thrilling in its readability, and it's a math book!