How a team of young people helped rebuild healthcare.gov
theatlantic.com
theatlantic.com
"The MPL team had three great technical accomplishments over its 16-
month-long life. First, it served as a crack team which understood the
infrastructure of the site and could resolve small issues as they arose.
Second, it built an insurance application, called App2, which signed up
new users in less than half the time of the original app. Finally, it
replaced the website’s crashy login system with a functional (and much
less expensive) one of its own design."
"It did most of this while living together in an unremarkable McMansion in suburban
Maryland."
"So the team that started by performing bug fixes on a sprawling, struggling mass
of code ended by writing critical, efficient infrastructure for the government.
Yet what the MPL team accomplished philosophically may be even more important:
It helped teach government bureaucrats how to think about building websites in 2015."
edit: stephengillie called it in his "submarine" comment below (see: http://www.paulgraham.com/submarine.html)
I think this is true. Governments and corporate's are starting to see that you don't need to go with GenericAdAgency12 anymore to get shit-hot, well-made websites going. This project was a textbook example of this.
The problem is the massive, unbelievable amount of red tape to get things done. Software stacks and browsers have to get approval and re-approvals, a years-long process. Code has to be written in something stable, not something trendy. The programmers spend more time writing documents and attending meetings than coding. There are meetings about meetings and the layers of management are infinitely recursive (managers for the managers).
Start-up culture and tech trends just don't work at the scale of government, for much the same reasons massive corporations move glacially. Just add even more layers of management, congress, civil service employees that cannot be fired, and tax dollars to the mix.
I enjoyed my time working in government web, but "youth" isn't really an answer.
> Start-up culture and tech trends just don't work at the scale of government, for much the same reasons massive corporations move glacially.
This is good example of an old truth that needs some unlearning. Old corporations move glacially. Some new ones work differently, though.
Government can also work differently. The folks at 18F and the US Digital Service are demonstrating that in ways both large and small. Indeed, when I was looking at a USDS job, a White House advisor told me plainly that agile, iterative approaches were clearly better for government because they were much more effective at risk reduction than producing piles of documents ever could be.
Is there a reason you didn't take the USDS job?
"how a team of immigrants helped bring in a strawberry harvest"
The success of this team is based on a relationship between two things: 1) eagerness to prove one self 2) willingness to suffer pain in order to prove one self
Having worked with a number of them, my experience has been that building things from scratch in big bureaucracies (that aren't software companies) is basically impossible. To much group think inevitably destroys the scoping process.
The teams that have impressed me in big bureaucracies tend to operate in the following manner: -take/buy something off the shelf that meets 60% of the requirements -use a team of seasoned project managers and engineers who have seen the darkest gloomiest bloodiest field of battle to incrementally patch and beat that software into the inflexible mold of their bureaucracy
This strategy usually works surprisingly well, but it is like comparing steering a slow moving container ship to a speed boat. It's part of the reason that the agile SaaS market is so successful and profitable.
I first went to work with an organization where a planning meeting was discussing a new feature. Because of my startup culture experience I asked a PM after the meeting "this sounds cool should I whip up a prototype tonight". The project manager laughed and said "this feature set is on the roadmap for 2014 (two years out at the time), we wanted your input but you should be focusing on your existing ticket pipeline". When you consider the scope of the organization the things that they were doing with their core platform was pretty incredible.
That being said it drove me personally nuts, and I quit.
Then some crack team (Accenture or "young people") comes in, gets permission to make the necessary changes and gets all the credit.
Mikey was the first SRE at the scene of healthcare.gov and is now running the new US Digital Service. He works with the folks mentioned in this article and lots of others trying to fix how the government writes software.
Wow, 30 milliseconds. Subtract out the RTT, and you have app request times on the order of 5ms. Outstanding. And impossible.
For an example of scale, 10,000 nodes with 96GB each costs less than $30M dollars. The government spent billions on the healthcare.gov project.
Having been subjected to the government's definition of "agile" before[1], I can tell you that nearly everyone who was at least barely competent technically[2] thought the processes were asinine. The problem is that nobody could convince the government customers and non-technical management of this. The government in particular was bad, as their reps (GS and contractor) often came in with a dictatorial, know-it-all attitude about the problems they were hiring us to solve.
This team managed to do what they did because they were outsiders brought in specifically to fix a high-profile fuck-up. I'm willing to bet that most of what they did was suggested at some point by people who worked on the broken system, but it was shot down. The success of the second effort was primarily political.
[1] The Gantt chart bit brought back memories that made me want to throw things.
[2] Despite stereotypes to the contrary, most developers and engineers working at government contractors are at least barely competent. Contractors with mature engineering orgs (Raytheon, Boeing, Lockheed, etc) are better as they tend to force the fuck-ups out pretty quickly.
This is true. But having recently talked with a number of people at 18F and the USDS, I think the enormous scale of the high-profile fuckup is serving as a great example to persuade people that things should be done differently.
For example, note that the government is trying to change the purchasing processes that drive heavyweight methods and delivery of monoliths:
https://18f.gsa.gov/2015/01/08/creating-a-federal-marketplac...
Having said that, there is one three-letter monolith that commands the plurality of the Federal budget and shows absolutely no inclinations to change. They have resisted intrusion by USDS and 18F thus far, and I can tell you from direct experience that their information systems and policies are atrocious. Hopefully 18F, USDS, and similar efforts can eventually penetrate the Five-Sided Playpen, but it is going to be a long, hard slog.
For the first time in decades the entire country experienced exactly the government they elected and are paying for. And the beauty of it is that it could not be hidden away, spun or silenced.
What's sad is to see an article such as this one pushing the idea that it was fixed...a "Mission Accomplished" message of sorts. The truth of the matter is that dozens of people should have been fired and a number of them should have ended-up in jail.
Nothing has been fixed. Sure, the website is better but the organization is still being run by he same incompetent morons who created the problem in the first place. That's the problem. And, yes, a bunch of young people can help fix it without writing a line of code by using their brains to vote rather than cargo cult and emotion.