I grew up in Silicon Valley, but I was young enough that I don’t have many memories of the first Internet bubble. There was the logoed duffel in my closet that long outlived Fogdog, an early online sporting-goods store acquired by eBay, and the classmate in fourth or fifth grade whose mom worked for Webvan, the online grocer, and one day struck it rich. (That day was probably November 5, 1999, when the company’s stock closed 65 percent above its initial offering.) I didn’t understand how a person working at a job could suddenly get rich in the course of a single day, but I didn’t want to look dumb, so I decided not to ask. A couple of years later, Webvan flamed out in spectacular fashion, after burning through more than $800 million in shareholder capital.

In 2013, Peter Relan, the founding head of technology at Webvan, published a post on TechCrunch discussing why the company had failed and how the next round of delivery start-ups could avoid the same fate. The first problem, he wrote, was customer targeting. Webvan’s strategy was to offer “the quality and selection of Whole Foods, the pricing of Safeway, and the convenience of home delivery,” a combination that attracted working- and middle-class shoppers. What it should have been doing instead, Relan concluded, was “providing a luxury” to a smaller, richer customer base. Second, he wrote, the company shouldn’t have invested in all that infrastructure. Webvan built cutting-edge distribution systems from scratch: giant networks of new algorithms, miles of conveyor belts, fleets of custom trucks with PalmPilot-wielding delivery drivers. At its peak, Webvan had a billion-dollar contract with the construction firm Bechtel for new distribution facilities around the country. Relan named Instacart and Postmates as lean start-ups that were learning from Webvan’s failure.

It’s easy to see Relan’s retrospective advice as obvious and Webvan as a predictable case of techno-folly, but what the company was trying to do was marshal technological and communications advances to offer consumers a relatively modest uptick in living standards. In the 20 years since, Americans have more or less given up on the idea that progress should simply improve everyone’s life. We now accept that every step forward comes with costs to someone somewhere, whether it’s content moderators in the Philippines or taxi drivers in New York. In the future, companies would not be making the same mistakes as Webvan, and the current crop of tech-based firms has learned to stay light and charge for convenience. Most of these services are no longer aimed at a large, inclusive consumer base but a small, wealthy one.

Take Instacart. The grocery-delivery start-up has a nominal valuation near Webvan’s peak market cap; but, on paper, it’s much smaller. Instead of trying to build a shopping infrastructure for the 21st century, Instacart is a middleman platform that connects buyers with pseudo-independent personal shoppers. When the company started out, it didn’t have relationships with stores, so it just marked up orders, adding costs for consumers rather than trying to minimize them. The personal shoppers would go to the grocery store, buy the products at the register, and deliver them. What was ostensibly a grocery service was in fact a digital market where you could hire someone to run an errand, at a considerable cost. This is a far cry from the Webvan distribution system: It neither seeks to build the infrastructure for new forms of stable employment nor attempts to bring costs down for most consumers.

In a 2000 report to the Securities and Exchange Commission, Webvan bragged that all of its couriers “are Webvan employees…. The courier training lasts two weeks and includes 36 hours of classroom training, 12 hours of driving training and 28 hours of on the job training…. Webvan’s couriers receive a competitive compensation package, including cash and stock options.” Commentators pegged Webvan’s delivery-labor costs at $30 an hour, or $45 in today’s money. Instacart, on the other hand, is notorious for cutting its pay and hiding tip options on the app, as well as refusing to classify its workers as workers at all. One analyst figured that to earn its valuation, Instacart would have to get its delivery-labor costs to $10 an hour, and then keep pushing down.

Sarah Kessler’s new book, Gigged: The End of the Job and the Future of Work, is an account of the new generation of Internet businesses that have migrated from the Webvan model to the Instacart one. A deputy editor at Quartz, Kessler tells these companies’ stories not from the point of view of their executives, but from the vantage of the workers on the lowest rungs. If the Webvan era, at its best, was an attempt at a highly automated 20th-century capitalism, Instacart and its compatriots had their start in the years after the 2008 financial crisis, when no one was looking to pay semiskilled workers $30 an hour, let alone $45.

In Gigged, we get a national tour of the new labor relations that provide the basis for Silicon Valley’s prosperity. Kessler began covering the gig economy in 2011, when she was herself a casualty of the post-financial-crisis labor market. Her first post, “Online Odd Jobs: How Startups Let You Fund Yourself,” was published on Mashable and offered a concise articulation of the way that many people viewed contingent labor at the outset of this new era. “You can really just create your own little economy around the things you’re good at,” one Skillshare teacher told Kessler enthusiastically. “It pays to get as many skills as possible,” said a user of Zaarly, a TaskRabbit competitor. This was the age of “an ATM in your pocket” and the dawn of “Uber for everything.”

The new middleman-app companies never planned to build futuristic infrastructure the way Webvan did; while becoming profitable, they were not interested in helping to raise living standards for consumers and workers. What they were interested in, above all else, was increased efficiency. But instead of using technology to reduce the role of labor in production through automation and cybernetics, they perform what is essentially arbitrage with human life. If Person A’s time is worth $50 an hour on the market, and Person B’s time is only worth $10, Person A should have a strong incentive to hire Person B to perform life’s unpleasant tasks. This kind of shallow thinking is what current Silicon Valley fortunes are made of.

It’s an old joke that ride-share companies are slowly inventing the bus, but it’s more accurate to say that they have reinvented the servant. Few Americans have ever been prosperous enough to afford the full-time, one-on-one attention of servants, but most can afford at least some small amount of personal service on occasion, if only the rare cab ride. The goal was to bring these personal services to middle- and upper-middle-income consumers at the expense of those who are forced to do the work. In a perfectly efficient world, people would be served by others to the exact degree that the market values their time more—and in the 21st century, the market doesn’t value most people’s time that highly. If middlemen-app companies can lower the barriers to service in such a world and charge a small fee in the process, they should, in essence, have a license to print money.

It has not escaped the critics of the gig economy that much of this work is also “women’s work” and that gender has long functioned, at least in part, as a technology that matches masters and servants.

One strength of Kessler’s book is that she is not interested in the efficiencies created by this new economy so much as the conditions suffered by those employed in it. What almost all of these workers have in common is that, for whatever reason, their time is not worth as much as other people’s in the 21st-century labor market. That makes them potential candidates for gig work.

But while some gig-economy start-ups claim that people seek out such work because they want the flexibility, the truth is that they seek it out because it’s all that’s left for them. “I haven’t really met many people in general who don’t value stability and safety,” Kessler writes. The “flexibility” is imposed, and workers do the best they can to adjust. The exception is the highly skilled coder who finds that he can do better for himself as an independent contractor on a project-by-project basis online than as a full-time staff employee. He should enjoy it while he can, because with all the tech companies pushing coding into school curriculums, he’s going to have a lot more competition in the future, which means less money. Just ask the cabdrivers.

Though many of Kessler’s subjects are not the average gig-industry employee, they can be portals to particular insights. For example, Kristy is a worker on Amazon’s Mechanical Turk platform (named for the 18th-century chess-playing “machine” that merely hid a human operator) who leads a “Turker” forum and uses computer scripts to cobble together a living from the site’s nearly valueless micro-tasks. Her story helps the reader to see just where the efficiency of gig labor comes from: Workers develop their own ways to speed up the completion of their tasks, even with no guidance.

In Capital, Marx theorized that, to maintain profitability, employers would eventually have to “depress wages below the value of labor power”—that is to say, below what the worker needs to subsist. Marx doesn’t elaborate much on this, and his interpreters, to the degree that they’ve noticed this section, have suggested that market dynamics and regulations tend to keep this level of wage depression from happening. After all, if you don’t provide workers with enough to live on, eventually they die, and then they can’t work for you anymore. But that’s a flat and economistic way to think about the world. What Kristy the Turker shows us is that, by depressing wages so low (in some cases, a penny a task), employers put the onus on workers to do the innovating themselves. One of the tasks that Kristy performed was identifying images to train image-recognition software. Clicking a mouse when you see something yellow at 3 cents a pop is terrible work and cannot possibly offer a living wage, but by finding and installing a script that let her use her keyboard instead, Kristy could make $10.80 an hour. By pricing the task so low, however, the employing firm successfully transferred the cost of software development to the Turker.

More than any technology—or even technology at all—what the gig-economy companies share is the strategy of shifting costs to workers. Why should Uber train drivers when people already learn how to drive on their own? Why write a user interface for image identifiers when you can pay them so meagerly that they have to figure one out themselves? Why supply cleaners with supplies if they’re not really “your” cleaners, technically? Why buy communication devices for your workers when they all have their own smartphones? And if you’re a potential worker who hasn’t invested in a smartphone… well, how do you expect to get a gig?

When Kessler attempts to seek out organizers to address these new labor relations, she’s frustrated. We can theorize all we want about the potential ways that gig workers can unionize, but the Uber driver she meets who wants to foment a strike doesn’t have a real base of fellow workers to appeal to, just a manic streak and a few grand to throw away on Facebook ads. He ends up trying to leverage a small payout from Uber to make him go away, but he’s unsuccessful here as well. From a worker’s point of view, there’s not a lot to be hopeful about in Gigged at all.

Even if we associate the gig economy with technology, cheap labor can’t speed scientific advancement, at least not the useful kind. It’s only when firms are forced to value their employees’ time that they’re compelled to innovate in ways that save labor and generate real social progress. Take a common right-wing meme about higher pay for fast-food workers: $15 per hour?? meet your replacement. The text is superimposed over a photo of a digital McDonald’s kiosk that says: order here. What the meme means is that capitalists have the technological ability right now to automate all sorts of unpleasant drudge work, but they don’t feel the need to make that kind of investment as long as they’re allowed to pay humans so poorly.

In the transition between the first and second round of Internet companies, we can see a clear adjustment of priorities. Now they orient their products and services toward a shrinking wealthy elite and draw their profits from a growing precarious subclass of workers. At its peak, Webvan had 4,000 employees and targeted budget-conscious consumers; Instacart has less than a quarter of that number and targets wealthier consumers who want other people to bring them stuff now. Genuine investment—especially in employee training—is to be avoided. Gig workers, as Kessler reminds us, do not compose a large segment of the world’s labor force, but neither did industrial proletarians when Marx and Engels started paying special attention to them. This category of workers who do more and get less will continue to grow in importance for the 21st-century economy. At this point it’s clear that this is mostly bad for most people.