An Open Letter to Duo

A green cartoon owl of exceptional malevolence---who is lying to you too.

Good Morning:

Dear Duo:

I want you to stop lying to me.

I know you think it’s cute. You think you’re just creating incentives for me to learn other languages. You’re just encouraging me to practice. But you, little green owl, are part of the problem. You are the reason why people think that meaning to do something or wishing you had done that thing is the same as actually having done it. You are the reason why people think they have accomplished things because in their hearts, they did them. You’re the reason we are a flabby society built out of validation and grade inflation. You are the cartoon manifestation of mediocrity posing as accomplishment. And I want you to stop—at least as regards me.

Let me back up and review the record. You are the mascot of an app called Duolingo, which teaches a great many people a great many languages. It’s a good program. It’s never going to make a person fluent in a foreign language, but it’s a very good way to make a person familiar with a language, literate in it, and for that person to get to a certain level of proficiency with grammar and vocabulary. I currently study six languages on Duolingo, so you and I have gotten to know each other rather well.

Your specific job is to nag and harangue people into studying every day, a job I actually respect. I know that people hate you for making them feel guilty. They misunderstand your role.

Key to that job is two basic concepts. The first is gamification. Duolingo operates like a set of games. You do lessons. You earn gems and points. You can buy things with the gems. There are competitive leagues. All of it is designed to give the user the incentive to practice more. It’s clever. It’s smart. And it actually works—even though the gamification is really silly. I have no beef with any of this, save that some of the characters are really dumb.

The second concept is the “streak.” You count every sequential day every user practices, and you dangle the threat of our streaks ending as part of the incentive to do at least one lesson every day. Again, it’s very clever. Nobody wants to lose a years-long streak, particularly as the app has this social component so that my “friends” can see my streak—and see it end. What’s more, you have created this powerful idea of friend streaks, wherein users and their social contacts practice in streaks together and thus have powerful incentives not to let their partners down by missing a day. Again, it’s fine. It works.

Here’s my problem—and both you and everyone else reading this who knows your has already discerned where I’m going with this: You’re lying to people about their streaks.

You have developed a number of different strategies for not letting people’s streaks end when they do, in fact, miss a day. Such that at this point, all Duolingo streaks are complete frauds. Nobody’s means anything.

You are far worse than, say, Ivy League grade inflation. All of the children are not just above average; they have years-long streaks going. And when they miss a day—as they, in fact, do—you cover for them. In fact, you organize the cover-up.

A few months ago, my long-running streak came to its natural end. I got on an airplane and traveled to a different timezone. The day thus vanished pretty fast for me. And I missed a day. It happens. I am not perfect. I can face that about myself.

You immediately slapped a “streak freeze” on my streak. What is a streak freeze? It’s a kind of mulligan which you created so that when people don’t, in fact, practice every day, they can pretend that they did. What’s more, you sell these streak freezes in exchange for the “gems” earned by people who fall for your gamification stunt. People also “earn” streak freezes as prizes over time. Practice X number of days, and you’ll earn a few streak freezes to cover days you might miss. These streak freezes accrue quickly, so that it quickly becomes hard to fail your way out of a streak.

But dude, I am your Martin Luther, calling you out for selling indulgences.

In fact, not only are selling indulgences, you are giving them away for free—forcing them on people who actually want to confess their sins and do their penance.

When my streak ended, I just wanted to start over, from Day 1, and have an #HonestStreak. You refused. For more than a week, you kept insisting that streak freezes had preserved my streak. You kept pretending that I had practiced hundreds of days in a row. I had to take a lengthy break from practicing my various languages merely to get the clock to reset to zero. I vowed that I would not allow you to lie to me any more.

I started a new streak and texted the estimable

Eve gaumond

that this one would be an #HonestStreak. My first #HonestStreak lasted 48 days. It took me more than a week of non-practice to reset it before you would acknowledge that I wasn’t still on it.

My next #HonestStreak lasted 16 days before I got wrapped up in an intense period of editing and running

Lawfare

. I’m not ashamed of it. Lindsay Halligan was rage-texting one of my reporters. I write a column two or three times a week. I am trying to follow a great many cases. And I seem to have become a non-profit executive who has to fund a bunch of people’s jobs. I can face the fact that I’m going to miss a day every now and then of concurrent study of French, German, Italian, Spanish, Ukrainian, and Hebrew. Really. My ego isn’t so fragile that I need you to lie to me about this. On Sunday, I missed a day.

There. I said it. It wasn’t even hard.

On Monday morning, you informed me that I had protected my streak, though you and I both know that I had done no such thing. You had protected my streak, because you are a liar. I texted Eve the evidence of your corrupt scheme:

Then you tried to get me to join the scheme:

You were persistent. You even lied to me about whether Monday was my “last chance” to help you cheat on my behalf. I knew there would be other chances:

Even the next day, you were still at it—offering to protect my barely two-week long streak with two-days worth of indulgences:

Again with the imputing of your own wrongdoing to me.

Even on Wednesday, you were pushing me to lie to the world:

It took until yesterday, after four full days of intentionally scotching my streak, for you to acknowledge that it was gone. Think about it: That means that as much as 25 percent of my streak (four days out of 16) would have been a fraud if I had merely acceded to your seduction. Think how many days of non-practice worldwide you are coaxing people into describing as streaks:

Duo, dude, yingele, green motherfucker that you are: I want an #HonestStreak option. It’s not too much to ask. I want a means of knowing how many days in a row I have actually practiced. No games. No indulgences. No tricks. (I will tolerate an exception for crossing the International Date Line, but that’s it). I don’t want credit for stuff I didn’t do. And I don’t want to have to take lengthy breaks from studying languages in order to force you to count accurately.

I’m putting you on notice about this. I am a good customer. I pay good money for your service. I’m not even asking that you stop lying to other people. We live in a world of disinformation, and a lot of people seem to get off on being lied to.

But I am not one of those people.

All I’m asking is that you stop lying to me about my own performance. Among other things, it hurts my feelings that you think my ego is so fragile that I would need that from a cartoon character owl.

Yours most sincerely,

/b

---

In honor of Sandwich Guy’s acquittal, here’s a sign

EJ Wittes

found on a lamppost the other day:

---

Yesterday on #DogShirtTV, the estimable Katherine Pompilio and I celebrated our latest video, and the estimable Mariia Hlyten had some thoughts on internet outreach:

---

Yesterday On Lawfare

Compiled by the estimable Isabel Arroyo

Sandwich Guy, Thrower of Hoagie—Or Hero?

Molly Roberts examines the federal proceedings against Sean Charles Dunn, who was acquitted of misdemeanor assault today after throwing a sandwich at a Customs and Border Patrol officer over the summer.

The reason for this intensity becomes clear: Both sides want to make an example of Sandwich Guy. The government, as a menace to society against whom that society must for the sake of law and order impose punishment, and the defense, as a dissident unfairly targeted for his deeply held ideological convictions. The trial ends in an acquittal. But the proceedings that lead to this verdict feel like a strange sort of performance art—highly amusing and highly menacing at the same time.

Are Federal Officials Immune From State Prosecution?

Bryna Godar analyzes whether federal officials are immune from state prosecution, explaining that courts have historically treated state-level criminal prosecutions as a limited but important check on federal power. Godar examines the two-pronged test that governs prosecuting federal actors and criticizes the Justice Department’s threatened tit-for-tat prosecutions against state and local officials.

This state prosecutorial power does come with important constitutional limits. In cases where federal officials are acting reasonably and within the bounds of their federal duties, federal courts have held that state prosecutions cannot go ahead. The reason stems from the Constitution’s Supremacy Clause: If states were free to prosecute federal officials for lawfully and reasonably carrying out their duties, states could thwart the entire operations of the federal government. Importantly, however, the Supremacy Clause does not categorically immunize officials for actions carried out during the performance of federal duties, as Blanche and Miller have asserted. Rather, courts conduct a situation-specific inquiry into both the scope of the federal officer’s authorized duties and whether the officer’s actions were reasonable (or “necessary and proper”) in carrying out a federal duty.

State Prosecutions of Federal Agents and the Presidential Pardon Power

Aaron S.J. Zelinsky, John Connolly, and David Reiser break down the procedure for prosecuting federal actors under state criminal law, explaining that such prosecutions would be removed to federal courts but remain beyond the constitutional pardon power of the president.

In that circumstance, state prosecutors would prosecute the case in federal court, before a federal judge, with a jury selected under procedures for the federal district. The federal court would apply state substantive law and federal procedural law, and any appeal from a prosecution for California offenses would take place in the U.S. Court of Appeals for the Ninth Circuit. Importantly, even though any resulting conviction would occur before a federal judge, the president would be unable to pardon any defendant, since the convictions (regardless of forum) would still be for state offenses and thus beyond his constitutional pardon power. Any conviction would still be subject to the pardon power of the relevant state authority.

Podcasts

On Lawfare Daily, Tyler McBrien sits down with Joseph Kellner to discuss Kellner’s new book “The Spirit of Socialism: Culture and Belief at the Soviet Collapse.” The two discuss how shifting senses of epistemic authority, cultural identity, and history informed post-Soviet civilians’ search for spiritual meaning, the belief systems that gained prominence, and the relationship between the fall of a political superpower and the decline of shared meaning.

---

Today’s #BeastOfTheDay is a horse in belly-rub position:

Video Source

---

Tell Me Something Interesting

A couple days ago, I—

EJ Wittes

—discovered a new paper on AI language capabilities with the breathtakingly nerdy title of “One ruler to measure them all: Benchmarking multilingual long-context language models.” Tolkien references aside, the paper itself is fascinating, and it’s sent me off on several days of linguistic research of my own.

Essentially, the researchers compared different AIs’ ability to process text in various languages by giving various LLMs a series of identical language processing tasks in 26 different languages. The researchers expected that the LLMs would perform best in English, because that’s the language in which the models have the most training material. But no, English isn’t even in the top five. Want to guess what’s number one?