iiinova — where your digital self keeps evolving

Yoel Roth @yoyoel.com · 12/09/2026

TGI Fridays. A repellant French onion soup if ever there was one.

130

Reposted by Yoel Roth

Natalie Lung @natalielung.com · 11/09/2026

Wrote about what this means for two-party consent states: bsky.app/profile/nata...

0245

Yoel Roth @yoyoel.com · 09/09/2026

I don’t think the risk here is to Apple; it’s (maybe?) to medical professionals who unwittingly compromise the privacy of their patients’ information by having these features enabled during patient interactions.

160

Yoel Roth @yoyoel.com · 09/09/2026

That would be my interpretation of Apple’s docs, yeah

030

Yoel Roth @yoyoel.com · 09/09/2026

Ok but more seriously: (1) I am confident Apple’s MANY excellent lawyers and privacy engineering people have thought about this; and (2) I don’t understand how it is possible Protected Health Information doesn’t end up in Siri Recaps if you have this enabled.

6723

Yoel Roth @yoyoel.com · 09/09/2026

*screaming internally in HIPAA*

925464

Yoel Roth @yoyoel.com · 09/09/2026

Apple quietly dropped a ton of detail about the Apple Watch’s new always-on don’t-call-it-transcription feature and it is FASCINATING. support.apple.com/en-us/148354

Like all of our products and features, Audio Intelligence on Apple Watch was built from the ground up to protect your privacy.
• Audio is not recorded or stored. Audio Intelligence features don't create audio recordings. No audio can be accessed by the operating system, apps, you, or even Apple.
There is no recording to share, forward, or produce if requested by any party, because no recording exists.
• Raw audio is never accessible. Audio flows into a protected buffer inside the S11 chip's Secure Exclave — a hardware-isolated compartment built into the silicon of Apple Watch and iPhone that processes sensor data separately from the rest of the system. Data that is processed and analyzed in the Secure Exclave can't be accessed by the operating system, apps, you, or even Apple. Once processed, this data is immediately deleted.
• Siri Recap and Live Rewind help jog your memory, not provide a detailed record. Siri Recap doesn't produce a transcript. It generates a brief, high-level summary — comparable to notes you might jot down after a conversation. Live Rewind shows a text snippet of what was just said in the last 15 seconds, so you can catch something you missed.

• No speaker attribution. By design, Audio Intelligence features don't identify and attribute speakers. Siri Recap may contain names if mentioned in the conversation, but doesn't directly attribute speakers (for example, "John said..." or labeling as "Speaker A" versus
"Speaker B"). Similarly, Live Rewind may transcribe what someone said word-for-word, but does not attribute quotes to speakers.
• Audible and visible signals for Live Rewind.
When activated, an audible chime plays from the speaker on your Apple Watch, even if your Apple Watch is on silent or you have headphones connected. A full-display animation and microphone indicator appear on the display, and a distinct double-press gesture is required to activate Live Rewind. This is to alert those nearby that the feature has been activated.
• Omitted content. Siri Recaps are designed to omit potentially harmful content, such as language that promotes self-harm or hateful speech, and sensitive information, such as financial data, government-assigned identifiers, authentication data, and certain personal identifiers.
• Be mindful of those around you. While audio is not recorded and stored, and no speakers are attributed, consider those around you where conversations might be private or sensitive.

57508142

Yoel Roth @yoyoel.com · 18/08/2026

I don’t see the problem

1100

Yoel Roth @yoyoel.com · 04/04/2026

It’s fine, just hold the power button until you hear the startup chime

1582

Yoel Roth @yoyoel.com · 09/03/2026

I’m in some highly specific group chats.

020

Yoel Roth @yoyoel.com · 06/03/2026

Honored for sure. That line (and the last one) absolutely killed me.

1140

Yoel Roth @yoyoel.com · 06/03/2026

I asked Claude to generate a book jacket blurb for a novel inspired by Heated Rivalry and the Lumen database, and, umm…

NOTICE AND TAKEDOWN
Every day, thousands of removal requests flow into the Lumen Database — the public archive where the internet's most powerful actors are forced to show their work. Most requests are routine. Some are weapons.
When copyright attorney Marcus Vega files a takedown notice targeting Clearwire, a scrappy digital rights nonprofit, he expects the usual: compliance, a form letter, maybe a sternly worded reply. What he gets instead is Eli Marsh - Clearwire's counternotice specialist, former EFF researcher, and the most infuriating person Marcus has ever had the professional misfortune of opposing.
Eli knows the law. He also knows Marcus - knows his firm's clients, his win rate, the pattern of industries he protects. What Eli can't account for is what happens when the two of them end up at the same internet governance conference in Geneva, arguing in bad faith over cocktails until they're not arguing at all.

The notices keep coming. So do the counternotices.
So do the emails that start as legal correspondence and become something neither of them can quite name.
Marcus has spent his career making inconvenient things disappear. Eli has spent his making sure they can't. Between them sits every question the internet has failed to answer about who gets to control what's public, what's private, and who decides.
Some things, it turns out, can't be taken down.
For readers of Heated Rivalry, Better Than People, and anyone who has ever found themselves deeply, inconveniently attracted to someone who sends documents with tracked changes turned on.

138511

Yoel Roth @yoyoel.com · 11/02/2026

I feel extremely attacked by this

0100

Yoel Roth @yoyoel.com · 18/01/2026

I kept having one where I’d tap an article and it would load to a blank sheet, which I could only resolve by force-quitting the app and relaunching. It happened often enough to annoy me into switching.

110

Yoel Roth @yoyoel.com · 18/01/2026

I got sick of some of the bugs in legacy Reeder and switched to Unread. It’s a delight. Highly recommended.

330

Yoel Roth @yoyoel.com · 02/01/2026

Sonic is very high up on the list of reasons I’d be reluctant to leave the Bay Area. 10 gigabit fiber for $60 a month, from an ISP that actually respects its users. We stan Sonic.

3421

Yoel Roth @yoyoel.com · 29/12/2025

Anyway, Brandolini’s law continues to hold: Bullshit is asymmetrical. Editorial failures like this put the burden on me to track how even purportedly scholarly works (mis)use what I’ve said and written, and expend time and effort to set the record straight. It’s untenable in the long run.

112617

Yoel Roth @yoyoel.com · 29/12/2025

I still can’t wrap my head around how this could have happened in any reasonable research and writing process. Like, I’ve written lots of stuff; I always start by bucketing sources into “agrees with/supports the point I’m making” and “disagrees with it.” How do you accidentally mix those up?

4973

Yoel Roth @yoyoel.com · 29/12/2025

I appreciate the forthright acknowledgement of the mistake from the author, and look forward to decisive action by @princetonupress.bsky.social to correct this error, and any others, that made it past their robust editorial process.

Dear Mr. Roth,

I am not sure how it happened, but you are right that I made a mistake in treating the words quoted from your essay as a representation of your views. It may be that when I returned to your essay (which I had previously read) in finishing a draft of the book, I interpreted the bolded text as a “pull quote.” That was a mistake, for which I am extremely sorry. It was my mistake and not that of my co-author.

I am working with Princeton University Press to have that sentence deleted from the book (all future printings and the Kindle edition). Once the Press’s offices re-open after January 2, we will explore the possibilities for an errata statement on the book’s web pages. I will also send out a message on Twitter acknowledging the mistake in misleadingly using the passage from your essay.

I want to offer you my sincere apologies for this very regrettable mistake. My aim and that of my co-author has been to try to cast some light on difficult and important matters about which there is much disagreement. It is far from my intention to mislead concerning the pandemic and decisions made in response to it. Our book has 935 footnotes, but this was not a footnoting error. I deeply regret having made this mistake.

Sincerely, Stephen Macedo

Thank you for the direct acknowledgement of the error, and for the commitment to addressing it. I look forward to confirmation that the steps you outlined have been taken, and hope that you and the team at PUP will do a thorough review of the text to ensure other such errors haven't snuck through.

Dr. Roth

830948

Yoel Roth @yoyoel.com · 27/12/2025

Requested a correction from @princetonupress.bsky.social. Let's see what happens.

Dear Princeton University Press Editorial Team,

I am writing to request a correction to In Covid's Wake: How Our Politics Failed Us by Stephen Macedo and Frances Lee. The book contains a significant misquotation that inverts the meaning of my published work.

In 2023, I published an article for the Knight First Amendment Institute at Columbia University titled "Getting the Facts Straight: Some Observations on the Fifth Circuit Ruling in Missouri v. Biden." In that piece, I directly refuted false claims about government jawboning of social media platforms. One claim I specifically pushed back against was the myth that the FBI demanded Twitter censor online speech.

And yet, I was recently surprised to find that In Covid's Wake quotes me as stating that "the FBI went beyond strategic information sharing and made direct moderation demands" — the exact claim I was refuting, presented as though I endorsed it. The book uses my refutation of a false narrative to support that very narrative, employing my name, former position, and scholarly writing to buttress the authors' argument at the expense of the integrity of my own views.

This is not a case of authors disagreeing with my interpretation or selectively quoting from a nuanced argument. The book attributes to me a position that is the precise opposite of what I argued. The passage the authors quote is a section heading that introduces my refutation of that claim, not an endorsement of it.

The misrepresentation is particularly harmful because it lends Princeton University Press's credibility to demonstrably false claims about government censorship. My article exists specifically to establish what actually happened versus what conspiracy theories allege. By quote-mining a section heading and divorcing it from the argument that follows, the book weaponizes my attempt to establish facts against the truth itself.

I request that the authors and Princeton University Press:
Issue a formal correction acknowledging the misquotation
Revise the relevant passage in future editions to accurately represent my argument
Post an erratum on the book's webpage
I recognize that scholarly interpretation involves judgment calls about sources. However, quoting someone's statement of a position they are about to refute as though they endorse that position is not interpretation — it is dishonest, misleading, and a self-evident factual error. This is not a matter of academic disagreement but of basic accuracy in representing published work.

I hope Princeton University Press shares my commitment to factual rigor and will address this matter promptly.

Sincerely,

Yoel Roth, PhD

201733303

Yoel Roth @yoyoel.com · 24/12/2025

There’s a lesson here about writing in the age of weaponized lies: Always precede false claims with giant red text, ideally wrapped in a <blink> tag, that says “THIS IS A LIE I AM REFUTING!!” But also, why is a book that makes this kind of lazy, bad-faith argumentation getting any accolades at all?

272171172

Yoel Roth @yoyoel.com · 24/12/2025

And yet, In Covid’s Wake quotes me as saying that “the FBI went beyond strategic information sharing and made direct moderation demands” — literally the opposite of what I argue — to buttress their claim that the government was too busy censoring speech to adequately deal with the pandemic.

91640155

Yoel Roth @yoyoel.com · 24/12/2025

One such refutation appears in my paper under the heading “The FBI went beyond strategic information sharing and made direct moderation demands.” I spend several hundred words explaining that that didn’t happen.

The FBI went beyond strategic information sharing and made direct moderation demands
In the Fifth Circuit's analysis, the line between permissible information sharing and threat alerting and coercive moderation demands is a porous one. If representatives of the government share indicators of compromise (IOCs) with private sector entities, is a coercive demand to take some kind of action implicit in that sharing? The Fifth Circuit takes this to be the case:
... we do find the FBI's requests came with the backing of clear authority over the plattorms.
Atter all, content moderation requests "might be inherently coercive if sent by ... [al law enforcement officer."
... although the FBI's communications did not plainly reference adverse consequences, an actor need not express a threat aloud so long as, given the circumstances, the message intimates that some form of punishment will follow noncompliance.®
The implications of this argument are striking. The Fifth Circuit appears to hold that any information-sharing from the FBI to platforms is de facto coercive, simply by virtue of who it originated with.
Even in situations where platforms actively solicit information from the government-as was regularly the case on election security matters between platforms and the FBI-the resulting
communications are taken to be coercive because of the FBI's standing as a law enforcement entity.

Receiving and acting on external reports is a core function of plattorm content moderation teams, and the essential nature of this work is an independent evaluation of reported content under the platform's own policies. The fact, cited in Missouri v. Biden, that platforms only acted on approximately half of reports from the FBI shows clearly that the standards platforms applied were not wholly, or even mostly, the government's.
Finally, it does not withstand factual scrutiny that platforms were so petrified of adverse consequences from the FBI that they uncritically accepted and acted on information sent to them by the government. The Twitter Files themselves document clearly at least two instances in which, presented with low-quality information or questionable demands, Twitter pushed back on the FBI's requests. In one case, the FBI passes on a request-seemingly from the NSA-that Twitter "revis[e] its terms of service" to allow an open-source intelligence vendor to collect data from the Twitter APIs to inform the NSA's activities. This request is arguably as close to jawboning as any interaction between Twitter and the FBI gets; yet, in response, I summarily dismissed not only the request for a meeting to discuss the topic, but the entire premise of the request, writing, "The best path for NSA, or any part of government, to request information about Twitter users or their content is in accordance with valid legal process." The question was not raised again.

51400123

Yoel Roth @yoyoel.com · 24/12/2025

The authors quote from an article I wrote in 2023, refuting false claims about government jawboning of social media platforms. One of the claims I directly push back against is the persistent lie that the FBI demanded Twitter censor online speech. knightcolumbia.org/blog/getting...

knightcolumbia.org

Getting the Facts Straight: Some Observations on the Fifth Circuit Ruling in Missouri v. Biden

51318113

Yoel Roth @yoyoel.com · 24/12/2025

A small (personal) example of this book’s intellectual dishonesty: My father-in-law is reading In Covid’s Wake, and excitedly told me he found a passage where I’m quoted. The quote in question is me saying the FBI worked to censor speech on social media. Huh? When did I say that?!

A passage from the book In Covid’s Wake, with a highlighted sentence reading: “Yoel Roth, the former head of trust and safety at Twitter, states that "the FBI went beyond strategic information sharing and made direct moderation demands," including demands involving domestic speech.”

7658671668

Reposted by Yoel Roth

Daphne Keller @daphnek.bsky.social · 21/07/2025

@yoyoel.com at TrustCon making the key point that big platforms can weather storms and absorb blows (litigation, regulatory churn) that kill their smaller rivals. The legal threat may look the same to an array of platforms, but the real question is who can survive it.

1417

Yoel Roth @yoyoel.com · 04/06/2025

The operating mantra is, “Yes, and… you’re banned”

090

Yoel Roth @yoyoel.com · 13/05/2025

This is an extraordinary and courageous piece of writing that wrestles with one of the great challenges of this moment: How do you decide whether to remain a part of broken institutions to try to fix them from within?

713847

Yoel Roth @yoyoel.com · 07/03/2025

Don’t make me snitch tag the menswear guy

0270

Yoel Roth @yoyoel.com · 21/02/2025

Seems like something we could figure out. :) Will be in touch.

130

Yoel Roth @yoyoel.com · 11/02/2025

I was just thinking of doing this! Good to hear it holds up.

020

Yoel Roth @yoyoel.com · 07/01/2025

Genuinely baffled by the unempirical assertion that Community Notes “works.” Does it? How do Meta know? The best available research is pretty mixed on this point. And as they go all-in on an unproven concept, will Meta commit to publicly releasing data so people can actually study this?

We are now changing this approach. We will end the current third party fact checking program in the United States and instead begin moving to a Community Notes program. We’ve seen this approach work on X – where they empower their community to decide when posts are potentially misleading and need more context, and people across a diverse range of perspectives decide what sort of context is helpful for other users to see. We think this could be a better way of achieving our original intention of providing people with information about what they’re seeing – and one that’s less prone to bias.

1021135234

Reposted by Yoel Roth

Ryan J. Gallagher @ryanjgallag.com · 07/01/2025

Community Notes don't work by themselves. So many things don't get labeled because people don't reach "consensus" on them, which is really what Community Notes optimize for, not factuality What an utter failure. So disappointing

914431

Yoel Roth @yoyoel.com · 04/01/2025

That teaching is prioritized and institutionally incentivized.

0140

Yoel Roth @yoyoel.com · 10/12/2024

One of the things we all really need from platform T&S teams — honest, authentic, human conversations about the impossible tradeoffs involved in this work — is made effectively impossible by the deranged ways a small number of people treat platform staff when they speak publicly. Ask me how I know.

3257179

Yoel Roth @yoyoel.com · 19/11/2024

Somehow I just KNEW you would understand this

1140

Yoel Roth @yoyoel.com · 19/11/2024

Oh no, the Signal icon very subtly changed colors and now it’s setting off my “what is weird right now” alarms every time I unlock my phone

142427