The Heaven & Earth Grocery Store is a meandering historical novel, mostly about intertwined Jewish, Negro, and White communities of Pottsville Pennsylvania around 1936. The Jews and the Negros live on Chicken Hill, a hardscrabble muddy neighborhood without paved roads or proper plumbing.

The story arcs around several main characters. Moshe runs a pair of dance halls that book Jewish and Negro musicians. His wife Chona is the daughter of the former rabbi and runs the titular grocery store, which offers a lot of credit to the neighbors. Nate is Moshe's house manager, a tough and taciturn black man who does everything that needs doing. Addie is Nate's wife, and also Chona's nurse as Chona suffers an unknown degenerative illness. Fatty is another black man who runs a jook, as well as a thousand other miscellaneous schemes, and Soap is Fatty's Italian best friend. And there's secondary and tertiary characters galore, down to microscopic differences in where the Jews came from, as well as the graduations of social rank among the Negros. Of course, the basic law is that they're all below even the meanest White man.

The book opens in 1972 with a mysterious skeleton, likely the victim of a murder, as if to gesture at a plot. But nothing happens for quite a while, until our main characters are sheltering a Deaf Negro boy named Dodo from the state, and local doctor and KKK leader Doc Roberts arrives at the grocery store. Plot ensues, Dodo winds up in a horrific state sanitarium, and everybody has to work together to get him back.

What works is the sepia-colored reconstruction of history. McBride is both Jewish and Black, and I can say that the Jewish half hit home pretty well. There are my people, though with a lot of the rough edges smoothed off. Characterization is less solid. An adage that I believe in is that novels are defined by the choices a character makes. In this book, there are less choices and more extrapolations of an essential nature of each character. Chona is principled and feisty. Doc Roberts is malevolent and self-centered. And so on.

Everything winds up tidily enough, but the artifice is evident. Compared to some literature I've read, individual bits are nice and there is a plot, eventually, but this book is less than the sum of its whole.

In early 1942, one of the darkest periods of the Second World War, US Navy Salvage officer Commander Ellsberg, recently rejoined, was given a vital project. The Eritrean port of Massawa was the only potential maintenance facility between Alexandria, which was under imminent threat from Rommel, and Durban, 10,000 miles from the front in South Africa. Also, the drydocks at Alexandria were tied up keeping the two sabotaged British dreadnoughts afloat, meaning that no other ship could be repaired. Cargo ships, their hulls covered with years of barnacles and seaweed, were reduced to sailing at 6 knots--sitting ducks for Axis submarines. The fighting cruisers and destroyers of what was left of the British Mediterranean Fleet could not repair damage, and were being attrited to uselessness.

Unfortunately, Massawa had been comprehensively sabotaged by the Italians before it was captured. Every single machine in the shops had been battered with sledgehammers and vital parts thrown into the sea. The harbor was chock full of wrecked ships, with bombs blasted in their side. The floating crane was sunk, and both floating drydocks had been sunk with eight charges apiece to make sure they stayed at the bottom.

For this critical mission, Ellsberg's resources were practically nil: himself, and any civilians he could scrape up. Every single American Navy salvage diver and mechanic was at Pearl Harbor, getting the battleships back in action. The British had nothing they could offer, stretched as they were. And worse, Massawa was a legendarily awful station, combining brutal tropical sun with sweltering humidity.

Ellsberg is a talented memoirist, and makes the drama of engineering at the ends of the earth and the absurdity of his situation come alive. There is plenty of absurdity. The troop transport he took to Africa was comprehensively blacked out as a precaution against U-boats, except for the red and green running lights, which its rigid captain refused to shut off in case they collided with another ship at night. Ellsberg had to beg, borrow, and steal materials, while there was an entire warehouse of new salvage equipment under control of an incompetent British contractor (a monocled, safari-suit wearing Colonel Blimp) that he couldn't touch. The American contract seemed to be actively sabotaging Ellsberg, building two entire useless base housing facilities before starting on industrial buildings, messing up payroll for over six weeks in a row until a full-blown strike occurred, and illegally issuing orders replacing Ellsberg.

Against these obstacles, with a bare handful of American salvage professionals and a motley labor force of Eritreans, Italian POWs, Arab carpenters, and miscellaneous Sudanese and Indians, he somehow accomplished miracles. To give an example, he and his small crew went through every fragment of machinery the Italians had left behind and managed to assemble a working lathe and mill, which they then used to machine parts to repair the rest of the machinery. Work was accomplished under the most brutal conditions, broiling on steel decks under the sun and inside stifling compartments at well over 120 degrees (he brought out a thermometer once-and the reading was enough to prompt everybody to knock it off for the day) to patch holes and make various wrecks water and airtight so they could be salvaged.

This is a story of endurance under the worst conditions, because it had to be done, and of leadership and persuasion. One anecdote sticks out. Ellsberg had hired large numbers of Eritreans as scrapers and painters, since they were the only group available in the numbers needed for this key jobs. The British regarded the Eritreans as some of the worst laborers in their entire Empire, and objectively, the average Eritrean of 1942 was scrawny, undernourished, entirely uneducated, and had spent their entire life get cooked by tropical heat. They were not doing their jobs, and could not be berated into working at any pace faster than a crawl. And their payrate was fixed at a miserable sum that could not be increased. Ellsberg gathered their sheiks (hiring was collective by tribe and managed by the sheiks) and explained that this job took three days. If it was not done on three days, they'd be fired and never hired again. If it was done faster than three days, they got paid for three. The Eritreans did it in two.

As he put it, "There is no worst labor in the world. Touch the proper chords-pride, incentive to produce, whatever fits the situation--and men will be found men, whatever their color, whatever their physique."

Tuf Voyaging is mid-80s science fiction fluff from George RR Martin before he was the Game of Thrones guy, a series of 7 linked short stories that flicker between charming and rancid, depending on how generous you're feeling.

The story introduces us to a group of rogue academics and criminals on a search for a ancient starship of immense power, an Ecological Engineering Corps seedship with a library of billions of cell samples, quick cloning tanks, and the power to remake worlds. They hire as their transport failing freelance trader Haviland Tuf. Tuf himself is an odd freak (absolutely autism coded, as the kids on TikTok would say). Physically, he's a pale hairless eight foot tall giant who speaks in a bass monotone. He's overly literal, honest to a fault, is a committed vegetarian, hates to be touched, loves dark ale and mushroom wine and strategy games, and prefers the company of his cats to any people.

When they arrive at the Ark, Tuf's trading ship is damaged by the defense systems and the crew immediately falls out over how they'll split the unimaginable wealth and power of the seedship. Betrayal and the hazards of a Big Dumb Object in space whittle down their numbers until Tuf is the last man standing. He decides to embark on a new career as an ecological engineer, and then the rest of the book.

The three main stories focus on S'uthlam, a city planet facing an immanent Malthusian crisis (the planet's name is an anagram for Malthus-subtle). Religious and social mores prevent any form of population control, and the planet's exponentially growing population of 40 billion plus is perennially years away from famine and/or an existential war with their neighbors. In three stories, Tuf gives them new crops, criticizes their society, and finally forces mandatory birth control, observing that as with cats, either you choose to sterilize them, or you find yourself tossing a sack of kittens out an airlock.

The intervening stories are a bit more fun. "Guardians" is the best, where Tuf helps a planet under assault from viciously evolving sea monsters reach a peace accords with the natives. The other stories have him undermining an aristocracy of animal pit fighters, and dropping plagues on an anti-technology zealot calling himself Moses.

GRRM knows what to do on a sentence to sentence level. But the setting and characters are a sketchpad for the moral lessons, and how much you'll enjoy the stories depends on how much you like Tuf's "Oh dear, no one appreciates my honesty and assumes that I'll use my god-like power to screw them over. Well, time to use my god-like to do what I think is best for them." I split the difference and gave this collection three stars.

The thing about kids is that they don't listen. I didn't listen much to my parents, my son doesn't listen much to me. As parents, we want the best for our kids, but realistically we can't just make them do it. And as a parallel, as a former academic, if professors could really indoctrinate their students, the first thing they'd do is indoctrinate them to read the goddamn syllabus!

While kids don't hear the words you're saying, with adults coming through as a kind of Charlie Brownish wurble, they are very adept at hearing the tone that you use. Mogel has a lot to say about being steady, calm, and curious, all of which are good advice. Don't yell, don't hector, don't use sarcasm, and leave room for love and understanding to emerge.

The flipside is that this slow/gentle approach works great when things are good, I'm not sure how well it works when things are bad. As parents we all have moments when we are tired, stressed, overwhelmed, or otherwise not putting our best foot forward. My son is three, so our conflicts have been over things like baths and potty training, which are in the grand scope of things pretty minor. While life is often rocky, what if you have a teenager who's seriously depressed, failing classes, and hanging out with the worst crowd?

A large portion of this book is sectioned by girls and boys, and talking to mom or dad. As a therapist, Mogel has seen a lot of families, so I'm moderately confident her stereotypes are more grounded in numbers than gut feelings, but this book is still based heavily on gender stereotypes.

Data science is not that hard. Simply clean and annotate a dataset, select one of the available algorithms from the basics like linear regression to the latest transformer architecture neural networks, train and optimize a loss function for accuracy, precision, and/or recall, make some pretty plots, and move on. You did it!

Unfortunately, the company is going to need you to keep doing that every single day, forever.

Deploying and maintaining models in production is machine learning engineering and operations, and it is in fact pretty hard. Designing Machine Learning Systems is a solid introduction about how to go from ad hoc data science to continual learning with machine learning engineering and operations.

The first and foremost issue is one of data shifts. The data coming into any system is continuously evolving, and entropy means that changes are away from the data that the model was trained on. This means that a useful ML product has to be constantly retrained and redeployed, even in the absence of

The second issue is that platforms and tooling for doing this is apparently not great. Code versioning via Git is solid. Model versioning via some kind of artifact store is okay, but varies via company. Data versioning is likely bad, requiring painstaking reconstruction from a data swamp (like a data lake, but full of sludge). And the totality of being able to maintain a consistent workflow around code, data, models, and compute is basically non-existent.

This book has a lot of good questions to ask and targets to aim for, especially in the later chapters (I found the first five or so chapters very basic), but fewer good answers, particularly around the key questions of what metrics to monitor and when to refresh models. I guess this is why they pay us.

One of the clearest symptoms of modern MBA brain poisoning is that everyone knows that it takes leaders a lot of time to get up to speed on a new organization, and even longer to demonstrate results. Yet the average tenure of an executive is a bit over two years. Companies are constantly shuffling around more or less identical haircuts in the hopes of something different happening.


Stock photo of a Real Business Situation. Can you help this man draw a plot that goes up and to the right?

In the desolate wasteland of business books, The First 90 Days is pretty good. Watkins makes several key observations: First, your success as a leader will depend on establishing credibility early. What counts as credibility is local to each organization. As such it is important to objectively figure out what kind of situation you're in, and act appropriately.

Start-Ups have limited resources and need to scale quickly, requiring agility and inspirational leadership. Turnarounds require decisive action to cut out dead wood and restore morale in broken organizations. Aggressive growth has to take a successful project or culture and make it big, without sacrificing what made it valuable in the first place. And finally, reframing and business as usual require slower and more sensitive skills.

Watkins also recommends a clear timeline, with 30 days to get your bearings and another 60 to deliver results. Work with your superiors and subordinates so they understand the need for deliberate speed, and things will go better. And as always, the politics are foremost.

Everybody knows how Rome fell. After a succession of crises, plagues, civil wars, and especially barbarian invasions, the once illustrious empire ended. Even a perfunctory examination shows some issues. Rome was sacked in 410 by Visigoths under Alaric, sacked again several times, and the last Western Roman Emperor, Romulus Augustus, was deposed in 476. Roman institutions limped along in some form or another for the whole 5th century. How did people at the time experience the fall of Rome?


Fantastic cover design on Gibbon's Decline and Fall, showing a pillar crumbling over time.

Mathisen conducts a close read of available sources for Gaul in the 5th century, focusing primarily on the letters of Sidonius Apollinaris. Mathisen establishes a Gallo-Roman aristocratic identity as one of the "good people", a criteria established by wealth, lineage, office holding, and literary accomplishment, and then discusses how that criteria shifted over time.

Wealth was always the foremost criterion. Aristocrats were distinguished by their estates. Holding on to them became trickier in the 5th century, requiring aggressive legal and physical defenses against barbarian and Roman neighbors. Lineage mattered, and intermarriage with barbarians was discouraged, but of course bloodlines were negotiable for the right price. Offices and careers shifted from the Roman civil service to the Church hierarchy, with some Bishoprics becoming family holdings, as well as ad hoc work for barbarian kings. And finally, literary matters were very important, though this is somewhat self-selecting. Sidonius collected his letters for publication, which was sponsored by his son, and wrote extensively about the importance of cultivating a literary circle.

Yet at the same time as Mathisen makes a convincing case for the ductility of Gallo-Roman aristocratic identity, it is impossible to deny its diminishment over time. Many aristocrats sought exile. Travel essentially stopped, first between Gaul and other provinces, and then within Gaul, as various barbarian kings divided up the region. Sidonius writing that letters were the best form of friendship seems to have been sour grapes for a man who clearly did not feel comfortable embarking on even moderate journeys. While courts continued to function for some decades, ultimately justice became a matter of helping oneself, contributing to cycles of feuds.

While Sidonius was recognizably a Roman aristocratic writer in a lineage we would recognize with Cicero and Pliny, no one followed him. By the 6th century, the time of his grandchildren, genealogical records show Roman and German names in the same century. The fall of Rome was perhaps less violent than we remember, but it was definitely comprehensive.

On the Bottom is Ellsberg's first popular book, an account of the salvaging of submarine S-51 from 160 feet of water in 1926 by a team of US Navy divers. Compared to the other books I've read, it's more technical and has less human interests, this is after all a team of competent professionals without any of the cultural conflicts that make his other books so interesting.

The story of men against the sea has enough drama. Deep sea diving is one of the most hazardous and extreme activities people do, and crude technology of 1926 meant that everything had to be done by hand in the dark and cold, where divers survived in a precarious equilibrium and a single mistake could prove fatal.


Mark V Standard Diving Dress, of the kind used by Ellsberg. From Will Kutscher

Perhaps the most terrifying way to die was a squeeze, where air pressure in the suit dropped below the water pressure and the diver was compressed into the helmet. Conversely, if air pressure built up too high, the suit would blow up, forcing the diver spread eagle and unable to operate their relief valves as they shot to the surface. Ascend too quickly and the diver would be crippled or killed by the bends. At depth, oxygen itself had an intoxicating effect, so try thinking while five drinks drunks, and moving while carrying hundreds of pounds of extra weight. Visibility at the bottom was between feet and nil, with much work having to be done by feel. One diver got helplessly loss for a half-hour in a 15 foot triangle between the sunken submarine and a house-sized pontoon. A narrow lifeline connected the diver to the surface, carrying air and a balky telephone, and if that line got snarled or the diver got stuck, there was little chance of help. Suits leaked, dives and decompression sequences took hours, and their were ordinary risks of exhaustion and pneumonia.

The plan to raise S-51 was complex. First the remaining good compartments of the damaged submarine would have to be made watertight and pumped full of air. Divers practiced on the sistership S-50 until they could turn the necessary valves blindfolded, and then did the same same thing on the bottom, in a listing wreck full of corpses. Repairs and special hatches had to be installed to keep the pressurized air in, which was literally backbreaking labor.

Next, a cradle of chains had to be laid around and under submarine. The bow and stern rose above the bottom, making this easy, but the two middle cables had to be passed through tunnels dug undersea. These were dug with water-pressure, leading to one of the more incredible feats of bravery I've read, where Diver Smith was in a 20' long tunnel that caved in on him. He managed to get the hose turned around and held it between lead-booted feet to dig himself out. Then he caught his breath and went back into the hole, a calm and cold-blooded heroism under literal and metaphorical pressure to get the job done.

After that, pontoons had to be submerged and laid alongside the submarine, a process Ellsberg compares to lowering a train car into place from the top of a 16 story building in a midnight gale. And once all this was done, there were the sundry matters of accidentally raising the submarine before a storm and having to sink and untangle the entire mess to do it again, running the wreck around in the East River and having to salvage it a third time in shifting tides, and finally bringing the ship and the men home.

As a first book, Ellsberg's writing is a hair weaker, and salvage doesn't have the same historical weight as World War 2. Yet for a certain type of nautical geek, this is a fantastic yarn and well worth the read.

One of the books that profoundly influences my worldview is Seeing Like a State. There, Scott describes the central problem of statecraft and of government as one of legibility; the state must make its citizens and their activities visible before it can appropriate revenue and orchestrate any plan for the general welfare. In the 19th century, this was a matter of censuses, maps, and ledgers. In the 21st century, it's about websites and databases. As anyone who has interacted with the American government experiences, this process of becoming legible is profoundly awful. Taxes suck, beyond the financial cost. The DMV is a pain in the neck. If you've been to the post office lately, you know (Bob's Burgers link, trust me). And god forbid you ever have to apply for social services or get the justice system to fix a mistake. The inability of government to do anything well, particularly anything involving digital service delivery, is a joke as tired as the deal with airline food. If statecraft is about seeing, the American government is stumbling around blind.


U.S. President Donald Trump looks up toward the Solar Eclipse while joined by his wife first lady Melania Trump on the Truman Balcony at the White House on August 21, 2017 in Washington, DC.
(Photo by Mark Wilson/Getty Images) (Mark Wilson, 2017 Getty Images)

Pahlka's analysis is based on her personal experience in the tech industry, leading her non-profit Code For America, and a year as deputy chief technology officer for government innovation in 2013, where she learned a lot of hard lessons in the Obamacare rollout. She's kept in touch with colleagues from that time. She has deep experience with the cultures of both tech and government. The American government's failures with technology over decades are primarily self-inflicted wounds, and one where there is a chance of hope, however slim, with a potential bipartisan

Pahlka's core premise is that there is basically a good way to write usable software, as developed by internet-era companies and described in the Agile manifesto. Start with something small and simple. Iterate rapidly, using user experience research to guide the development of the product. Trust the expertise of the people involved and their ability to make good decisions. Have someone empowered to make decisions in charge who doesn't lose sight of providing a success product to end-users. There's a lot of (justified) anti-Tech and anti-Agile opinions out there, but she isn't some kind of one-true-wayist or egomaniacal platform god-king. This is the mundane tech stuff of "How do we build and maintain a website that doesn't absolutely suck?"

The largest barrier is cultural. Government is a hierarchy. At the very top are We The People, who elect representatives who write laws, which are turned into policy by senior bureaucrats, the most exalted of whom are at the Office of Management and Budget, which become program requirements and bids negotiated by mid-level bureaucrats and contractors, which is built by software developers who have no agency, which becomes a system used by junior bureaucrats and inflicted on the public. This a classic waterfall method, which as Pahlka acidly observes is "a promise not to learn anything."

Hierarchies are also embedded into the details of the policy, which for any specific program will likely be a combination of Federal, State, and County guidelines, as interpreted by multiple agencies, including both the primary and partner agencies, judges, and a constantly updating set of legal changes and refines at multiple levels. Enacting policy is genuinely complex.

And finally, and most fatally, bureaucratic culture is built around consensus and avoiding blame. You get promoted by following the rules exactly, and saying "I followed the process" is the only shield when things go wrong. Which means that there is no one who can say "This requirement is dumb, and it is killing the rest of the program. I am removing it. If necessary, we can do it later", short of going all the way back to Congress or having a major lawsuit. Americans are rightfully suspicious of the super-bureaucrat in the vein of Robert Moses or J. Edgar Hoover, but in our fear of empowering a dictator, we've instead empowered nobody.

Culture aside, there are a few specific laws and norms which are particular pernicious. A longstanding interpretation of the Paperwork Reduction Act effectively prohibits agencies from surveying and interviewing their users without a full on Federal IRB. The Administrative Procedures Act requires a multi-step detailed bidding process for literally everything. Various civil rights decisions argue that if a program is not maximally inclusive,

And a variety of laws and norms say that Federal agencies should buy, not build, what they make. Which is all fine and dandy, except that while in many cases a lot of digital technologies are commodities, like web hosting and database servers, the specifics of the logic that is encapsulated ARE THE POLICIES! In a fight between a human's opinion of how code works, and the computer's reality of how the code is working, the human will lose every time. Agencies which do not have some in-house capacity to operate their own data systems are incapable of seeing like a state.

The end result is policy vomit as product. The entire program gets thrown into a poorly designed questionnaire with no particular organization, confusing wording, and an actively hostile attitude towards actually getting anything done. For example, unemployment relief basically relies on former paystubs, and good luck if you don't know where those are (I'm pretty organized, and I'd have to search), or had informal employment as a gig worker or something.

This book is full of horrifying tech war stories, elaborations of what you'll read at the excellent Statecraft blog on fixing the Veteran's Administration and the Department of Labor. The VA had an electronic form system which would only work on the specific versions of Internet Explorer and Adobe Acrobat installed on VA computers. The California Unemployment system collapsed under COVID because millions of people were dumped into a manual queue based on minor data entry errors associated with a 0.25% instance of fraud, which could only be remedied by a small number of 20+ veterans due to the arcane nature of the system. A next-generation GPS satellite kept missing latency targets due to a requirement to translate data from standard UDP to a complex enterprise service bus back to UDP because of a game of telephone from mid-90s IT standards. Most horrifyingly, Immigration was unable to reunite family after Trump's family separation policy because families who had previous been under a single case number were now under ones for adults and separate one for each child, and there was no possible way to link case numbers in the INS database.

Ironically, I finished this book in a DMV lobby, waiting two hours to submit a piece of paper in place of a broken website. But I agree entirely with Pahlka's critique and argument. My last job was at a bank, where I was an in-house dev responsible for automating a bunch of reports that were not being done properly by vendor software. I had a toaster's worth of compute, a few gigs of database space, and with that I could do things in two weeks which would take our vendors six months and $3 million.

Liberals want a government that delivers services. Conservatives want a government that is efficient and gets out of the way. In the 21st century, this will require a new mode of supplying government services, new types of leaders, and a culture that is willing to be accountable and learn from mistakes. We should automate what we can, and trust the discretion of human beings to what can't be automated. Better things are possible. After the fiasco of the healthcare.gov launch, a follow-up program to help Medicare pay more to doctors who keep people alive launched more or less smoothly. In countless ways, government is getting better. The questions is if government can rebuild trust by being effective before cynics in positions of power bring down the whole state.

Common advice is to write what you know, which means that a fair parody of serious literature as a genre is a story about an aging male author with some kind of driftless personal life being prompted to go on a transformative adventure, usually in the erotic company of a younger woman. Taken as a member of that genre, The Paper Men is an awful novel. Taken as a parody, it rises to occasional flashes of brilliance.

Wilfred Barclay is a certified titan of English literature, coasting on the reputation of his first novel, and also a miserable and curmudgeonly alcoholic. He crosses paths with the ambitious American academic Richard Tucker, who seeks to become the authorized biographer of Wilf Barclay. And over the course of several years, the two men grind each other to pieces.

Barclay's disdain and hatred of Tucker is the animating spirit of this book, a vile gut level enmity that propels him across Europe in an alcoholic fugue. Barclay's misogyny is also a major character trait, he doesn't much care for his ex-wife, daughter, or any of the other women he encounters, except his lust for Tucker's younger wife, but compared to how he treats Tucker himself, the misogyny is small change.

Three scenes illuminate the book: The opening, where Barclay mistakes Tucker for a badger rummaging in his trash bin and shoots him with an air gun, an extended encounter at a Swiss mountain lodge where Tucker attempts to bargain his wife's body for Barclay's assent, and then saves Barclay's life in a mountaineering accident, and the conclusion, which is worth not spoiling. A fantastically savage and ironic spirit of farce animates these scenes.

Unfortunately, they're connected by about 150 pages of nothing much in particular, a venting of bile and delirium tremens that is neither entertaining not informative.