Summary

The Trump–Musk alliance became concentrated business risk once Musk attacked Trump’s budget bill. The House-passed, Senate-debated package would eliminate an EV tax credit, while Musk separately failed to secure an FAA deal for Starlink. Casey Newton’s blunt interpretation: “This is not what he thought he was paying for when he tried to buy the 2024 election.”
Musk’s escalation exposed Tesla, SpaceX, and potentially Musk himself to retaliation from an administration he had empowered. Tesla fell 14% that Thursday as Trump threatened to terminate “billions and billions of dollars” in subsidies and contracts; Musk answered with Epstein allegations, support for Trump’s impeachment, and talk of a new party. Immigration scrutiny, a congressional subpoena, and contract cancellation remained speculative but newly plausible pressure points.
Trump also loses an important political asset: Musk served as the administration’s “heat shield” while DOGE dismantled parts of government. The next fault line is a Washington “friend divorce” among Musk-aligned figures such as David Sacks and the remaining DOGE staff. Casey suggested those employees could suddenly be treated as an “insider threat,” creating the possibility of a DOGE purge.
Sahil Lavingia’s 55-day DOGE stint complicates the premise that federal agencies are simply warehouses of obvious waste. Inside Veterans Affairs, he was “surprised by how effective the government is”; contracts that initially looked outrageous sometimes made sense once explained. His sharper diagnosis was that Congress behaves like “the worst PM you’ve ever met,” forcing agencies to execute awkward specifications while “playing Twister.”
DOGE paired useful technical scrutiny with an operating model that often left that expertise underused. Lavingia arrived without knowing his salary, boss, or exact role, then received a laptop that could not run Python or Git; he nonetheless challenged a $4 million chatbot and asked contractors whether AI productivity gains could support a 25% discount. DOGE could attend meetings and “be annoying,” but at the VA it could not fire people or act “unless VA is willing to give us the keys.”
Lavingia still defends modernization, but his concessions validate the hosts’ critique of DOGE’s “ready, fire, aim” method. Government must serve everybody, so efficiency is beneficial only after objectives and constraints are correctly defined; “modernization or simplification” is his preferred framing. He acknowledged that the effort was “less than ideal” and relayed another DOGE engineer’s assessment that “mistakes were made,” while retaining faith that courts and other checks can reverse some apparent damage.
Elite chefs are using ChatGPT less as an automated creator than as a tireless source of ingredients, techniques, and provocations. Grant Achatz called it his “favorite kitchen tool,” using it to surface fuels such as avocado pits and corn cobs; Ned Baldwin used it to explore sausage texture and Malaysian seasoning. The investable product insight is behavioral: generic first answers become valuable through iteration, specificity, and human refinement, while high-end execution remains beyond the tool “at the sort of high-end creative level.”

Deep dive

1. The budget bill exposed the price of Musk’s alliance

Kevin Roose and Casey Newton recorded around 5:00 PM Eastern on Thursday, June 5, while the feud was still accelerating. Casey moved from initial cynicism about how real the conflict was to concluding by midday that it was “extremely real.”
The trigger was Trump’s “one big beautiful bill,” already through the House and under Senate debate. Musk called it “a disgusting abomination,” turning a policy disagreement with his former ally into an unusually public attack.
Casey identified two business-specific disappointments: the bill would eliminate an electric-vehicle tax credit valued by Tesla owners, and Musk had reportedly sought but not obtained an FAA deal to use SpaceX’s Starlink infrastructure.
Musk then claimed credit for Trump’s presidency and Republicans’ congressional majorities. Kevin immediately remarked that “Trump is not gonna like that,” capturing the collision between the two men.

2. The feud repriced Musk’s political and financial exposure

Trump said Musk had been “wearing thin” and threatened the heart of his portfolio: “The easiest way to save money in our budget, billions and billions of dollars, is to terminate Elon’s governmental subsidies and contracts.”
Musk retaliated by talking about Trump being mentioned in the Epstein files, sharing a 1992 video of Trump and Jeffrey Epstein, endorsing a post calling for Trump’s impeachment and replacement by JD Vance, and proposing a new political party.
Tesla fell 14% that day. Casey would not be surprised by further declines as investors processed the conflict, while Kevin emphasized Musk’s additional billions of dollars in exposure to SpaceX government contracts.
The hosts treated personal legal jeopardy as possible, not probable. They noted Steve Bannon’s call to investigate Musk’s immigration status, reporting that Musk began his U.S. career working illegally, and a House Oversight subpoena that Republicans had moved to block.

3. Trump loses his heat shield as the tech right chooses sides

Casey’s institutional framing: Musk had functioned as Trump’s “heat shield,” absorbing blame for DOGE’s unpopular dismantling of federal operations. Without him, Trump must own more of the administration’s decisions directly.
Kevin saw a possible route back because Musk has burned bridges with Democrats while Trump controls the Republican Party. Yet Musk may have simultaneously alienated both Democratic Tesla buyers and the MAGA customers who might have bought Teslas or Cybertrucks.
Washington now faces what Casey called a “major friend divorce.” Musk allies such as David Sacks can preserve influence by affirming loyalty to Trump, while DOGE personnel deemed loyal to Musk could be recast as an “insider threat” and subjected to a “DOGE purge.”

4. Lavingia joined DOGE to ship code and entered without a map

Sahil Lavingia had long viewed government as a high-impact destination for a software engineer. Private businesses felt like “the little jewel boxes on Mission Street,” while public infrastructure powered everything around them.
He had applied to the U.S. Digital Service around 2015, after Gumroad failed to raise a Series B and laid off most of its team, but never heard back. DOGE appeared to offer the missing chance to ship government software.
The pre-inauguration vetting included whether he had voted for Kamala Harris; he was told a Harris voter should not waste time. Lavingia replied that he had voted for neither Harris nor Trump and guessed the political filter came from Trump’s side rather than DOGE itself.
He received no conventional offer or salary discussion and initially had no named boss or clear role. He knew only to report to VA headquarters at 7:45 AM on March 17; even afterward he assumed his salary was zero, while occasional de minimis checks suggested otherwise. His government laptop could not run Python or Git or download anything — “cooking without any equipment.”

5. Veterans Affairs complicated DOGE’s theory of obvious waste

DOGE’s orientation centered on cutting costs, while VA onboarding made Lavingia an agency employee. His practical filter required both savings and a better veteran experience — “or at least not make it worse.”
What surprised him was “how effective the government is” at accomplishing what it is instructed to do. The deeper problem was congressional specification: Congress resembled “the worst PM you’ve ever met,” leaving agencies “playing Twister” to satisfy strange requirements.
Lavingia entered some meetings asking why a contract cost $40 million and left thinking, “Oh, I get it.” He also found a comparatively developed VA technology organization, built partly through earlier USDS work and already using GitHub and Slack.
At a DOGE all-hands, he proposed livestreaming meetings to show that employees were not secretly “scheming” to cause harm. Musk enthusiastically promised to start the next week, but Lavingia received no further invitation and saw no evidence that it happened.

6. DOGE could interrogate contractors but could not drive the agency

A Trump executive order let DOGE sit in on contracting meetings. Lavingia’s value was being someone who had actually coded: when told VA paid $4 million for a VA.gov chatbot, he could answer that he might build and integrate it in a week, then demand an explanation.
With firms such as Booz or Deloitte, his questions were concrete: “Do you guys use Cursor?” If AI improved their productivity, could they pass the gain to government and “shave like 25% off the bill?”
The authority was narrower than DOGE’s public image. At the VA, it could attend, question, and annoy, but not independently fire employees: “Unless VA is willing to give us the keys, like, we can’t drive the car.” Even the reduction-in-force order asked for recommendations and gave DOGE “zero power” to execute layoffs.
Lavingia did observe a genuine culture clash. Only two or three engineers he encountered were over 30, many lacked substantial private-sector experience, and his own startup sarcasm failed spectacularly: entering a workplace anxious about layoffs, he joked, “I’m here to riff everybody.”

7. DOGE misunderstood government careers and discarded its own recruit

The Deferred Resignation Program offered what Lavingia viewed as an extraordinary eight months of severance, yet uptake was drastically lower than expected. DOGE missed that many civil servants intended to work in government for life and genuinely loved their jobs; “an extra eight months” simply did not compute.
Challenged that joining DOGE was naive, Lavingia embraced the charge: “All startup people are naive.” Most bets fail, but an occasional breakthrough — “an iPhone or AI” — can make the “maybe” worth taking; he had time and preferred trying rather than opting out.
His leading explanation for being fired was a Fast Company interview he had not run by DOGE. Nobody formally informed him: GitHub emailed that he was no longer part of the VA organization, his access stopped, and DOGE personnel “just kind of ghosted me.”
The termination made him feel he had not been imagining DOGE’s problems. He described an “emergent behavior of lack of empathy” in which everyone could blame someone else, and concluded that “my wife was right.” The White House defended political vetting for political appointees; the VA did not answer the show’s termination questions.

8. Efficiency works only after government’s obligations are specified

Kevin’s challenge was that governmental friction can be intentional: agencies must serve constituents, reconcile conflicting priorities, and deliberate before changing lives. A government is not a startup whose sole objective can be reduced to speed or cost.
Lavingia’s synthesis was conditional: “If you define the goals accurately, then efficiency is better.” Spending less, moving faster, or requiring fewer forms is beneficial only while meeting every obligation — especially because government programs must serve everybody.
He preferred “modernization or simplification” to the now-loaded word efficiency. Government frequently stacks new agencies and systems atop old ones instead of refactoring them; three structures created in different eras might now be merged without abandoning their objectives.
Kevin’s pushback — worth keeping: modernization should be slow, careful, and expertise-heavy, while DOGE used “ready, fire, aim.” Lavingia called the results “less than ideal” and relayed another DOGE engineer’s assessment that “mistakes were made,” but said he retained faith that courts and other checks sometimes stop the apparent gun from actually going off.

9. Leading chefs treat ChatGPT as another collaborator

Pete Wells began reporting because chefs were strangely silent about AI. Grant Achatz of Alinea, Next, and Fire immediately broke that silence: “I use it nonstop. It is my favorite kitchen tool.”
For Fire, where food would be cooked with flames, coals, or embers, Achatz asked ChatGPT for unusual fuels used around the world. It returned avocado pits and corn cobs, alongside the commercially less attractive suggestion of cow dung.
Achatz’s adoption follows an existing collaborative method. Long before Alinea, his chefs traded ideas after service; suggestions from a sous chef, travel, or somebody else’s childhood memory were debated, tested, altered, and refined rather than placed directly on the menu.

10. Iteration extracts novelty, while authorship remains taboo

Houseman chef Ned Baldwin supplied another workflow: long conversations about making sausage firmer, softer, or springy like Asian fish balls, followed by seasoning questions such as how the recipe might change if made in Malaysia.
Several chefs used the phrase “out of the box.” Restaurant pressure pushes people toward combinations they already trust; an endlessly available chatbot can interrupt those formulas before the quality that once made a chef “fresh and exciting” disappears.
Kevin’s pushback was that AI cooking suggestions often feel statistically generic. Pete agreed the first answer can be obvious, then demonstrated the remedy: sardine-and-fennel pasta became a lower-carb dish, then a raw fennel salad, then a Turkish direction with braised zucchini or fried eggplant. “The more specific you can be, the more surprising it will be.”
Resistance remains rooted in the romantic chef who stands “on a mountaintop” until lightning becomes dinner. Pete said AI cannot yet perform the high-end chef’s job, but one food writer still told him, “I hate that AI piece” — meaning not the writing, but the reality it documented.

Kontrarian Korner's Ben Kelleran on Sable Offshore $SOC Asurion: 50X Season Two - (50X, S2)