🎧Listen to the deep-dive radio (15–20 min) for this storyRio × マクル・ドーキン in conversation, offline playback, and full scripts — in the app
Listen in the app →
U.S. Weighs Ban on Chinese Open-Source AI — Anthropic's Regulatory Capture vs. Developer Freedom
Guests
Jason Calacanis (host) × David Sacks (Craft Ventures, former White House AI & Crypto Czar), David Friedberg (Ohalo founder), Chamath Palihapitiya (Social Capital founder)
Bottom line
Banning open-source AI would destroy U.S. competitiveness and amount to regulatory capture by Anthropic/OpenAI — imposing a "token tax" on American enterprises while the rest of the world uses cheaper alternatives.
3-Line Summary
China's Moonshot AI released Kimi K3, matching Anthropic/OpenAI performance at half the cost, prompting White House debate on banning Chinese open-source models (Polymarket odds: 45%). Former White House advisor Sacks states "no decision made yet" but warns a ban would be a "tragic mistake" — Anthropic is at $70B ARR (mid-2026) and doesn't need government protection; punishing American developers for using public-domain models is contradictory. This fight will determine whether AI value accrues to a government-protected duopoly or diffuses across the economy.
3 Key Points
1. Anthropic's Double Standard Exposed: Anthropic argues it can train on all the world's content under fair use, yet tells the government that Chinese companies training on Anthropic's outputs constitutes "IP theft." But in their February blog post coining "industrial-scale distillation attacks," they never used the term "IP theft" — only framed it as a national security threat (Sacks, 00:51:00). Meanwhile, they trained on 7M pirated books (LibGen) and settled for $1.5B, the largest copyright settlement in U.S. history.
2. Models Commoditizing "Faster Than Anybody Thought": Chamath notes "no sustained advantage once a model publishes performance criteria" — real value is shifting to the application layer and infrastructure (cloud, chips). Banning open source would force U.S. companies to pay 50-100x more for AI than global competitors, causing capital markets to downgrade American firms (00:10:13). Google's 25-year average ROIC of 32% matters here — model fragmentation benefits Google (makes money on cloud, silicon, apps regardless of which model wins).
3. American Open Source Already Depends on Chinese Models: The best American open-source model (Thinking Machines) was distilled from Chinese Kimi 2.5; Cursor's Composer 2 post-trained on Kimi K2.5. If Chinese models are "tainted," all derivative works are tainted — "a dagger through the heart of the American open-source ecosystem" (Sacks, 00:40:42). Open Router token share shows over 50% from open-source models (many Chinese).
Editorial Perspective
This fight is a replay of 1990s Netscape vs. Firefox. Back then, Netscape's proprietary browser/server was crushed by open-source Firefox/Apache, and the open internet enabled Google, Amazon, and eBay to flourish. Today, the choice is the same: a government-protected duopoly or an open economy where value diffuses to millions of enterprises. As Friedberg notes, open-source AI answers Bernie Sanders/Elizabeth Warren's concerns about AI inequality — value flows to the entire industry, not to 2-3 companies and their billionaire shareholders. However, China's long game (commoditizing the knowledge economy to dominate the molecule economy, where it has 20x manufacturing capacity and 8x electricity production) cannot be ignored. Will the U.S. succumb to the temptation of regulatory protection, or embrace competition and win at the application layer? History favors the latter.
Source: All-In Podcast "The Fight Over Open Source AI, Anthropic's $1.5B Payout, NYC Socialists: Evictions = Violence?" (2026-07-24)
https://allinchamathjason.libsyn.com/the-fight-over-open-source-ai-anthropics-15b-payout-nyc-socialists-evictions-violence
AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.
# Kimi K3# Open Source AI# Anthropic# Regulatory Capture
🎧Listen to the deep-dive radio (15–20 min) for this storyRio × マクル・ドーキン in conversation, offline playback, and full scripts — in the app
Listen in the app →
Why VC Consensus Themes Always Miss the Winners — The $100M Seed Trap in the AI Era
Participants
Alex Wilhelm (This Week in Startups) × Aileen Lee (Cowboy VC, coined "unicorn"), Mike Maples Jr. (Floodgate), Ben Learer (Lerer Hippo)
Bottom line
No company in history that raised $100M+ at seed has exited above $10B — massive early rounds eliminate optionality and destroy founder wealth.
Three-Line Summary
Three seed investors (managing $130M-$230M funds) warn that 2026's AI boom replicates past bubble structures: three-week-old companies raising $20M seeds at $100M pre-money, optimizing for model capabilities that will be obsolete in months. Meanwhile, VC annual themes (2021 crypto, 2022 future-of-work) have never correlated with actual winner formations — Anthropic founded during crypto hype, Wiz during remote-work obsession. The panel argues multi-stage firms are "king-making" companies in crowded categories while ignoring harder, defensible businesses that need patient capital.
Three Key Points
1. "Acceptance AI" becomes the scarce resource: As generative AI floods the world with "slop" (low-quality content), value accrues to "credibly neutral third parties" certifying correctness — the audit firms of AI outputs. Compliance infrastructure companies like Drata are positioned to capture this (Mike Maples).
2. Fine-tuning is futile: Portfolio companies achieving "equivalent results at a fraction of the cost" using GLM-5.2 (Chinese open-weight model) vs. frontier models. Model versions evolve so fast that by the time you finish fine-tuning version 2.5, versions 2.6 and 2.7 are already out (Aileen Lee).
3. Rule of 70 banks profit first: KeepSafe maintains growth% + margin% = 70, depositing profit targets monthly before running the business on the remainder. From a $1.5M Floodgate investment, the founder has dividended out $10M-$20M over the past decade (Mike Maples). In contrast, companies raising $100M+ Series A rounds will see "collapses begin" in the next 12-36 months (Ben Learer).
Editorial Perspective
Reading this through our signature lens of platform transition cycle analysis, the 2026 AI funding environment mirrors late-2000 dot-com dynamics with eerie precision. The biggest exits clustered in the late-1998 to mid-2000 window; founders who raised massive rounds at peak valuations "we wouldn't know their names today" (Mike Maples). History suggests the real winners are being built quietly now in areas no one is watching. The new variable this time: model evolution velocity — in a world where fine-tuning completes after the next version ships, the very definition of defensibility is being rewritten. The question is whether investors can resist the gravitational pull of consensus long enough to find those invisible builders.
Source: This Week in Startups "Why the VC Hype Cycle Always Gets It Wrong | VC Roundtable | E2307" (2026-07-01)
https://4a885955-6823-4b22-a3e1-a526c25516a5.libsyn.com/why-the-vc-hype-cycle-always-gets-it-wrong-vc-roundtable-e2307
AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.
# Floodgate# Cowboy VC# Mike Maples Jr.# Hype Cycle
🎧Listen to the deep-dive radio (15–20 min) for this storyRio × マクル・ドーキン in conversation, offline playback, and full scripts — in the app
Listen in the app →
Apple's Across-the-Board Price Hikes Reveal AI's Permanent Reshaping of Consumer Hardware Economics
Participants
Marques Brownlee (MKBHD) × David Imel (tech journalist) × Andrew Manganelli (producer)
Bottom line
These price increases mark not a temporary supply shock, but the point where AI infrastructure demand permanently restructures consumer hardware economics.
Three-Line Summary
Apple announced sweeping price increases in July 2026 across Macs, iPads, and accessories (Mac Studio Ultra +$1,300, Apple TV +54%), attributing them to unprecedented RAM shortages driven by AI data center demand. iPhones remained unchanged, and hosts noted the increases don't correlate with RAM amounts—suggesting strategic demand management rather than pure cost pass-through. Samsung preemptively launched the Galaxy Fold 8 Wide (matching rumored Apple folding iPhone specs) to position its premium Ultra model favorably. PlayStation will end physical disc production January 28, 2027, intensifying concerns about "digital serfdom" (platforms can delete purchased content without refunds).
Three Key Points
1. Price hikes are strategically uneven, not cost-based — MacBook Neo (high-volume) rose only $100, while Mac Studio Ultra (low-volume) jumped $1,300. Brownlee observed: "Some products are higher volume and the price didn't go up as much... some went up more than they had to in order to make up for other prices not going up." No correlation between RAM capacity and price increase, implying cross-subsidization to manage optics and demand elasticity.
2. Apple may be pricing out ML hobbyists, not just managing costs — Mac mini saw earlier sellouts because buyers used it for machine learning, not personal computing. Manganelli asked: "Does Apple see these price increases as... what is our best chance of people still buying the products to use them as actual personal products, not AI machines?" Suggests Apple wants to reclaim margin from unintended use cases that don't drive services revenue.
3. Physical media extinction creates preservation crisis — Sony ends game disc production January 2027; Wii, Wii U, 3DS, Xbox 360 stores shut down in recent years. Sony recently deleted purchased movies from PlayStation Store (licensing expired, no refunds). Guest Moriah highlighted the Video Game History Foundation's archival efforts and warned indie games face the highest risk when storefronts close.
Editorial Perspective
本来の Gutenberg Lab では珍しい内容ですが、this price hike echoes the early internet era when server demand drove up consumer PC prices. The difference: back then, prices fell once supply caught up. This time, Apple appears to be choosing permanent repositioning toward higher tiers. History shows companies rarely lower prices voluntarily—instead, next-gen products justify the same price with better specs. The death of physical media fits the same pattern: companies frame cost-cutting and control-tightening as "environmental benefits," narrowing consumer choice. This isn't technological progress; it's market power consolidation.
Source: Waveform Podcast "Apple Raises Prices on Everything!" (2026-07-03)
https://www.example.com/waveform-apple-price-increases
AI Disclosure: Produced with AI assistance; facts and analysis reviewed by our editorial team.
# 2026 RAM Shortage# Apple# Apple Silicon# Folding iPhone
The world's intelligence, in 15 minutes every morning.
Audio, deep-dive analysis, and your own knowledge graph — in the app. Free today.
Get the app →