All-In PodcastTrump's First 100 Days, Tariffs Impact Trade, AI Agents, Amazon Backs Down
CHAPTERS
- 0:00 – 2:30
Cold Open: Flights, Private Jets, and Bestie Banter
The episode opens with jokes about private flights, missing planes, and the All-In crew’s internal economics before introducing the guests and topics. The tone is light and self-deprecating, setting up a contrast with the heavy political and economic content to come.
- •Jason jokes about needing to wrap to catch a flight to Miami and teases Chamath and Friedberg about private vs commercial travel.
- •The familiar All-In intro plays, reinforcing the brand and running gags (Rain Man Sacks, Queen of Quinoa, etc.).
- •Ryan Petersen (Flexport) and Aaron Levie (Box) are introduced as returning guests for a policy-heavy episode.
- 2:30 – 8:30
Setting the Stage: Guests, Tariffs, and Political Labels
Jason welcomes Ryan back and briefly revisits Flexport’s CEO transition before turning to tariffs as the reason for his appearance. Aaron jokes about not really being a ‘fan favorite’ and recounts being labeled a ‘Biden-loving soy boy’ in the comments, while positioning himself as a pro–free market Democrat.
- •Ryan describes reclaiming the CEO role at Flexport and notes how tariffs have reintroduced major uncertainty into global logistics.
- •Aaron describes internet backlash to his last appearance and embraces the role of free-market advocate from the left.
- •The crew riffs on leftists embracing Milton Friedman and the stock market, framing ideological realignments around Trump and markets.
- •Early mention of Trump Derangement Syndrome vs Trump Defender Syndrome foreshadows polarized takes on the administration.
- 8:30 – 15:00
DC Power Circles and the New Trump-Aligned Private Club
Conversation shifts to Sacks’s government role and a new Trump-aligned private club in DC backed by Don Jr., with Sacks and Chamath as members. Jason laments not being invited and the group uses the club to illustrate changing Republican factions and social infrastructure around Trump.
- •Sacks explains he’s working roughly half-time in government and badges in and out like a normal employee.
- •Jason needles Sacks and Chamath about their membership in a new DC private club; he jokes about MAGA hats and high dues.
- •Sacks clarifies there are expensive founding memberships and more modest tiers; Chamath confirms he is one of the top-tier founders.
- •The club is positioned as ‘younger, hipper, Trump-aligned Republican’—a counterweight to old, Bush-era DC clubs and a safe space away from ‘fake news reporters’ and lobbyists.
- •Aaron jokes about starting a ‘Kamala club’ as a satirical mirror.
- 15:00 – 23:00
Grading Trump 2.0: First 100 Days Rundown
The panel offers overall assessments of Trump’s first 100 days, touching on border policy, executive orders, markets, and cultural shifts. They highlight divergence between Trump and Biden/Kamala trajectories and set up the central disputes around tariffs and economic strategy.
- •Jason frames the milestone: 143 executive orders, major indices down ~7–10%, 10-year yields down ~40 bps.
- •Ryan introduces the ‘OODA loop’ concept, arguing the administration is moving so fast that opponents, journalists, and mainstream Republicans can’t mount organized resistance.
- •He notes everyone expected tariffs given Trump’s explicit affection for the word, but the sudden implementation after ‘Liberation Day’ (April 2) shocked importers who already had goods in transit.
- •Aaron highlights AI as a bright spot: US policy has become decisively pro–open source and pro–domestic AI innovation, though tariffs and weak high-skilled immigration policy are headwinds.
- •He argues an alternative universe was available: keep doubling down on what’s working in the economy and AI while using targeted fixes instead of broad tariff shocks.
- 23:00 – 31:00
Chamath’s Report Card: Border, FDI, Tariffs vs Transparency and Comms
Chamath gives Trump’s first 100 days a B+, citing A+ performance on foreign direct investment and border enforcement and strong marks for tariffs and market resilience. He criticizes the administration’s failure to release various promised document troves and poor communication around tariffs and markets.
- •Chamath claims US FDI commitments are approaching or exceeding $1T, calling it a legacy that will outlast Trump.
- •He praises the border policy for driving illegal crossings effectively to zero, separating that outcome from deportation tactics.
- •Tariffs earn an ‘A’ in his view because they expose dangerous dependence on China’s brittle supply chain while markets and rates remain relatively stable.
- •He grades document releases (Epstein, MLK, full JFK) a ‘D’ due to delays and lack of explanation, and tariff communication a ‘C’ for not preparing markets and businesses for volatility.
- •Netting these, he lands at a B+ and predicts history will see these early moves as important structural progress.
- 31:00 – 41:00
Sacks on Border, DEI Collapse, and Reprivatizing the Economy
Sacks, speaking as both a bestie and administration insider, highlights three big achievements: ending the border crisis, collapsing DEI/wokeism, and beginning to ‘reprivatize’ the economy by cutting government bloat, deregulating, and empowering private innovation. He contrasts this with Biden-era narratives and spending.
- •On the border, Sacks says Trump’s restoration of Remain in Mexico and other policies sealed the border far faster and more completely than anyone would have predicted, disproving claims that new laws were needed.
- •He frames the Biden administration as gaslighting the public about the border crisis for years, then claiming their hands were tied by Congress.
- •On culture, he asserts ‘wokeism has completely collapsed,’ with DEI programs ended in government, disparate-impact-based affirmative action shelved, and meritocracy/colorblindness reasserted, with universities as the main holdouts.
- •He credits Trump with ending Biden’s AI executive order, stopping the ‘war on crypto,’ rolling back green projects like offshore wind, and unleashing oil and gas.
- •He argues these moves are setting up a future ‘Trump boom’ though the effects will take time to show.
- 41:00 – 48:00
JCal’s Centrist View: Wins on Border and Spending, Concerns on Rule of Law
Jason positions himself as an independent searching for consensus. He gives Trump a ‘B’ vs Biden’s ‘C–’, praises the border crackdown and efforts to rein in government waste, but flags risks around economic uncertainty, rule-of-law norms, and conflicts of interest.
- •He argues most Americans want the border secured and fentanyl stopped; Trump delivered where Biden gaslit and obfuscated.
- •DOGE (cutting government overspending and fraud) is another big win aligned with public sentiment.
- •He worries that rapid, opaque policy changes create economic uncertainty that makes business planning difficult, especially around tariffs and trade.
- •He flags due-process concerns around deportations, talk of a third Trump term, and scandals like meme-coin grifts as mirroring the Hunter Biden discomfort on conflicts of interest.
- •Jason emphasizes the ‘peace dividend’—Trump’s moves to de-escalate in Ukraine and avoid larger wars—as crucial, urging focus on economic growth and peace as Trump’s core mandate.
- 48:00 – 57:00
War, Nukes, and the Ukraine Off-Ramp
Chamath and Sacks zoom in on foreign policy, especially nuclear risk and the Ukraine war. They argue Trump is uniquely fixated on avoiding nuclear catastrophe and has pulled the US back from an escalatory spiral with Russia that Biden and Kamala likely would have continued.
- •Chamath recounts a recent dinner with Trump where he invoked his uncle’s lessons on the horrors of nuclear war; this reinforced Chamath’s confidence that Trump will always seek an off-ramp to avoid existential conflict.
- •He notes that in a nuclear conflict, debates about tariffs, documents, or DEI become irrelevant, elevating nuclear avoidance above all other policy issues.
- •Sacks reconstructs the Biden-era escalation path in Ukraine: initially warning that Abrams tanks, F-16s, ATACMS, HIMARS, and strikes on Russian soil might trigger ‘World War III’ or ‘Armageddon,’ then gradually delivering all those capabilities.
- •He cites New York Times reporting that US generals and intelligence were deeply embedded in planning Ukrainian strikes, effectively making the US a co-belligerent.
- •Sacks credits Trump’s team (e.g., Steve Wykoff) with re-opening direct diplomacy with Russia for the first time in years and halting the slide toward WWIII.
- 57:00 – 1:07:00
Tariffs Hit the Real Economy: Flexport’s View from the Trenches
The discussion returns to tariffs with Ryan detailing how the new rates are impacting import flows and supply chain operations. They unpack timing, hacks like bonded warehouses, and the bleak vs optimistic paths for trade and small businesses.
- •Ryan reports a ~60% decline in ocean-freight bookings from China to the US after reciprocal tariffs jumped to ~154%—much higher than anyone planned for.
- •Tariffs apply to goods departing China after April 9 (Eastern), and most ships now arriving are subject to the new rates.
- •Importers must pay duties at or soon after arrival, but bonded warehouses let them defer payments and potentially benefit from future tariff reductions.
- •A second ‘hack’ involves sending goods into Mexican or Canadian bonded warehouses, then importing into the US later when tariffs may be lower.
- •Ryan outlines the bleak scenario: tariffs stay high, escape hatches close, and many small import-driven businesses—especially apparel, consumer brands, Amazon sellers—go bankrupt because China’s role is now about manufacturing ecosystem quality, not just cheap labor.
- •He notes that even with prior 25% tariffs, many firms couldn’t relocate because China’s ecosystem remains unmatched for certain categories.
- 1:07:00 – 1:16:00
Levie’s Critique: Mixed Signals, Whack-a-Mole, and Missed Acceleration
Aaron Levie attacks the incoherence and execution of the tariff strategy. He argues the administration is sending mixed signals about whether tariffs are revenue tools to offset income taxes or bargaining chips for freer trade, while running an ad hoc, whack-a-mole approach to business fallout.
- •Levie says messaging is contradictory: some officials frame tariffs as long-term revenue replacements for income tax (implying permanence), while others frame them as temporary leverage for better trade deals.
- •He notes public statements from people like Howard Lutnick about the ‘end state’ of the economy are not accompanied by a clear, compelling vision, leaving businesses unsure if the future means Americans on assembly lines or next-gen jobs orchestrating robots.
- •He stresses the importance of ‘change management’—leaders must paint a vivid, credible end state to rally teams and capital, instead of issuing sudden shocks.
- •Levie’s alternative playbook: fire Navarro, issue a mea culpa, lean into massive carrots (Stargate, TSMC, NVIDIA chip fabs, 5% tax rates for onshore manufacturing), accelerate permitting/deregulation, and surgically use tariffs only in truly strategic sectors (chips, pharma, AI) rather than across-the-board disruption.
- •He emphasizes that in a free market, multinationals already choose supply chains optimally; heavy-handed government redirection risks central planning and economic distortions conservatives would normally denounce.
- 1:16:00 – 1:30:00
Media, Messaging, and the OODA Loop: Who’s Failing Whom?
The group debates whether the chaos is mainly due to Trump’s rapid-fire OODA loops or a media ecosystem obsessed with trivial controversies and blind to substantive policy details. They cite examples like Wall Street Journal coverage of Tesla governance rumors vs tax incentives for factory PPE.
- •Chamath accuses mainstream media of falling into Trump’s traps and then neglecting real policy: they fixate on MS-13 tattoos and blurred lawn placards while virtually ignoring impactful tax changes like full expensing of PPE and incidentals for factories, which his wife’s pharma company could use.
- •He argues that such details dramatically change ROIC/ROE for many S&P 500 firms but get no coverage, while sensational (and allegedly false) Tesla board stories make front pages.
- •Levie responds that Fortune 500 CEOs listen more to Goldman and JPMorgan than MSNBC, and the administration still bears responsibility for coherent change management even if the media is lazy or biased.
- •Ryan ties this back to tight OODA loops: Trump’s team rolls out a policy, hears from impacted industries (e.g., autos), then rapidly patches in exemptions—creating a perception of chaos but also evidencing real-time course correction.
- •They agree communication is muddled and that the burden is on both the administration and media to convey precise, actionable policy details, not just high-level vibes or partisan narratives.
- 1:30:00 – 1:42:00
Amazon, Temu, and the Tariff Line-Item Fiasco
A brief segment covers Amazon’s attempt to surface tariff/import charges at checkout, the political backlash, and how this incident encapsulates the wider whack-a-mole dynamic in tariff policy. The panel brainstorms what a more strategic approach could look like.
- •Temu already itemizes import charges; Amazon began to experiment with similar tariff disclosures but quickly reversed course after intense criticism, including being called ‘treasonous’ by the White House press team.
- •Jason and Ryan see itemizing tariffs as good consumer transparency and a potential tool to nudge buyers towards US-made goods (no tariff line) vs Chinese goods (visible tariff cost).
- •Aaron argues this is further evidence that policy is being managed via one-off interventions and presidential phone calls, rather than a coherent framework—Amazon got a direct call; thousands of smaller retailers will not.
- •Chamath focuses on Amazon sellers’ emails claiming that Amazon favors foreign suppliers (especially Chinese) that can circumvent US rules, undercut local sellers, and evade enforcement due to lack of US legal presence.
- •Ryan notes US law allows foreign companies to import without a US entity, making it easy to misclassify goods or safety standards while competing on price, and hard for regulators to punish offenders.
- •The crew sees this as a major policy opportunity: reform import rules, de minimis thresholds, and marketplace accountability to restore a fair playing field for domestic sellers.
- 1:42:00 – 1:55:00
Macro vs Micro: Are We Over-Correcting or Finally Acting?
The tariff debate crescendos as Aaron insists the strategy could have been executed with zero ‘shock and awe,’ while Chamath and Sacks argue that absent a dramatic break, entrenched globalist inertia would have continued. They spar over timelines, market reactions, and what counts as success.
- •Levie references Scott Bessent’s own critique of Bidenomics—deregulate, grow GDP, then reduce taxes and spending—and argues Trump could have used that accelerationist path instead of chaos.
- •Chamath counters that many of those deregulatory ‘carrots’ are already being pursued, but media coverage is skewed towards drama, so executives often don’t even know about them.
- •Sacks points out that Larry Summers’ main argument against tariffs in their prior debate hinged on a temporary market downturn; since ‘Liberation Day’, markets are actually up, undermining claims that the policy is obviously failing.
- •He stresses it’s too early to declare defeat: the Trump trade team (Bessent, Lutnick, Greer) must still ‘stick the landing’ by converting leverage into concrete deals—Lutnick has hinted a parliament-approved deal is already queued.
- •Levie insists that the initial market freak-out and political pressure are precisely what forced Trump to walk back the harshest tariff details (e.g., auto parts), so you can’t say ‘no need to panic’ when panic triggered the fixes.
- •Chamath underlines national security stakes: China strategically annihilated US rare earth and magnet capacity via WTO-enabled subsidies, leaving the US dangerously exposed in EVs and motors, and this was never seriously addressed until now.
- 1:55:00 – 2:04:00
National Champions, Supply Chains, and Strategic Optionality
The besties step back to the geopolitical big picture: China’s national-champion model, potential conflicts over Taiwan or India, and how tariff shocks are finally forcing the US to confront structural vulnerabilities. Aaron continues to warn against ‘chaos monkey’ policy despite agreeing with the goals.
- •Chamath emphasizes scenarios where China could weaponize US dependencies: cutting off pharma APIs, rare earths, batteries, or critical components in response to US siding with Taiwan or India.
- •He points out that China’s ‘national champions’ blur boundaries between public and private, law and capital, giving them strategic advantages the US has not matched.
- •He insists the US needs its own answer—if not identical, at least a strategic industrial policy—and that such topics only became mainstream after Trump’s tariff moves.
- •Levie concedes he agrees on the strategic aims (chips, batteries, APIs, AI) but argues you don’t need a mass, sudden supply-chain pivot to address a relatively small portion of imports; targeted, well-telegraphed moves would suffice.
- •Sacks pushes back that for 25 years almost no one in power challenged the globalist trade regime; many critics now offering ‘perfect plans’ were silent while vulnerabilities deepened.
- •Jason eventually calls time on the circular argument and shifts the conversation to AI, noting they’ll have to ‘agree to disagree’ on process while mostly sharing end goals.
- 2:04:00 – 2:13:00
AI Agents 101: From Chatbots to Full-Fledged Digital Workers
The conversation pivots to AI agents as 2025’s defining tech trend. Sacks explains the Manus demo paradigm and how agents will plug into many SaaS tools, while Jason and Aaron give concrete enterprise examples of agents already being deployed.
- •Sacks describes Manus’s two-pane UI: a chat window plus a live view of the agent working across search, browser, code terminal, and document editor, building a to-do list and checking items off.
- •He emphasizes that Manus didn’t invent agents but advanced the UI paradigm; the deeper innovation is standards like MCP that let agents understand data and actions across many SaaS apps.
- •Jason shares his venture firm’s agent use case: ingesting 20,000 startup applications and investment updates per year, auto-linking them with past notes, and surfacing suggested questions and competitor insights to partners via Slack.
- •Aaron reframes AI from simple co-pilots to ‘agentic systems’ that can operate indefinitely across large data and tool surfaces to accomplish complex tasks end-to-end.
- •He predicts this will massively expand software TAM because AI agents now target labor budgets: instead of selling 10 lawyer seats, you might sell AI capacity equivalent to 20 paralegals’ work.
- •Both stress that most immediate value is in new work that was previously too expensive to do—e.g., mass contract review, multilingual marketing campaigns—rather than replacing existing employees one-for-one.
- 2:13:00 – 2:20:00
Practical Enterprise AI: Flexport Calls Truckers, Box Automates Knowledge Work
Ryan and Aaron give specific, grounded examples of how AI agents are already creating value in logistics and knowledge work without wholesale job replacement. They also correct a doom-heavy narrative by emphasizing expansion of the pie rather than pure substitution.
- •Ryan explains Flexport’s AI system calling hundreds of thousands of truck drivers daily to match loads to drivers via phone, a function that was too costly even with offshore call centers but is now economically viable.
- •He notes the AI calls drivers, not customers, so the reputational risk of an occasional weird conversation is manageable compared to the upside in engagement and utilization.
- •Aaron argues that Valley discourse overemphasizes displacement: in practice, enterprises are mostly using AI to tackle ‘never-done’ work (unreviewed contracts, ignored invoices, missing campaigns), not to fire staff.
- •Jason echoes this from venture: they now use AI to compare legal docs and detect term deltas they previously would have skipped or outsourced to external counsel.
- •The panel suggests that over time, AI will cause job role shifts (e.g., lawyers focusing on higher judgment tasks) more than a simple reduction in headcount, especially in the near term.
- 2:20:00 – 2:28:00
Deterministic vs Probabilistic Software: The Compliance Wall
Chamath lays out a fundamental challenge: regulated industries are built on deterministic software with clear audit trails, while LLMs are probabilistic and can hallucinate. This creates a huge need for new QA disciplines—‘improvement engineering’—before agents can fully penetrate high-stakes domains.
- •Chamath notes that in sectors like finance, healthcare, and life sciences, errors aren’t just bugs—they trigger enforcement, fines, or legal liability (e.g., bad KYC lets a wire slip to Syria; clinical misclassification delays a life-saving drug).
- •Replacing deterministic, stepwise code with probabilistic LLM outputs requires far more emphasis on QA, testing, and monitoring than traditional software—suddenly the least glamorous role becomes the most critical.
- •He warns of forthcoming class actions once model-driven failures cause real-world harm, and argues that we’re only beginning to understand the legal infrastructure needed for accountability.
- •Ryan adds a concrete example with tariff classification: even high-90s accuracy in HS codes leaves non-trivial error exposure—but because customs uses simple decision trees today, if they adopted an LLM as their system of record, its outputs would in effect become ‘truth’.
- •Chamath introduces the term ‘improvement engineering’ for teams focused on error-rate reduction and guardrails, comparing it to his creation of growth teams at Facebook—a new specialty central to value creation.
- 2:28:00 – 2:40:00
Exponential AI Progress: Models, Chips, Compute, and the Million-X Question
Sacks lays out why he thinks we’re nowhere near a hype peak technologically. He details how algorithms, chips, and data center scale are all improving at exponential rates that compound multiplicatively, while Aaron grounds this with empirical eval results and cost caveats.
- •Sacks separates change management from tech progress and argues the latter is still in early innings: LLMs have evolved from basic chatbots to reasoning models and now agents, with each iteration qualitatively expanding capabilities.
- •On hardware, he traces NVIDIA’s road map from H100 to H200 to GB200 and beyond (Rubin, etc.), with each generation roughly 3–4x more powerful and enhanced with innovations like NVL72 rack-scale systems.
- •Data center build-out is scaling from 100MW to 300MW to gigawatt-class facilities; GPU clusters at companies like xAI and OpenAI are moving from 100k to projections of millions of GPUs in a few years.
- •At roughly 10x every two years across models, chips, and deployed compute, Sacks estimates a potential million-fold effective increase in available AI compute over a presidential term, though some of that will show up as lower costs rather than raw capability.
- •Aaron tempers this with current evals: even top models get ~90% accuracy in complex document extraction in single passes; higher reliability requires multi-pass strategies, chunking, reasoning models, and highly tuned prompts.
- •They note that areas with objective validation (code, math) are improving fastest, while domains like law or medicine will progress more slowly due to harder ground truth and validation costs.
- 2:40:00
Closing Banter: Flights, Private Jets, and Bestie Sign-Off
Jason wraps the episode, again joking about missing his flight vs the other besties’ private travel. They end on their usual irreverent soundboard clips and inside jokes, underscoring the mix of high-stakes policy analysis and informal camaraderie that defines the show.
- •Jason insists he has to leave for Miami and teases Chamath and Friedberg for charging private flights to the All-In budget while he flies Southwest.
- •Sacks brags he hasn’t missed a flight in ~15 years; the group riff on what ‘missing a flight’ even means to private jet flyers.
- •Jason signs off by reasserting his tongue-in-cheek title as ‘world’s greatest moderator’ and reintroduces running gags (Rain Man, Queen of Quinoa, ‘wet your beak’).
- •The show closes with the usual audio stingers about besties, orgies, and dogs defecating in Sacks’s driveway, signaling a return from policy depth to comedic brand identity.
