Thursday, May 28, 2026 · 139 newsletters
Trump's Texas Takeover
Texas Senate · Paxton · AI agents · Iran · energy · stablecoins · data centers · Pope Leo · China · creator economy
Published on Thursday, May 28, 2026.
Pulled from 138 newsletters in yesterday's inbox. One political story ate most of the oxygen, but the AI conversation underneath it shifted in an interesting way: the model wars are quietly being replaced by a fight over plumbing.
Politics: Trump's Texas Takeover
The biggest story of the day, by a wide margin, was Ken Paxton crushing John Cornyn in the Texas Republican Senate primary runoff. The reactions ranged from grim to gleeful, and the convergence is the story: a corruption-plagued state AG just dispatched a four-term incumbent because Donald Trump endorsed him. That is now the only test that matters in a Republican primary.
Republicans got the candidate they earned. Bill Kristol and Andrew Egger at The Bulwark called it exactly that, framing the night as the moment the GOP base completed its transformation into something more like a cult than a coalition. Rick Wilson was less restrained: Cornyn, he wrote, won statewide nine times, then got "absolutely wrecked by one of the most corrupt figures in American public life, all because Donald Trump said so." Lincoln Square's open letter traced Cornyn's arc from respected Texas Supreme Court justice in 1997 to political corpse in 2026, treating the obituary as the lesson.
Democrats see an opening, and they are not subtle about it. Paul Krugman argued the Senate race will now be a referendum on healthcare, with Paxton's brand of corruption forcing him to lean fully on Trump for cover. Dan Pfeiffer at The Message Box called Paxton "an absolute gift" and laid out the case for James Talarico as the rare Democrat who can actually run statewide in Texas. Rick Wilson's other piece pushed the same theme: Tejano turnout plus a clean candidate against a "bottom-feeding criminal" is a winnable map. Matt at Crooked and Pod Save America treated the result as a flat gift for the DSCC; Bloomberg's morning brief led with the "Paxton crushes Cornyn" framing.
The dirty-tricks side of the night was its own story. Judd Legum at Popular Information walked the breadcrumbs from a "progressive" super PAC called Lead Left to a Republican operative, who spent roughly $900,000 boosting a sex-therapist Democratic primary candidate in Texas-35. That candidate lost to Johnny Garcia anyway, but the operation is now in the open. Democracy Docket ran two stories worth pairing: Paxton's ongoing crusade to relitigate 2020, and a separate filing showing election deniers and January 6 rioters lining up for a piece of the $1.776 billion DOJ "slush fund" that Sarah Longwell and Mary McCord spent an hour unpacking on the same day a federal judge tore into the DOJ's "vindictive" Broadview Six prosecution.
The convergence is clear. Trump used the primary to enforce loyalty over competence and won; the Democrats now have a real shot at a Texas Senate seat they would not have had against Cornyn; and the DOJ scandal pipeline (Broadview Six, Kilmar Abrego Garcia, the slush fund) keeps feeding the same story Lincoln Square's strategy session is now openly asking: is anyone in the chain of command actually fit to be there?
AI: The Model Wars Are Over, the Plumbing Wars Begin
The clearest thread across builder-facing newsletters yesterday was that picking an LLM is no longer the interesting decision. Max Mitcham's "The LLM Doesn't Matter Anymore" put it most bluntly: every founder is still fighting GPT vs. Claude vs. Gemini, but any frontier model in 2026 performs well with the right infrastructure wrapped around it. The differentiation has moved to context, tools, and orchestration.
Skills, not prompts. Ken Huang's SkillOpt write-up framed the shift as moving from stochastic prompting to deterministic skill engineering, with self-evolving agent skills as the next layer. The Pragmatic Engineer interview with Dax Raad on OpenCode and Peter Yang's slide-building tutorial in Claude Code both pointed at the same thing from opposite ends: the unit of value is a reusable skill, not a clever prompt.
Agents as labor, not features. Guillermo Flor at Product Market Fit walked through a one-hour Claude setup that runs 40 hours of CFO grunt work a month: data pulls, reconciliations, variance explanations, scenarios. The GTM Engineer's case study on Vanta showed Dust turning every seller into an agent builder. ByteByteGo's breakdown of Airtable's search layer explained the unglamorous retrieval work that makes any of this usable. Noah Brier's Forward Deployed episode 6 with Rohit Krishnan went deeper still, sketching out what market mechanisms for agents actually look like.
The labor question is no longer hypothetical. Noah Smith argued at Noahpinion that the new white-collar job description is keeping AI on task: less making, more supervising. Every's "After After Automation" tied that to the Vatican weighing in on AI labor and to Codex playbooks. The Linear voice-AI deep dive with Bessemer's Mike Droesch made the same point one layer down: voice agents are now boring infrastructure, the interesting fights are about routing, escalation, and trust.
The take: when the model is no longer the moat, what is left is workflow, distribution, and judgment. The newsletters that treat AI as a tool category are getting less interesting; the ones treating it as a labor question are getting better.
Energy and Iran: The Forever War Trap
A second cluster, smaller but louder than usual: the Iran "deal" that keeps almost happening. Lincoln Square's Anchor Watch with Bobby Jones reminded readers that the "four-day war" with Iran is now in week twelve, with Trump declaring victory on roughly a weekly basis. Foreign Affairs Today led with Lawrence Freedman on "Iran and the Forever War Trap," arguing that the search for a quagmire exit has produced a different kind of dead end. Lincoln Square's evening brief with Edwin Eisendrath and Susan Demas tied it to the domestic side: gas prices climbing, Trump calling the hikes "peanuts."
Hormuz is the pricing signal. Maritime Analytica's briefing on the strait treated "Hormuz is open" as a misleading headline: the actual industry guidance turns on insurance exposure, crew endurance, and management judgment, and the answer to "should every ship go" is not yet yes. Their second piece, HMM's Q1 results, was a quieter signal in the same direction: profit survived, margin shrunk, and that is how container shipping crises usually start.
Munitions and climate. Contrary Research's "The American Missile Crisis" made the case that US munitions stocks have been drawn down past the point of comfort by Ukraine and Israel both. David Callaway on greenhushing flagged the inverse trend on the climate side: the most ambitious climate fighters are now the quietest, because saying the words out loud is a political liability.
The convergence is grim: the political pricing on energy and security is doing more work than the underlying facts, which means the facts can move a lot before the politics catches up.
Money: Tokens, DeFi, and Robinhood's Robots
Finance newsletters spent the day talking past each other in productive ways. Brew Markets and Tech Brew both led with Robinhood letting AI agents trade on the platform; Snacks noted the US now hosts a dozen trillion-dollar companies, with Micron the newest entrant on AI memory demand.
The infrastructure layer is the interesting layer. Tearsheet's new Letter from the Editor argued finance is becoming ambient infrastructure underneath everything else, and asked what we are giving up for the convenience. Sam Boboev's Fintech Wrap-Up bundled three reports on card issuing, banking in the tokenized economy, and the three types of digital money. Fintech Business Weekly's interview with zerohash's Edward Woodford was a useful read on the same plumbing question from the crypto side.
DeFi's identity crisis is still ongoing. Bankless ran "DeFi's Uncertain Future," tying agents and the x402 protocol to a slightly hopeful read on safer onchain finance, with BTC at $75.1k and ETH near $2.1k for context. The Breakdown ran a Disclosure-is-all-we-need piece on the same theme: regulatory clarity, not new primitives, is what unlocks the next leg.
The other end of the stack. Linas's deep dive on Starling Bank's 2026 annuals is the standout fintech read of the day: 2.5x Monzo's profit on half the revenue, plus a £70M SaaS pipeline that quietly changes the Starling thesis. Dwayne Gefferie's piece on mule-account handovers made a useful point on the fraud side: accounts that pass KYC clean still change hands months later, and the industry has no good answer.
If you read these together, the picture is a financial system that is rebuilding its primitives (tokens, agents, ambient identity) without rebuilding its enforcement layer. That is the gap.
Infrastructure: Thank God for Data Centers
Tied to the AI thread but worth its own section. Not Boring's "Thank God For Data Centers" made the broad case that the infrastructure buildout is the only thing keeping the US industrial story coherent right now. Ben Thompson at Stratechery wrote up the SpaceX IPO as a financial mystery that resolves only if you take seriously the prospect of data centers in orbit, which he does. a16z's "Avoiding Death on the Yellow Brick Road" put the policy frame around it: the buildout is fragile, and the wrong policy turn can kill it.
The take: the AI labor story and the AI capex story are now the same story, and whether American manufacturing recovers depends in large part on whether the grid and the chips show up on time.
The Vatican Weighs In
A genuinely unusual cluster: Pope Leo XIV's encyclical on AI was the single non-tech story that the most tech writers covered. Mike Solana at Pirate Wires took the pope's "Babel" warning seriously, arguing that the framing (humility in the face of tools we do not fully understand) lands harder than most tech-ethics output of the last year. Lincoln Square folded it into the daily political read. Every threaded it into "After After Automation" as a labor argument. Two readers, three reads, one signal: when a Catholic pope publishes the most quoted AI-ethics document of the week, the secular AI conversation has lost a step.
China: Long Lens, Short Demand
Thinner than usual but worth flagging. Trivium China led with Xi's "stabilizing and enhancing confidence" framing and a podcast with Jon Czin on the US-China stalemate. ChinaTalk's "China's Robotics Dream Began in 1972" traced today's humanoid race back to the founding work of one engineer, and is one of the better pieces of long-form context in my inbox this week. Aletheia's reminder for Vincent Chan's call on China strategy framed the immediate market read as a "dichotomy of strong supply and weak demand," which is roughly the entire China macro story in six words.
Ideas Worth Reading
The LLM Doesn't Matter Anymore by Max Mitcham. The piece that crystallized the day's AI mood: infrastructure beats model selection in 2026.
Debunking the Five Biggest Myths of Psychology from Jay and Dom at The Power of Us. A clean takedown of the conformity / obedience / bystander stories you remember from intro psych, with what the replications actually showed.
Bored One Minute, Overwhelmed the Next from Big Think. A workmanlike read on attention as the underlying tax on modern life.
Why I Don't Write Every Day from Fish Food for Thought. A pushback against the King-Murakami discipline gospel that does not collapse into excuses.
EQ Is Not Just About What You Say from Wes Kao. Body language as the other half of executive communication, with concrete tactics.
When Models Disagree, Transcription Accuracy Improves by Mark Humphries at Generative History. A real result about handwritten-archive transcription that has broader implications for any LLM ensemble work.
Outside Interests
There's Injustice in Palestine That Needs to Be Repaired from Vittles. Nisreen Fox interviews chef Fadi Kattan and Sara-Asad Mannings on doing hospitality work during the Gaza war.
The Future of Summer Camp, Post-Camp Mystic from Anne Helen Petersen's Culture Study. A long podcast on what summer camp becomes after the Camp Mystic flooding, paired with the NYMag feature she points at.
Ambient Scrutiny and iPad Rages from Casey Lewis at After School. Off Campus is now the third most-watched Amazon Prime series debut of the year, which is its own data point.
Polymarket Users Long A24's Backrooms from Emily Sundberg at Feed Me. Sundberg on prediction markets crossing over into entertainment release windows.
Weekly Digest: May 27 from Tetragrammaton. Rick Rubin's weekly culture pull. This week: Paul Klee's "Juggler in April" and Robin Spalding on water.
Data Worth Noting
Numlock News: Flowers, Bees, Honey by Walt Hickey. The USDA is decommissioning the Beltsville Agricultural Research Center over a $500 million maintenance bill, with knock-on effects for honeybee research that few people are pricing in.
Capital Is Flowing Again, Just Not Where the Consensus Says from Tearsheet's daily brief. Citi Wealth's new "Citi Sky" AI teammate, built with Google, as a tell on where the next capital deployment goes.
A Machine God from John Ellis at News Items. A short, sharp read on Cade Metz's reporting and the cultural register that the AI conversation has now drifted into.
Three Takeaways for You
The Texas Senate primary is a regime change in slow motion. Trump enforced loyalty over a four-term incumbent and won by a wide margin; the price of that victory is a general-election candidate who Republicans privately think can lose Texas. That is the bet now: Trumpism keeps the base but bleeds the seat. Watch James Talarico's fundraising over the next two weeks for the first real signal.
The AI conversation has decisively moved on from model selection to infrastructure and labor. Pieces by Mitcham, Noah Smith, Guillermo Flor, and the Pragmatic Engineer interview with Dax Raad are all making the same argument from different angles, and the Vatican's encyclical is the surprise frame that ties it to the labor question. If you build for a living, picking a model is the easy decision; the hard one is what skill, what context, what handoff, what human stays in the loop.
If you only read three pieces today, I would suggest: Bill Kristol on Texas Republicans for the political stakes, Max Mitcham on why the LLM doesn't matter for the AI frame, and Maritime Analytica on Hormuz for the unglamorous read on the war that nobody quite admits is still ongoing.