From the Eddy
Back after a long gap with nine verdicts, one Greek summer, and a new rule of thumb. Health care was government in the '90s. AI is government now.
This is what I read about AI this week, with one-stamp verdicts. The stamps are first-person reactions on purpose. This column is about my journey through this stuff, not a neutral roundup. Each story gets one primary verdict, plus Florida implications when the connection is the point.
The verdicts:
Bookmarked it — reference material I’ll come back to.
BS — obviously baloney; fake outlets, AI-generated content masquerading as journalism, pink slime.
Changed my mind — shifted a view I held going in.
Didn’t make me smarter — read it, learned nothing, here’s the warning.
Gave me hope — left me a little less worried about where this is going.
Hit close to home — landed personally.
Made me smarter — genuinely taught me something.
Not in my purview — interesting, but outside my beat. I’m flagging it, not interpreting it.
Raised more questions — opened doors instead of closing them.
Set off my hype alarm — claims outrunning the evidence.
Plus Florida implications, when the connection is the point.
Plus one tracking stamp for pieces I’m watching develop:
On the radar — tracking it, no verdict yet.
This is the second time a long stretch has passed between From the Eddy columns, and the gap tells the opposite story of the one you might expect. The intersection of AI and Florida politics has been claiming more of my attention than ever.
Since the last edition, I moderated my first AI panel. I spent two FloridAI columns inside the intersection of AI and health care, including a look at what happened when AI came for the radiologists. And the tool itself has rewired my job. I now use Claude to review our reporters’ published work, flagging errors and tightening style, and that editorial loop has drawn me deeper into the daily operation of Florida Politics than I’ve been in years. I feel empowered by it.
The reading never stopped either, and a pattern has emerged in my inbox. I used to tell people that health care is government. It consumes so much of the budget that you can’t quite separate it into its own silo, yet it looms so large that you almost have to. AI is developing the same gravity. Stories about AI and government now arrive at the pace health care stories did in the ‘90s, when Bill Clinton made his push, and again when Obamacare moved from bill signing to implementation. AI is government now, or close enough that the distinction is fading.
Away from the desk, our family has been sprinting through campaign season while savoring the last summer before our daughter finishes grade school. This year’s vacations have followed a Greek theme, most recently a trip to New York for a Percy Jackson camp in Central Park. Homer would approve of the itinerary, if not the subway.
Somewhere in all that travel, this reading list came together. Here’s what earned a stamp.
— Camila Jose Gomez of the Tampa Bay Times reports on the AI-generated ad from Charlie Crist’s St. Petersburg mayoral campaign, which asks viewers what would happen if they showed up to work 34% of the time. The three “voters” who answer are AI avatars, fed their lines by prompt, and the swing at incumbent Ken Welch is built on the Times’s own reporting about his City Hall attendance. “I could have spent $28,000 for a 12-hour film crew to be out in the 92-degree Florida sun in five locations and got the same result that they got probably with 12 prompts in 30 minutes,” consultant Anthony Pedicini said. Experts judged the ad “tame” because it sticks to fact-checked reporting instead of distorting anything. What AI adds is imagination, wrapping vivid visuals around the least cinematic material in politics, an argument about time cards. In the run-up to “The Odyssey” arriving in theaters, I read Gareth Hinds’s graphic novel adaptations of the Iliad and the Odyssey, and they were great because they illuminated what is at heart a big comic book tale. AI ads do the same work for voters, filling in the imagination.
Verdict: Hit close to home — Michelle helped produce this one. The AI-in-campaigns story doesn’t get much closer than my own kitchen table.
— Derek Thompson surveys what he calls the four horsemen of the AI bubble apocalypse: spending risk, revenue risk, political risk and technological risk. The two weeks before the piece supplied the omens. Alphabet posted a negative free cash flow quarter for the first time in its history as a public company, Meta’s stock plunged nearly 10 percent in a day, South Korea’s AI-fueled market crashed more than 40 percent in a few sessions, and China’s Kimi K3 arrived near the frontier at a third of the price. Thompson compresses the whole mess into a single sentence, writing that “the capabilities of AI are becoming more powerful, while the economic underpinnings of the AI buildout—and, as we’ll discuss, the political support for AI—are becoming more vulnerable.” What makes the piece worth your time is that he pairs every risk with his best counterargument, and he ends on opportunities and risks instead of certainty. The storms gathering on the horizon are real, and not one of them is AI-generated.
Verdict: Raised more questions — including some with real money attached. I finished the piece wondering about the chip stocks sitting in my own Robinhood account, and Thompson would be the first to tell you he doesn’t know the answer either.
— Damon Beres of The Atlantic argues that Apple’s new rental program is a trap. Apple Upgrade replaces the old installment plan with a true lease run through Klarna. Rent a $1,999 MacBook Pro at $54 a month for two years and you have paid $1,296, still owe $703, and the likeliest outcome is you trade in and restart the cycle, renting in perpetuity. The reason prices are climbing in the first place is AI, whose appetite for memory chips has pushed up the cost of consumer electronics across the board and put a $10,000 MacBook Pro into the conversation. (The new iPhones, incidentally, come in a shade called Deep Blue, a choice the chatbots further down this column would endorse.) “I walked into the trap myself—even saw it coming—but I am nonetheless ensnared,” Beres admits. The AI divide was already about who can afford the subscriptions, and now there is a hardware layer underneath it, with the price of a good seat at the table going up.
Verdict: Bookmarked it — I just bought a new Mac myself because I needed more computing power to run my beloved Claude at optimum speed, so I felt this one in the wallet. Watching what happens with Apple’s rental program from here.
— POLITICO’s Nick Reisman reports that Kathy Hochul is enacting the nation’s first state-level pause on large-scale data center construction, an executive order halting new permits for up to a year while New York writes energy and environmental standards. Already-issued permits stand, and facilities built for research, education and health care get carve-outs. “As data center development threatens to hike up utility bills, deplete our natural resources, and create uncertainty for New Yorkers, it’s my responsibility to take action and lead,” Hochul said. The politics are with her. A Siena poll in June found New York voters back a one-year pause, 46 percent to 21 percent. This is one way of doing it, and it previews the trouble with a state-by-state approach to regulating AI infrastructure. New York can do what it wants, and the interesting question is what the other 49 states do with the opening.
Verdict: Raised more questions — starting with what actually happens when a state decides to go against data centers, and who benefits when it does. Florida implications — if New York’s pause becomes a playbook, the projects don’t disappear, they relocate. A big state with abundant rural land is exactly where developers look next. Nobody wants to say Florida has a competitive advantage here, but it’s the truth.
— Daniel Rivero of WLRN profiles the 21-year-old behind Florida Data Centers, a statewide tracker that pulled in a million visits in its first month online. James Boger, a West Palm Beach web developer, built floridadatacenters.orgafter the proposed $2.6 billion Project Tango landed in Palm Beach County, and the site now reads like a running journal of fights most outlets aren’t covering. He has no insider sources, just public records aggregated from county commission and utility websites. He also sells a $9 service that files opposition letters with county commissioners, with more than 1,600 sent. “By the time the developers come in and file the plans and put them on the board’s table, there’s a giant list from all the residents saying, ‘Do not build these here. We don’t want them,’” Boger said. The withdrawal of an 813-acre Citrus County project was reported only by a hyperlocal Facebook group, which tells you plenty about the news hole Boger is filling. Thompson’s political-backlash horseman is an abstraction until you see it organized, county by county, on one map.
Verdict: Bookmarked it — I want one place to track every data center fight in the state, and this is the closest thing to it. You should bookmark it too. Florida implications — with projects live in St. Lucie, Okeechobee, Martin, Polk and Citrus counties and a new state law preserving local zoning control, this is becoming one of the most potent political issues in Florida.
— Evan Gorelick of The New York Times ran hundreds of prompts for The Morning newsletter to answer a question most users never think to ask, which is where a chatbot’s answers actually come from. Ask ChatGPT, Claude or Grok its favorite color and all three say blue, usually “deep blue,” because models amplify the trends in their training data and blue is the world’s favorite favorite color. The same tendency explains the em dashes and the “it’s not X, it’s Y” constructions. The moral questions are where the models split, because developers override consensus morality with bespoke instructions, from Anthropic’s nearly 80-page constitution to the Grok guidance that ended in “MechaHitler.” “When a chatbot tells you what to do, it is channeling, in some unknowable mix, the average opinion of the internet and also a developer’s own sense of what ‘right’ should sound like,” Gorelick writes. He closes with the right advice, which is to keep asking, but remember who is answering.
Verdict: Made me smarter — and after a career adjacent to graphic design, I too am a fan of deep blue, the kind Michael Mann and Christopher Nolan build whole films around. Nice to learn the machines and I share a palette, even if we arrived at it differently.
— This one is a support article from Anthropic, not journalism, and it earned its place anyway. The document explains how Claude’s Cowork feature works on Team plans, and it contains the sentence I care about most: “Starting August 3, 2026, Claude Cowork is in beta on web and mobile,” with cloud sessions that carry your work across desktop, web and mobile. That sentence matters to me because a couple of weeks ago Michelle and I flew to New York for our daughter’s camp, and I learned the hard way that my Cowork projects lived on my desktop back home, not in the cloud. Every project file I’ve built for Florida Politics, FloridAI and Sunburn was out of reach. I was flying blind, and I didn’t understand how blind until I couldn’t touch any of the working memory I’ve spent months building inside Claude. The dependency crept up on me, which is its own data point about where this technology already sits in my operation.
Verdict: Bookmarked it — and circled the date. Cloud sessions arrive today, and I don’t plan on flying blind again.
— OAN’s Angel Russell interviews criminologist David Maimon about what happens when stolen personal data meets AI, and his answer is that criminals are combining breach data with generative tools to mint fake identities and target Americans at scale. Maimon is worth listening to on this. He is Head of Fraud Insights at the fraud-prevention firm SentiLink, a professor of criminal justice and criminology at Georgia State University, and director of its Evidence-Based Cybersecurity Research Group, where he spends his time undercover on dark-web marketplaces watching fraudsters work. He testified before the House Oversight Committee on the same subject in mid-July. And the Florida angle is already here. In February he described Medicare fraud schemes built on older Floridians’ stolen identities and fictitious billing companies. “Everybody gets hit by fraud,” he told The Floridian, pushing back on the idea that only the elderly are targets. The fake-people problem is no longer theoretical, and the people building them have better data than most campaigns.
Verdict: Raised more questions — starting with how anyone verifies anyone in a world where identity itself can be generated. This one scared me, and I don’t scare easily.
— Meta announced in a newsroom release that it will now alert parents who use Instagram’s supervision tools when a teen’s conversation with Meta AI shows signs of suicidal thinking or self-harm. A corporate release gets read with the usual discount, but the details here suggest real care. Every flagged chat gets reviewed by a human before an alert goes out, and the company gathered feedback from more than 75 clinicians who specialize in teen mental health. “If a teen’s intent is ambiguous, we’ll err on the side of caution and alert the parent,” Meta writes. The parent alerts are live now in the U.S., U.K., Australia and Canada. The companion feature, contacting emergency services when someone appears at imminent risk, is still in development, and that is the part to watch. A promise from a platform is not the same thing as a shipped product, and this release contains one of each.
Verdict: Gave me hope — especially as the parent of a teenager growing up in a world AI is shaping. I don’t spend my days worrying about her, but this is a beat where you stay eternally vigilant, and a tool like this helps.
Now your turn. What earned a stamp on your reading list this week? Drop it in the comments or send it my way. The next edition’s already filling up, and I’ll be reading.
Ed. note: This story was drafted with assistance from AI. Editorial judgment, sourcing, and final review were performed by Peter Schorsch and the Florida Politics editorial team.