One of the problems with those exaggerations is it makes accountability harder. E.g. an agency overcounts something and reporters use the scary number despite being aware it's likely a bad database query. That means the agency can fix the query & claim progress without doing anything substantive.
When I was in traditional newsrooms doing data journalism, there were always editors and reporters who felt every number had to be maxed out to "have a splash" and "get people's attention" -- the statistical equivalent of "if it bleeds it leads."
I spend too much time thinking about how Al Qaeda was born in part out of a left-ish (Nasserist) Egyptian government throwing right wing agitators like Sayyid Qutb into a brutal prison system where they went even further.
You just never lose a step, do you?
Also: "Every hour an American spends staring at a car bumper is an hour they aren't at the dinner table or on the job" seems quite telling in terms of a specific vision of what USians should be doing with their lives.
Defendants who accept plea deals that waive their Fourth Amendment rights can lose their legal recourse for warrantless police searches for years. As of 2020, 96 percent of people living under these waivers in Richmond, Virginia were people of color.
And blessings be upon you for it! I've been tracking a bit of inaccurate information a newsroom I worked for published (knowingly, which is sad). And it's still the first thing GPT, Perplexity, Claude and Gemini consistently use, citing the story, when you ask for that specific statistic.
"Ugh. Are you a Platner apologist?"
We go through this every single time where I correct some misinformation and people jump in and claim that I must support whatever politician or whatever is being discussed. No! I just hate fake news!
We have updated our the #Missouri Vehicle Stops report with 20 more years of data. See long term trends, download the data + join on Census geoid. We'll have plenty more to say about this soon! vsr.recoveredfactory.net
"Where disadvantages are highly correlated and mutually determining, a fixation on causal inference may be misplaced. The thought experiment of manipulating a single condition lacks realism. Social scientists cannot and should not control away that which society has stitched together."
Traffic stops are the most common way Americans interact with law enforcement. My first native social data video as an independent data journalist!
www.youtube.com/shorts/Oboc3...
Our #missouri vehicle stops tool is quickly improving w/ cool improvements in the works -- vsr.recoveredfactory.net
Shiii mucho. Por ejemplo, en inglés, la palabra "curate" es tanto un verbo pero en contextos religiosos especificos es un sustantivo que significa "clérigo". Escribí una serie de pruebas para comprobarlo y realmente se nota la diferencia.
Los glosarios de DeepL son excelentes -- perfecto para mi cliente con muchas palabras con significas religiosas como "El Muy Reverendo" -- les pedí que me crearan una Google Sheet de palabras en los dos idiomas y luego la configuré en DeepL como glosario
Para un proyecto del que hablaré pronto, he estado utilizando mucho la API y los glosarios de DeepL. El sitio es Drupal y tuve que escribir código personalizado para traducir contenido viejo de forma retroactiva.
Sin duda compatibilidad de CMS -- las plataformas de boletines tienen dificultades; Wordpress y Drupal pueden mejor en algunas formas pero 😬😬😬. DeepL también comete errores sutiles, como referirse a mi esposa como mi mujer jaja.
We built Recovered Factory to be multilingual from the jump — English and Spanish, SvelteKit frontend, Kit for email, Paraglide JS for UI strings, DeepL + human editing for translation, Stripe for $. Full write-up with code and hard-earned lessons:
recoveredfactory.net/en/multilingual-how-and-why
The most important thing we learned: forget about parity. You don't need every post mirrored in every language at the same time. Once you let go of that, the whole system gets simpler and the content gets better.
For example, the English and Spanish versions of this post vary significantly.
None of the major platforms — Ghost, Beehiiv, Substack — are really built for publishing in more than one language. Kit was workable enough to build around. McConaughey apparently figured this out too, and Kit loves to put him in their marketing.
Researching multilingual newsletter platforms, I discovered that Matthew McConaughey and I came to the same conclusions: Life's barely long enough to get good at one thing, so be careful what you get good at. And if you need a multilingual newsletter, @kit.com is one of the few workable options.
The big lessons:
* Do invest in production, each language + platform is different
* Don't worry too much about content parity and instead build for flexibility
* Be ready to build around content production tools that haven't caught up with machine translation
Check out github.com/eads/missour... -- we should swap notes and merge whatever you're doing in terms of data enrichment into the pipeline.
Sure, why not? If you want! We definitely like contributions but it's also something I'm happy to handle. But also, we've got a system that's already linking VSR agencies to Census places, so we should definitely talk. That's how we have the Census data on the agency pages now.
Ohhhhhh yes, we really should! If you want an open source contribution to your name, you could even jump on it:
github.com/recoveredfac...
But I can handle in the next few weeks, we have a big backlog of cool stuff coming. MCP support, chat with an LLM about the data! Spanish. Augmented data.
A robotic email from Glassdoor asking whether CBS could cut costs by getting rid of Bari Weiss.
Let's all take a moment to admire this platform weaponization.
This is an apparently algorithmic robo email from Glassdoor to me based on my preferences.
Great job hacking the algorithm, Segment Manager!
The US government collected over $1 billion in immigration fees then refused to process the applications. No denials. No refunds. Just silence. This is the largest fee fraud in the history of the American immigration system. Here's what's happening. 🧵
JUSTICE SOTOMAYOR, dissenting from the denial of certiorari. Petitioner Priscilla Villarreal is a reporter who was arrested for doing something journalists do every day: posing questions to a public official. Specifically, Villarreal twice texted with a police officer to corroborate information Villarreal already knew about events that had occurred within her community. That officer voluntarily provided the information Villarreal sought, and Villarreal published those facts, consistent with her role as a journalist. Six months later, Villarreal was arrested for asking those questions. Making matters worse, Villarreal alleges that the arrest followed a months-long effort by a police department and district attorney’s office to retaliate against her because they disliked much of her reporting on their activities. Of course, that reporting was often critical of them. It should be obvious that this arrest violated the First Amendment. Yet the Fifth Circuit held that the officials were entitled to qualified immunity, and now Villarreal is left without a remedy. The Court today makes a grave error by declining to hear this case.
The Supreme Court refuses to review an odious 5th Circuit decision granting qualified immunity to Texas officials who arrested a citizens-journalist in retaliation against her posing questions to public officials. Sotomayor has an appropriately furious dissent: www.supremecourt.gov/orders/court...
MuckRock and @themarshallproject.org led a session to talk about their work on investigating book bans in prisons across the country.
Learn about the history of these bans and learn how to request and examine banned-book lists in their community.
I'm a @propublica.org alum. These are reasonable demands that benefit the reporting. I hope leadership finds reasonable compromises and demonstrates to others like @themarshallproject.org (also an alum) how to treat their staff with dignity & hold leadership to the same standards as their subjects.
