Profile cover photo
Profile photo
Tab Atkins Jr.
Tab's posts

Post has attachment
Hahaha, this is a pretty interesting summary of my year.

Only stylistic quibble is that it started off with my name, but showed a picture that was of other people (my wife and her classmate).  It should probably try for images that it knows include me, or don't include people at all.

Post has attachment
All the photos from the Google Holiday Party in SF City Hall.  The building was absolutely beautiful, friends are lovely, food and drink was great; a good time was had by all.
2014 Holiday Party
15 Photos - View album

Post has shared content

"There’s a whole category of native apps that could just as easily be described as “artisanal web browsers” (and if someone wants to write a browser extension that replaces every mention of “native app” with “artisanal web browser” that would be just peachy)."

Post has shared content
Elonis v. US is an extremely important court case, and I'm following it very closely. Unfortunately, Sarah Jeong's article at +The Verge gets nearly everything about it wrong. 

It does get the facts right, which can basically be summarized as: Anthony Elonis made a large number of posts on the Internet in which he threatened, graphically and at great length, to rape, torture, and murder his ex-wife and several other women, including an FBI agent. He was arrested, charged with making threatening communications, and promptly convicted. 

His defense, which is at the heart of what's going before the Supreme Court on Monday, can be summarized as: "I was just writing rap lyrics! I had no intention of actually threatening anyone."

This is a kind of bizarre defense in the case, (the so-called "rap lyrics" at issue were, shall we say, quite clear and vivid) but at its heart is an issue which the Court has specifically asked the two sides to brief, and upon which several amici have weighed in as well: should the test for a threatening communication be based on an "objective standard" -- i.e., that a reasonable person would interpret it as a threat -- or a "subjective intent standard," i.e. that the person intended it to be a threat?

This is the point, unfortunately, where the article completely falls apart, because Jeong confuses "reasonable person" standards with "community standard" standards, and "subjective intent" with "mens rea." So let me explain the terms at issue here.

The "reasonable person" standard is a concept in common law, and it's used in cases where there's a question about how a communication should be understood, or how a reasonable person would behave. For example, if you seal and lock a tank and put a sign on it saying  "WARNING: This tank is full of hydrofluoric acid!," a court would say that a reasonable person would have been warned by this, and if they decided to break the lock, open the tank, and bathe in it anyway, then this was not negligence on the part of the tank owner. You can read about this standard here:

A "subjective intent" standard, as the Court is considering in this case, would require that, to prove that something is a threat, the prosecutor would need to prove that the person intended to threaten someone. It's very difficult to prove what a person subjectively felt, unless they happened to tell their friends later how their purpose was to threaten someone, and as a result this would make it nearly impossible to convict anyone under this statute.

Jeong's misinterpretation comes from confusing these with two similar, but very different, ideas. She fears the objective test because it might let anything anyone says be prosecuted as a threat, because "when it comes to the Internet, where context or tone may be more difficult to perceive, this objective standard has obvious drawbacks: is the "reasonable person" going to be a teenager who plays League of Legends or a grandfather posting on a fly fishing forum?" 

What Jeong is thinking of here is a different standard -- the "community standard" which was historically used in obscenity laws, and which has been a notorious legal disaster for the reasons she describes. There is no coherent community standard. The "reasonable person" of a reasonable person test, however, isn't meant to be an average person of any community at all: it "is a composite of a relevant community's judgment as to how a typical member of said community should behave in situations that might pose a threat of harm (through action or inaction) to the public." (Emphasis mine)

Likewise, Jeong says that "it’s hard to see a downside" if the Court imposes a subjective intent requirement. To explain why this isn't true, let me tell you a bit more about subjective requirements, which are really part of the idea of mens rea, the "guilty mind" which is required to prove some crimes. (And if you want to know about it, the absolute best explanation ever is at

Basically, some crimes have in them a requirement that, to prove the crime, you need to prove some level of intent. Mens rea comes in five levels:

None: You had no evil intent at all. Maybe someone died by an accident you couldn't control. 
Negligence: You failed to take a precaution which a "reasonable person" (there's that standard again!) would have taken.
Recklessness: You consciously chose to disregard a substantial risk your actions would create. 
Knowledge: You knew that there was a substantial likelihood that your actions would lead to that outcome.
Intent: You committed that action with the intent of that outcome.

For homicide, for example, these five levels would correspond to an accident (no crime), negligent homicide, manslaughter, second-degree murder, and first-degree murder. (With the names and details being a bit different in different places) 

Subjective versus objective tests boil down to how mens rea is determined. In an "objective test," we use the reasonable person standard; for example, if you fire an automatic weapon into a highway, we assume that a reasonable person would know that the likely consequence of such an action would be that people would be killed, and so it would meet the "knowing" bar, unless the defense could somehow prove that the person had no idea what they were doing at the time. (e.g., they had good reason to believe they were firing blanks) (See )

In a "subjective test," you have to prove what that particular person had in mind at the moment that they did it, which basically means "you need a confession." It's very rare in law to have a purely subjective test for anything, because of this; the only case that immediately pops to my mind is conspiracy charges in cases where the actual crime hasn't happened yet (e.g., you catch them before they rob the bank) and so you have a very high bar to prove that they were up to no good. In this sort of case, you tend to need fairly overwhelming evidence: even finding the conspirators with a bunch of guns, detailed maps of the bank, and the schedules of the guards may not be enough.

This is why most laws that rely on something subjective (e.g., murder) actually use a hybrid test, where certain actions can be "imputed" to imply intent. To take a good example from , if Alice goes over to Bob's house, pours gasoline on his door and lights it, and he dies in the resulting fire, even if she claims that her intent was merely to drive him out of the neighborhood -- it's just arson, officer, honest! -- the consequence was so foreseeable (again, to a "reasonable person!") that under most laws, the court could simply impute that yes, she really did mean to kill him.

If the Court were to adopt the most extreme version of the subjective intent requirement, which is what Elonis is asking them to do, then you couldn't do this, which means that you would basically need a confession in order to convict: no matter how obvious the threat was (and Elonis' threats were pretty goddamned obvious), if they say "I didn't mean to do it," there's no crime.

This is what Jeong misunderstands about a subjective intent requirement: this is not merely something which would make threats like an ordinary crime, but something which would make threats virtually impossible to prosecute. 

The reason this case is so important is that harassment and threats are an ever-growing problem in our world, one which has become especially severe in the past few years, and they are already very difficult to prosecute for practical reasons, often involving complex jurisdictional issues. If the court rules that an objective standard is appropriate, then threat law stays roughly as it is: you need to make a statement which would make "a reasonable person" fear for their safety. If they rule for a subjective standard, then anything is fair game.

So in response to Jeong's key assertion: yes, there is a downside to a subjective standard, namely that threats -- even the sort of threats that drive people out of their homes and into hiding, threats followed up with a horse's head at the foot of their bed -- become effectively legal. An objective standard doesn't mean that anything you say on the Internet becomes a crime: it's subject to the exact same standard as it is today, the "true threat" doctrine, which (as anyone who's ever worked with stalking victims or tried to prosecute it can tell you) is already an extremely high bar to reach.

h/t +Ralf Haring for pointing me at the article.

Post has shared content
Four stories about work:

One: Immediately out of college, I took a job working as the receptionist at a homeless shelter. The social worker quit, and I added those duties to my own, despite not knowing how to execute them. Then the executive director quit, and I took his job.

There was no duty pager for staff. If there were incidents, and there were, the police responded without staff support. I would only find out about the response the next day, from the on-site manager -- who, in general, was a trusted client. The trusted client was cooking meth.

I found a new one. But I took the phone as well. I was the only employee, and so I was the call rotation.

Clients would show up drunk. My phone would ring. There would be fights. My phone would ring. The sheriff would show up to serve summonses. My phone would ring. And although my workday wasn't usually any longer than ten, eleven hours, several times a week, I would be rousted out of bed to go handle some emergency or another. 

I was bad at my job. This was in part for the usual reasons; viz., that I had no training. But almost every morning, at 4 AM, like clockwork, I would wake up shuddering, and I would drag myself to the bathroom, and I would vomit out of anxiety. I didn't sleep. And so basic things like grant billing and meeting scheduling began to fall apart as fatigue-driven errors crowded in on every side.

Eventually, I left, having kept the organization from imploding. But I did very little other than that: I held on white-knuckled, and managed a hard landing for an organization that could have cratered. But I did an awful job. Even my emails from that era -- some of which I've kept -- are riddled with typos. 

Two: Insofar as I have been successful, I owe my success to that job. Everywhere I've gone, from that point forward, the very first thing written on my resume has been "executive director." It's what took me to ATVP, and what got me into law school, and after law school, this set of things got me my clerkship, and after my clerkship, it convinced people that I was hard-working enough to cut it in the job I got at Google.

And so I've landed here.

Earlier, +David desJardins accused me of being contemptuous of startup culture. I've never worked at a startup. But I've been in a similar place: working insane hours, never being entirely off the clock, cutting my own salary to keep the money from running out before the task was complete. It made me terrible at my job. The more I scrambled to keep the money from running out, the less I deserved it.

There are strong personal incentives to the sort of work which depletes all of your resources; the sort of work that renders you incompetent. I benefited from them. I still benefit from them. But I can still name the people I harmed, and can still trace, in detail, the things I was terrible at.

In the aftermath of negligence, it's tempting to offer yourself the cold comfort that there was no better way. As hard as it is to make decision that makes others suffer, it's tougher still to admit that that decision was unnecessary -- that you passed up an easy win because the incentives were wrong.

Three: We have evidence from medicine that there are probably some easy wins across the board, and across all kinds of high-performing jobs. Since the 1960s, doctors have insisted that working residents 100 hours a week was necessary for training, and that failing to do so would cripple medical productivity. 

The evidence, such as it was, looked fine: doctors unwilling to work the hours washed out, and doctors who worked longer hours were more successful. Not only that, they treated more patients. This seems adequate to empirically demonstrate that working more hours is better.

Of course, when you examine the rate of medical error, you find that overworked doctors are terrible at their jobs. As you would expect, the rate of medical error among people working more than sleeping is tremendous.

So were we rewarding terrible doctors? 

No. Because there's a confound: the same people willing to work hundred-hour weeks are more dedicated to their job, and otherwise better workers. But we had somehow managed to engineer a system which rewarded the production of low-quality work by high-quality workers, and which -- even more perversely -- used the production of low-quality work as a proxy for high-quality status. 

Unless software development is utterly unlike any other human activity -- and the mythical man-month suggests that it's not -- then startup founders are simply engaged in the same status-demonstration game that doctors were, and there's no reason to take them seriously. Conceding that they'd been wasting their time (and others' time) raises troubling questions that they'd rather not answer.

Four: A coda. In medicine, the policy changed. Often, nothing improved: the incentives didn't change.

Doctors' hours were reduced from thirty-six hour shifts to sixteen-hour shifts. During their previous shifts, doctors slept, but they slept on the job. Now, working sixteen hours at a time, doctors never slept. The fatigue issues that were the core of the problem? They didn't go away. Doctors still needed to signal that they were working hard; the incentives to ignore the policy and exceed its bounds were still there.

Caring about incentives and caring about outcomes are often deeply at odds. Even knowing what I know now about work, I wouldn't change what I did: it was absolutely, unequivocally in my best interest. But the evidence is good that exhaustion makes us vulnerable to catastrophic failure, and that the incentives within subcultures of workaholism are deeply broken.

Post has attachment
Treated myself to a manicure today. Switched it up a bit, went with a brassy red for Thanksgiving.

Post has shared content
Best practices for publishing conference videos based on my experience with the +JSConf channel.

- While +Vimeo  may give you more "street-credibility" than +YouTube  the latter is way better to build a sustainable audience (over short term social network traffic). It also has no limit in upload traffic per day, which is important when you want to upload 2-4 talks in HD per day or even a whole conference.
- You probably want to do live encoding/production of conference videos as it ends up being much less work, but you DO NOT just want to dump all videos at once, the day after the conference. This just means that your YT subscribers will get 1 notification for 40 videos and they will watch one of them. The same with posting on social networks: People see all those posts and then watch 1-2 talks. Instead spread things out to 1-4 releases per day.
- Use a YT channel that is owned by a +Google+ page, since they can have multiple managers.
- License: Unfortunately Youtube is inflexible on the creative commons licenses they support. We've seen our videos ripped off by commercial sites that make people pay to see content that is free on YT and we don't want that. We thus use the default license with this extra blurb at the end of the description: "License: For reuse of this video under a more permissive license please get in touch with us. The speakers retain the copyright for their performances."
- Manually pick the best thumbnail in YT.
- Don't use the YT editor as it opts the video into showing ads. I've seen this used to i.e. add an opening sequence: Better do it offline.
- Don't make a YT channel per year or event. Use one channel for all conferences to build a long term audience. This is why all JSConf videos are under The URL sucks but it can't be changed.
- Video title format: Talk title first or speaker name first. Mention the conference like "JSConf US 2014" last. Don't mention the track.
- Copy the description of the talk into the video description.
- If you also publish the videos on your site, paste a link to that page into the YT description. This makes proper attribution much more likely when other sites syndicate your content. If you don't publish on your own site, just paste a link to the homepage.
- When publishing the videos on social networks, @ or + mention the speakers, so they can retweet. This adds their audience to your audience. Unfortunately this means you should not use the build in YT publishing tools as they want to post the same message to all networks.
- Try to find a link to the speaker's slides and paste it into the YT description.

All the next points are hard to get right with live encoding, so only do it if you can afford post production:
- Luxury bonus: Record a separate audience sound signal and mix it in later for laughs and clapping.
- Luxury bonus: Zoom in on slides if the slides are more important than the speaker.

Let me know in the comments with more suggestions for best practices and I will update the post.

Post has attachment
October Formal Friday means black and orange and black and orange.

Post has attachment
A highly abbreviated version of my trip to New Orleans with my wife, who was attending a veterinary conference.  (I was just there to eat food and look pretty.)

Post has attachment
We had a wonderful time tonight celebrating our 9th anniversary.  Ate at "La Fondue" in Saratoga, which I can strongly recommend now.
Animated Photo
Wait while more posts are being loaded