Persevering in a maddening media landscape.

July 10, 2025

We are suffering through an annoying, distracting, and exhausting news cycle in which many of the top stories are about “politics” in the worst sense of the word—partisan fighting that is reduced to a “media scorecard” that declares a winner in latest installment of the “incredible shrinking news cycle.”

The problem with most political reporting is that it treats the two major parties with false equivalency. One party is actively seeking to dismantle democracy. The opposing party is seeking to save democracy. But on the “media scorecard,” the “wins” and “losses” over the last four hours are treated with equal dignity, as if trying to destroy democracy and attempting to save it are the same thing. They are not.

If Republicans “get away with” their efforts to undermine democracy by engaging in lawless activity, they earn praise (or at least attention) from the major media. If lawful, peaceful efforts by Democrats to preserve democracy do not yield immediate results, Democrats are criticized, second-guessed, and covered with social media snark. What is missing from nearly all reporting is the fact that there are moral and constitutional dimensions to the stories that masquerade as “news” in major media outlets.

Reading news in the amoral and constitution-free framework adopted by major media is enough to make people sick of the news. I received the following note from a reader today:

I feel like I've been slowly spiraling since November and am worried about so many people right now. Following too much news every day has taken a big toll. I wanted to stay informed, but I think it's also worn me down and out, along with everything coming from the administration. I wonder if it's the same with others.

The reader asks, “I wonder if it is the same with others?” Yes! I hear similar sentiments from readers nearly every day. Indeed, I frequently hear, “I have stopped following the news”—as in “completely.”

Given the “news exhaustion” that many people are experiencing, it is a tricky task to write a newsletter about the daily news. (I am not complaining or seeking sympathy; just making an observation born out of concern for my readers.)

Today’s news cycle is a prime example. Major media outlets are leading with a story leaked by the White House to Fox “journalist” Peter Doocy that the DOJ has launched a criminal investigation of former FBI Director James Comey and CIA Director John Brennan.

The “leak” from the White House to a friendly reporter at Fox is the top story in the New York Times today. See NYTimes, Trump Officials Take Steps to Target Comey and Brennan, Who Investigated Trump. (Behind a paywall; I am not going to waste a gift subscription, for reasons described below.)

The “leak” to Fox News is not news. It is, instead, an effort to distract from the news that Trump's DOJ has squelched the release of the Epstein client list—presumably because Trump's name appears on flight manifests and visitor logs. By feeding “red meat” to Fox News, Trump is hoping to distract major media and his base from the outrage over the non-release of the Epstein client list.

The New York Times snapped at the bait. As MSNBC anchor Lawrence O’Donnell explained, no charges will ever be filed from the “leaked” investigation. Indictments are not the point. The point is distraction. And it worked. [Note: Trump investigated Comey during his first term. Bill Barr declined to prosecute Comey.]

But we—consumers of the news—are subjected to the political feeding frenzy of Trump leaking news about a pointless investigation of Comey and Brennan. As I wrote yesterday, it is enough to make a person weary.

The flip side of the non-story about the investigation of Comey and Brennan is Trump's announcement of a 50% tariff on all products from Brazil to punish the country for prosecuting its former president (Jair Bolsonaro) for—wait for it—attempting a coup to remain in office after he lost the 2022 presidential election!

Indeed, Trump accompanied the tariffs with a letter demanding that Brazil cease its prosecution of Bolsonaro! (Trump wrote that the trial of Bolsonaro is “a WITCH HUNT” and that Bolsonaro is “not guilty of anything.” Calling Dr. Freud!) See The Guardian, Brazil’s president rebuffs demand to cease inquiry into Bolsonaro after Trump tariffs.

Let’s pause here. In the media, this story is being reported as “Trump imposes 50% tariff on Brazil”—as if that action by Trump is a lawful exercise of the presidential power. It is not.

As I explained in detail earlier this week, the president has no authority to impose tariffs. The US Court of International Trade unanimously concluded that Trump's imposition of worldwide reciprocal tariffs was unconstitutional and was not authorized under the International Emergency Economic Powers Act (IEEPA).

But assuming for purposes of argument that Trump did have emergency authority under the IEEPA to impose tariffs, imposing tariffs to interfere in the sovereign affairs of another nation that is prosecuting a former president is not a lawful basis for exercising emergency powers under the IEEPA.

So the biggest story on Wednesday is that Trump acted in an unconstitutional manner to impose tariffs on Brazil. Did that never-before-in-the-history-of-our-nation scandal top the news? Nope! Not even close.

Sure, major media reported on the 50% tariff and Trump's letter demanding cessation of the prosecution of Bolsonaro, but there is no hint in any of the reporting that Trump was engaging in utterly lawless activity that has no grounding in the Constitution or legislative authority. See NYTimes, Trump Pledges 50% Tariffs Against Brazil, Citing ‘Witch Hunt’ Against Bolsonaro.

The Times does note that “Mr. Trump's effort to use tariffs to intervene in a criminal trial in a foreign nation is an extraordinary example of how he wields levies as a one-size-fits-all cudgel.” What is missing from that sentence—and the article in its entirety—is the fact that Trump has no power to impose tariffs and, even if he did, using them to interfere in a criminal prosecution in another country is an abuse of presidential power of epic proportions!

Like you, I am weary of such incompetent reporting. But we must look past the incompetence and see behind the headlines. Trump is violating the Constitution every day of his presidency. That is why we march, write, post, donate, and agitate. While some pundits view grassroots activists with bemusement or annoyance, we are the only constituency that recognizes the danger and is taking action to counteract it. And yes, that can be exhausting.

The good news is that there are tens of millions of us. No single person must carry the weight of the nation’s future on their shoulders alone. That is a collective burden shared by millions across many years. True, we must persevere even when we are sick of the “news.” In reality, our resistance is the “news.” Future historians will look back on this period and accurately report the most significant story of 2025 was that “millions of Americans rose up to save democracy.” That is the news from today. The rest is detail.

---

A reflection on Grok’s antisemitic slurs and the challenges presented by AI chatbots

An important news story of the last few days involved antisemitic statements by Twitter’s artificial intelligence agent, known as GROK. Musk’s engineers recently “tweaked” the algorithm for Grok, allowing it to provide answers without regard to “political correctness.” Shortly thereafter, Grok began to refer to itself as “MechaHitler” and made antisemitic slurs involving the tragic loss of life in the Texas floods. I won’t go into the horrible additional details, but you can read about them here: NPR, Elon Musk's AI chatbot, Grok, started calling itself 'MechaHitler'.

In a development that might be related to the Grok outburst of antisemitism, Twitter’s CEO, Linda Yaccarino, resigned on Wednesday. See Politico, Linda Yaccarino resigns as CEO of X.

Musk claims that Grok has been “retrained” to prevent similar statement in the future.

This story matters on many levels, including the fact that AI is being thrust upon us whether we want it or not. Many users of browsers (like Safari, Chrome, Edge) may not realize that the first-order answers to queries are being generated by AI plug-ins to browsers.

I am not an expert in AI, but I want to take a slight detour to provide a high-level, simplistic explanation of what is going on with Grok—and why it matters in dealing with the disinformation environment. For those of you who are experts in AI, please grant me some slack in trying to explain AI in a few paragraphs.

The first thing to understand is that current “AI models” are often referred to as “limited artificial intelligence.” That is, they are algorithms designed to perform specific tasks. To date, no one has created a “general artificial intelligence” model—and probably never will. (Sam Altman disagrees.) “General artificial intelligence” would mimic human intelligence in all its complexity, breadth, and nuance.

While it is theoretically possible to create general artificial intelligence, the computer processing power, training time, and energy requirements might exceed the total energy and time available in our universe. For example, a single ChatGPT search requires ten times as much energy as a Google search. Extrapolate that simple rule of thumb to a general artificial intelligence executing trillions of requests per second, and you are talking about serious amounts of energy.

Although most people assume that AI models like ChatGPT, Gemini, Grok etc. are artificial “intelligence,” they are a specific type of software program known as “large language models” (LLMs). Large language models work on the principle of predicting the frequency (or probability) of the next word in a phrase or sentence based on a prompt provided by the user. Faced with a prompt like, “Helena, Montana is . . . .”, an AI chatbot might return the words “beautiful,” “the capital of Montana,” or “the home of the annual Festival of the Trees.”

When a large language model is created, it must be “trained” by analyzing huge amounts of existing data contained in the vast reservoir of human knowledge. Training an AI model demands huge amounts of computer processing time and energy. In the next decade, the percentage of global energy devoted to AI is predicted to rise to 6% of global energy consumption!

Now, back to Grok. Large language models predict word frequency based on their “training.” Imagine that Grok is trained only on the historical content of all of the tweets on Twitter. If you asked Grok who won the 2020 presidential election (in my hypothetical), the high likelihood is that Grok will tell you that Trump won the 2020 election. Of course, Grok is not limited to tweets for its training, but Musk sees training Grok on tweets as a “differentiator” of Grok’s “real-time” responses on current events. See Medium, Grok AI and Real-Time Learning: How It Leverages X for Up-to-Date Responses.

The most recent “iteration” of Grok was programmed to “stop being woke” and to avoid “political correctness.” It is a short hop from prioritizing information in the cesspool that is Twitter to spewing antisemitism, holocaust denialism, and praise of Hitler.

But it gets worse. For reasons no one has satisfactorily explained, large language models frequently “make up” answers that have no basis in reality. These “made-up” answers are called “hallucinations.” A simplistic explanation for hallucinations is that large language models don’t “know” anything; they are merely “predicting” the next likely word based on training against vast amounts of data.

Depending on the AI bot and the nature of the test, hallucinations run from about 0.5% to 39%. Go back and re-read that sentence and keep it firmly in mind next time you use an AI bot and decide not to check the answers it provides.

So, we have a new technology that many people do not understand. They are asking questions and getting answers that may be the product of a computer hallucination. What could go wrong? We are already fighting in an information war. AI chatbots are going to make it worse—or better, depending on how the AI bots are used.

I have made an editorial and personal decision not to use AI chatbots in writing this newsletter. I do so because I believe that an important part of this newsletter is my personal relationship with the readers—which is dependent in large part on my “voice” and “personality.”

But I occasionally use an AI chatbot to identify underlying sources of original information. An experience in writing tonight has caused me to re-think even that limited use of AI chatbots.

Here’s what happened: Senator Adam Schiff gave a 37-minute speech on the Senate floor on Wednesday. He listed 10 ways that Trump is destroying America. A reader sent me a link to the 37-minute video and recommended that I share it with my readers. The video is here: I Just Got Off the Senate Floor - by Sen. Adam Schiff.

I did not have time to watch the entire video (as I was writing the newsletter). So, I searched for the text of Senator Schiff’s remarks using Google—but was unable to locate a text version of Schiff’s remarks.

I then asked one of the big three AI Chatbots to “provide a summary of Senator Schiff’s speech on the floor of the Senate on July 9, 2025.” I was hoping the prompt would result in a response that linked to a text version of Senator Schiff’s remarks—thereby allowing me to read the remarks, rather than listening to 37 minutes of audio.

The AI chatbot responded with a very nice five paragraph summary of Senator Schiff’s remarks—complete with links to underlying sources supporting the summary. As you should always do when using AI chatbots, I then opened the sources cited in the AI-generated summary and discovered that every source cited in the answer was from a speech given by Senator Schiff over the last six months—but NOT from his remarks on the Senate floor on July 9, 2025.

So, here’s what happened: I asked for a summary of remarks given by Schiff on July 9, 2025. The AI chatbot didn’t have the information necessary to answer my question. The correct answer was, therefore, “I can’t answer your question.” Instead, the AI chatbot made up a speech by patching together unrelated remarks given by Senator Schiff over the last six months!

The answer provided by the AI chatbot was a hallucination—in its entirety.

Imagine, then, if you are in a discussion with two friends about what the Big Ugly Bill does or does not do. You find yourself in a factual disagreement with your two friends. You consult Google, Friend One consults ChatGPT, and Friend Two consults Grok. You are likely to come up with contradictory or incomplete answers.

Here’s my point: The antisemitic slurs generated by Grok in response to a small tweak by programmers shows that large language model AI chatbots are creatures of their creators and the data on which they are trained. Musk has emphasized Twitter as a source of answers by Grok. Proceed with caution—and always check AI answers by examining source documents.

---

Concluding Thoughts

Thanks for sticking with me through a discursive newsletter. I hope that both articles help readers deal with a news environment that seems disconnected from the most important story of the day—which is your efforts to defend democracy. It doesn’t get any more important than that! The lack of acknowledgement of those efforts by the major media—and the failure of the media to acknowledge the daily lawlessness of Trump's actions—make it frustrating to continue the work of the resistance. But we have no other option.

We must continue our struggle in the face of a news environment that prioritizes political scorecards over the survival of our democracy. It has always been so. We must not bemoan that fact but must instead forge ahead confident in the knowledge that we are on the right side of history!

Talk to you tomorrow!

Daily Dose of Perspective