Todd Park (third from left) with members of his digital team (from left): Haley Van Dyck, Vivian Graubard, Park, Jennifer Anastasoff, Mikey Dickerson, Erie Meyer, and Brian Lefler. 
Michael George
The White House confirmed today the rumors that Todd Park, the nation’s Chief Technical Officer and the spiritual leader of its effort to reform the way the government uses technology, is leaving his post. Largely for family reasons—a long delayed promise to his wife to raise their family in California—he’s moving back to the Bay Area he left when he began working for President Barack Obama in 2009.
But Park is not departing the government, just continuing his efforts on a more relevant coast. Starting in September, he’s assuming a new post, so new that the White House had to figure out what to call him. It finally settled on technology adviser to the White House based in Silicon Valley. But Park knows how he will describe himself: the dude in the Valley who’s working for the president. President Obama said in a statement, “Todd has been, and will continue to be, a key member of my administration.” Park will lead the effort to recruit top talent to help the federal government overhaul its IT. In a sense, he is doubling down on an initiative he’s already set well into motion: bringing a Silicon Valley sensibility to the public sector.
It’s a continuation of what Park has already been doing for months. If you were at the surprisingly louche headquarters of the nonprofit Mozilla Foundation in Mountain View, California, one evening in June, you could have seen for yourself. Park was looking for recruits among the high-performing engineers of Silicon Valley, a group that generally ignores the government.
Achievements that Internet companies seem to pull off effortlessly are tougher than Mars probes for federal agencies.
There were about a hundred of them, filling several lounges and conference rooms. As they waited, they nibbled on the free snacks and beverages from the open pantry; pizza would arrive later. Park, a middle-aged Asian American in a blue polo shirt approached a makeshift podium. Though he hates the spotlight, in events like these—where his passion for reforming the moribund state of government information technology flares—he has a surprising propensity for breathing fire.“America needs you!” he said to the crowd. “Not a year from now! But Right. The. Fuck. Now!”
Indeed, America needs them, badly. Astonishing advances in computer technology and connectivity have dramatically transformed just about every aspect of society but government. Achievements that Internet companies seem to pull off effortlessly—innovative, easy-to-use services embraced by hundreds of millions of people—are tougher than Mars probes for federal agencies to execute. The recent history of government IT initiatives reads like a catalog of overspending, delays, and screwups. The Social Security Administration has spent six years and $300 million on a revamp of its disability-claim-filing process that still isn’t finished. The FBI took more than a decade to complete a case-filing system in 2012 at a cost of $670 million. And this summer a routine software update fried the State Department database used in processing visas; the fix took weeks, ruining travel plans for thousands.
Todd Park 
Michael George
Park knows the problem is systemic—a mindset that locks federal IT into obsolete practices—“a lot of people in government are, like, suspended in amber,” he said to the crowd at Mozilla. In the rest of the tech world, nimbleness, speed, risk-taking and relentless testing are second nature, essential to surviving in a competitive landscape that works to the benefit of consumers. But the federal government’s IT mentality is still rooted in caution, as if the digital transformation that has changed our lives is to be regarded with the utmost suspicion. It favors security over experimentation and adherence to bureaucratic procedure over agile problem-solving. That has led to an inherently sclerotic and corruptible system that doesn’t just hamper innovation, it leaves government IT permanently lagging, unable to perform even the most basic functions we expect. So it’s not at all surprising that the government has been unable to attract the world-class engineers who might be able to fix this mess, a fact that helps perpetuate a cycle of substandard services and poorly performing agencies that seems to confirm the canard that anything produced by government is prima facie lousy. “If we don’t get this right,” says Tom Freedman, coauthor of Future of Failure, a 61-page study on the subject for the Ford Foundation, “the future of governing effectively is in real question.”
No one believes this more deeply than Park, a Harvard-educated son of Korean immigrants. Mozilla board member and LinkedIn founder Reid Hoffman had secured the venue on short notice. (“I do what I can to help Todd,” Hoffman later explained. “We’re very fortunate to have him.”) Park, 41, founded two health IT companies—athenahealth and Castlight Health—and led them to successful IPOs before joining the Department of Health and Human Services in 2009 as CTO. In 2012, President Obama named him CTO of the entire US. Last fall, Park’s stress levels increased dramatically when he caught the hot-potato task of rebooting the disastrously dysfunctional HealthCare.gov website. But he was also given special emergency dispensation to ignore all the usual government IT procedures and strictures, permission that he used to pull together a so-called Ad Hoc team of Silicon Valley talent. The team ultimately rebooted the site and in the process provided a potential blueprint for reform. What if Park could duplicate this tech surge, creating similar squads of Silicon Valley types, parachuting them into bureaucracies to fix pressing tech problems? Could they actually clear the way for a golden era of gov-tech, where transformative apps were as likely to come from DC as they were from San Francisco or Mountain View, and people loved to use federal services as much as Googling and buying products on Amazon?
“We have a window of opportunity—within this government, under this president—to make a huge difference,” Park says.
Park wants to move government IT into the open source, cloud-based, rapid-iteration environment that is second nature to the crowd considering his pitch tonight. The president has given reformers like him leave, he told them, “to blow
everything the fuck up and make it radically better.” This means taking on big-pocketed federal contractors, risk-averse bureaucrats, and politicians who may rail at overruns but thrive on contributions from those benefiting from the waste. It also will require streamlined regulations from both the executive and legislative branches. But instead of picking fights, Park wants to win by showing potential foes the undeniable superiority of a modern approach. He needs these coders to make it happen, to form what he calls a
Star Wars-style Rebel Alliance, a network of digital special forces teams. He can’t lure them with stock options, but he does offer a compelling opportunity: a chance to serve their country and improve the lives millions of their fellow citizens.
“We’re looking for the best people on the planet,” he said. “We have a window of opportunity—right the fuck now—within this government, under this president, to make a huge difference.
“Drop everything,” he told them, “and help the United States of America!”
It’s ironic that the greatest opportunity for government IT in a generation had its roots in the greatest and most public government IT disaster of all time. When HealthCare.gov debuted on October 1, 2013, the site didn’t work in so many ways that it’s impossible to list them all. It took eight seconds to respond to a mouse click. It miscategorized minors in Louisiana as incarcerated prisoners and thus ineligible for health care. It crashed so often that of the millions who came to the site, virtually no one was able to complete an application. The failure threatened not only the controversial Affordable Care Act but the legacy of the Obama administration.
Park had not been involved in creating the site—the contractors working for CMS, the sub-agency of Health and Human Services charged with building HealthCare.gov, had never signaled that anything was amiss. But as the designated fixer, he realized that he would need outsiders, engineers schooled in a different style of computing than those who botched the project. Fortunately, he had a secret weapon.
In August 2012 Park had introduced a program called Presidential Innovation Fellows, drawing experienced tech people into government for six-month stints assigned to specific projects in the White House. Park envisioned the PIF program as a way to pepper government with volunteers who would create useful tools whose ultimate value would be proofs-of-concept that things might be done better using modern practices. (For instance, one focus was creating a Blue Button for users to instantly get health care records from Medicare, Veterans Affairs, and other government agencies.) Then, as he says, “magic” might happen, broadening the outlook of government lifers. The PIFs were, in effect, Trojan horses with coding skills and presidential endorsements. Some 700 people applied for the 18 spots in the first round, and the winners came from places like Google, TurboTax, and an assortment of startups. So when it came time to rescue HealthCare.gov, Park had a cadre of in-house tech adepts ready and willing to serve.
Mikey Dickerson Michael George
Key members of the HeathCare.gov tech surge included Ryan Panchadsaram, a PIF who had helped create Blue Button and had recently become deputy CTO to Park; Greg Gershman, a former IT contractor and recent PIF; Paul Smith, a tech entrepreneur who had worked on Obama’s reelection; Jini Kim, a former Google product manager who had just begun her own health care tech company; and Mikey Dickerson, a veteran Google site reliability engineer, who became somewhat of the group’s leader by virtue of running the daily stand-up meetings that organized what had once been chaos. By the time they started, HealthCare.gov had cost taxpayers over $300 million. “Your average team of 10 in San Francisco would be thrilled to get $3 million,” one of the Ad Hoc members says, voicing a sentiment the group held universally.
They got a firsthand look at why government IT is so expensive—and dismal. Even as the site collapsed, the dozens of contracted developers toiling on the site seemed weirdly detached, methodically producing code for arcane new features. Only later did the members of the tech surge realize that this behavior had been preordained by the way the federal government wrote its contracts. Perhaps to satisfy as many of the handful of big contractors as possible, contractors were hired to work only on discrete pieces of the puzzle—the features of the website, the security protocols, the accessibility requirements, and hundreds of other details. But none of the contracts dealt with overall performance issues, like the speed at which the website should respond to a user’s input. No contractor was responsible for even making sure the site was operational.
Gaps like this were standard practice, an artifact of a procurement system routinely manipulated by contractors and protected by their political allies. “As soon as you do one of these projects you break it up into pieces and hire five contractors to work on it,” Dickerson said soon after completing his work on the rescue. “They aren’t helping each other at all, nobody cares about the delivery of the project, everybody only cares about who’s going to get awarded the next contract. So everything they do is meant to make the other contractors look bad.” That means, when a setback arises, more energy is spent avoiding blame than accepting responsibility and fixing the problem.
More broadly, the government’s entire approach to technology was top-down and inflexible—the precise opposite of the so-called agile methodology that has driven Silicon Valley innovation for the better part of a decade. “In most things people rely on software that’s highly flexible and built to incorporate the most current best practices,” says Mina Hsiang, a designer who spent two months helping fix HealthCare.gov. “But the government asks for it to be built in the same way we make F-16s—plan it all out in advance, charge hundreds of millions of dollars.”
“If I hear one more person tell me we can’t use New Relic,” Dickerson announced at one meeting, “I’ll punch him in the face.”
In part, that’s because the agile system is based on an iterate-and-improve strategy that tolerates small failures in the service of an ultimately solid result. That’s a terrifying prospect to government bureaucrats, for whom even the tiniest temporary setback could provide ammunition to political foes. What’s more, federal services must address factors like accessibility, security, and adherence to sometimes obsolete standards—a complicated morass of requirements that favors a top-down centralized approach to planning.
Still, with the fate of the Affordable Care Act on the line, it was clear that HealthCare.gov had to break out of such standard operating procedure. The Ad Hoc members had leeway to push the established contractors they worked with to adopt modern practices. “The message the contractors got loud and clear from the White House is that no amount of blame-shifting is going to spare you from this tornado,” Ad Hoc member Gershman says. “Your only way out is to get your act together and make the site work.”
Even with its presidential imprimatur, the Ad Hoc team sometimes struggled to implement the newer approach. Contractor employees, for example, balked at taking up New Relic, a software product that monitors a server or application’s performance in real time. (Previously, the engineers had to rely on human testers to tell them whether the system was running slow or working poorly). After one such encounter, Dickerson blew up. “If I hear one more person tell me we can’t use New Relic,” he announced at one meeting, “I’ll punch him in the face.”
Eventually, the outsiders won the grudging respect of the lifers, as they brought order to the site through careful monitoring, automated testing, and a collaborative, methodical, common-sense approach to bug-fixing. Though the Ad Hoc engineers didn’t exactly transform a lumbering beast into a gazelle, they successfully patched HealthCare.gov to the point where it could at least perform its mission. In April, President Obama told the nation that despite its woeful debut, HealthCare.gov ultimately was key to the government exceeding its goal of 8 million people signed up online for new health insurance policies. Few in the know could dispute a secondary outcome of the tech surge: It proved that even in complicated government projects, Silicon Valley’s agile style really worked. And it gave Todd Park a perfect opportunity to pounce.
The scene is pure Silicon Valley: a group of young, casually dressed engineers leave their Mac computers atop trestle tables and gather in a corner of a loftlike space for the daily stand-up meeting. They’re introduced to just-hired colleagues, report the progress of their projects, and agree on what to do in the next few days before returning to their screens and hacking away.
Welcome to 18F, a digital SWAT operation within the General Services Administration. It’s a kind of extension of the HealthCare.gov squad, a group of tech adepts who can be tapped by various federal agencies to create new products—like websites or tools for easier access to government data.
The inspiration for 18F came from a wildly successful program in the UK called the Government Digital Service. Code for America founder Jennifer Pahlka, who spent a year starting in June 2013 as Park’s deputy CTO for innovation, had been blown away by how the GDS had modernized Britain’s IT process, and began proselytizing for something similar in the US. (She was actually visiting the GDS when Park called her to lure her to do a stint of government service.) The White House wondered how something like that could happen in the US and found a kindred spirit in the GSA’s administrator, Dan Tangherlini. The GSA head is a big fan of the Silicon Valley mindset; not long after taking office in the summer of 2013, he decided to do away with his building’s offices in favor of an open-seating scheme that lets employees book their spots online. So when he heard from former PIF Greg Godbout that the White House wanted to set up a small digital task force within the GSA, Tangherlini welcomed the prospect.
“We’re trying to take down the walls, trying to squish hierarchy and empower people,” Tangherlini says about the GSA and 18F.
You can already see that 18F—named for the street corner where its offices are located—isn’t your usual government agency. There are those Macs, first of all. And there’s also the fact that Godbout’s team found a loophole in federal hiring practices that allows them to streamline the byzantine process by 70 percent. “We’re trying to take down the walls,” Tangherlini says, “trying to squish hierarchy and empower people.”
Currently, 18F goes into agencies only when invited, and its successes have been small-scale. Its showcase is Not Alone, a web service launched inApril to support the president’s initiative on campus assaults—it gives students and university administrators easy access to resources about sexual abuse. “We did Not Alone from zero to a launched website in a month,” 18F engineer Aaron Snow says. As of August, over 50,000 people have visited the site.
While 18F provides a good testbed for Park’s philosophy, Pahlka and others felt that a more ambitious effort was needed. The UK’s GDS was structured so that it reported to directly to the highest levels of government. To really transform the bureaucracy, a true American counterpart should be close to the White House, with the presidential clout that implied. And so, in August, the White House announced the formation of a new task force called the U. S. Digital Service. It’s similar in some ways to 18F, except instead of simply building new products, itwill be focused on fixing broken systems and processes across the federal government. “Think about this as kind of a world-class group of technical experts,” federal CIO Steve VanRoekel says, “the archetype being the people we brought on to turn around HealthCare.gov.”
Park’s choice to head the Digital Service was Ad Hoc veteran Dickerson, whose experience in DC converted him into a passionate reform advocate. (He’s been urging tech executives to promote sabbaticals for government service.) His experience in the HeathCare.gov rescue, his expertise as a Google infrastructure engineer, and his willingness to threaten sullen bureaucrats with a punch in the face, make him an ideal selection. In a White House video made on his first day on the job, Dickerson noted that the number one question people asked him was whether he would dress the way he did at his previous job (i.e., like a Googler, with all that sartorially implies). The answer is pretty much yes—no jacket required, jeans OK, with a slight compromise of wearing a shirt instead of a T-shirt with some geeky insignia. Being able to retain his casual garb, he notes, is non trivial. “That’s just the quickest shorthand way of asking, ‘Is this just the same old business as usual, or are they actually going to listen?” So the reform effort has at least passed that test.
Upgrading the entire bureaucracy will be more difficult, especially since much of it retains a deep-seated aversion to the kind of processes Park hopes to promote. So part of the White House strategy involves what Park calls myth-busting. For example, he found that people in the federal IT community widely believed that government regulations forbade agile development processes. Park tasked his minions with poring over the Federal Acquisition Regulations—the Talmud of what can and can’t be done—to identify any such directive. There was none. “Actually, you can procure for agile,” he says. So his office created a TechFAR Handbook that outlines how to speedily purchase modern digital tools.
One of Martin’s biggest priorities at Veterans Affairs is creating a disability evaluation tool for vets applying for benefits.
Already, the efforts have begun to show results. Even while he was rescuing HealthCare.gov, Park brought in a new team to build Marketplace 2.0, the next generation of the site. In the process, they accomplished something even more remarkable: permission to employ the offsite data center of Amazon Web Services. Normally, government IT sites run on federally managed data centers. To add capacity, administrators must file a formal request for more servers, a time-consuming process that is useless to address a sudden surge in usage. Third-party cloud services like Amazon’s can automatically assign more servers in real time, handling even a sudden tsunami of requests. Countless private companies, including Netflix and Pinterest, take advantage of this, but a slew of regulations, largely dealing with privacy and security concerns, stood in the way, and it took months to get an OK. “It was less about security than it was about checking a lot of boxes,” one team member says. But when Americans go back to the website to enroll in 2015 health care plans, some of the site will indeed run on Amazon’s servers, dramatically increasing reliability.
“That was a very huge victory,” says Joey Liaw, a tech entrepreneur who worked on the Marketplace 2.0 team, because the steps taken by the Marketplace group to get authorization can be reused by other teams.
A bigger test is shaping up at the Department of Veterans Affairs, which is facing a shocking backlog in processing disability payments: Claims are supposed to be addressed in under 125 days, but the national average is more than twice that, and hundreds of thousands of former soldiers are stuck in wait-list purgatory. While much of this is due to medical understaffing, a better IT system could ease some of the problems. Or so believes Marina Martin, a former PIF who became the agency’s CTO last year.
The VA is a prime example of antiquated government computing; its main systems run on a 1960s vintage system called MUMPS, a dead digital language that few people under retirement age know how to use. When Martin arrived at the VA, she set up a development system that uses modern languages and tools like Ruby on Rails and Hiroku. “We had it up in a weekend,” she says.
Martin had no staff during her first year, yet she managed to implement a one-stop online Veterans Employment Center matching job-seeking vets with businesses looking for workers. She also replaced the VA’s complicated login system with one that lets employees use their Google or LinkedIn accounts.
One of her biggest priorities is creating a disability evaluation tool. Vets applying for benefits face a laborious, paper-based process in which health care professionals medically evaluate them, and then a VA tester determines whether the condition is service-related. The new tool would use machine learning and big data to automate evaluation in a significant percentage of straightforward cases, freeing resources for the agency to move more quickly on more complicated claims.
To build it, Martin has turned to Nuna Health, a startup headed by the Ad Hoc team’s Kim, and Ad Hoc LLC, another new outfit formed by tech surge colleagues Smith and Gershman. “Our goal is to do government contracting differently,” Gershman says. “Hopefully, the White House can work on IT reform so small companies can bid competitively with the big ones.” Indeed, Park and his PIFs have been chipping away at the process, implementing a system called RPF-EZ, through which certain projects are offered to small tech businesses.
Kim believes that an influx of people like her can help foment the change needed to overthrow the stifling bureaucratic regime that brews disasters like HealthCare.gov. But that depends on the government’s success in luring great people to revive the dead zone of government IT. This won’t be easy. In particular, for the reform effort to become truly widespread, it will need the support of agency heads and legislative changes that grant more flexibility in practices, hiring, and procurement. Park’s strategy depends on something rare in Washington: common sense. He believes that by using modern approaches to solve a few key problem areas, the undeniable superiority of the approach will convince lawmakers, administrators, and the next president to formally remake government IT into a smart, agile, geek-run IT platform. History, however, shows that when it comes to government IT, inertia always comes out on top, common sense be damned.
Park and his minions think that this time, it’s different—provided they don’t squander this moment. He insists that his departure from the CTO post won’t drain momentum from his crusade. While he plans to travel back to DC two or three times a month, he thinks that he can energize the undertaking even more from his new perch in the hotbed of high-tech. “I’ll focus super intensively on recruiting top tech talent, channeling the best ideas of the valley into a smarter government IT effort, and make sure we have the best possible sense of how technology is evolving,” he says.
In other words, Park is still pushing, because the moment for change is, as he himself has noted, “Right. The. Fuck. Now.”
