The Way We Use AI is Changing

Nathaniel Whittemore on The AI Daily Brief (June 11 2026). The thesis is the one this episode contributes to the vault — the Advantage Gap between AI power users and casual users is widening fast, and OpenAI's planned ChatGPT super-app overhaul is best understood as a UX bridge to drag the median user closer to the power-user mode. The headlines half is government-equity-in-AI-labs (Bernie / Trump / Sacks) and the SpaceX-Google / Nvidia-SK Hynix supply-chain plumbing.

Key takeaways

1. Both Bernie Sanders and Trump are proposing government equity in major AI labs — and they're not that far apart. Bernie's: tax 50% of AI-company equity into a sovereign wealth fund. OpenAI's pitch (Sam Altman met Bernie this week): donate equity to seed a public wealth fund, possibly routed to Trump Accounts for children. Sacks: I can almost support it as a "stupidity tax for the AI job-apocalypse narrative" — but warns of central government AI as a CCP-social-credit risk. Brad Gerstner: shares to citizens directly (Trump accounts / pooled private trust) yes, government ownership no. Full context in Government Equity in AI Labs.
   • Whittemore's "Overton window has become a flapping Overton door" line.
   • Ras Rax optimistic frame: "This is how you make the AI revolution something the whole country can support."
2. SpaceX is now the largest neocloud on Earth. Google signed a 3-year deal, $920M/month, for at least 110K Nvidia GPUs (Oct 2026 → June 2029). Same structure as last month's Anthropic-SpaceX deal. SpaceX now ~550K GPUs (>2× CoreWeave); Starlink at $15B ARR — GPU rental is now SpaceX's biggest business per Yuchen Jin. Elon's xAI built ~$40B of data centers; rental revenue from Anthropic + Google ≈ $26B/yr → 18-month payback. The "Elon as Earl of Compute" framing.
3. Nvidia secured HBM via multi-year SK Hynix deal. Jensen Huang flew to Seoul (grilled pork belly + soju with SK chairman) — face-to-face supply-chain diplomacy is his year's playbook. Multi-year deal + design-partner status on next-gen memory chips for physical/personal/infrastructure AI. Vera Rubin chips ramping. Backs up Token Maxing's edge-AI thread; pair with the SK Hynix capacity-double in The Next Wave of Enterprise AI.
4. The ChatGPT super-app overhaul is OpenAI's UX bridge for the advantage gap. FT broke it as "biggest ChatGPT overhaul since launch" — coding tools, AI agents, external apps surfaced. Cynical reads: David Gewirtz "couldn't find the next big thing"; Hedge Markets "a chatbot is hard to put a multiple on — this is a feature for the S1." Whittemore's pushback: the investor class can't imagine a non-financial motive. He thinks the real driver is giving casual users the experiences power users are already getting.
5. OpenAI's own engagement data quantifies the advantage gap. Sarah Friar (CFO): free users ≈ 7 turns/day; Plus ≈ 15; Plus (the $20 tier) ≈ 3× free; Pro ≈ 11× free. Power users aren't just more — they're using AI differently.
6. Loops are the next abstraction beyond tasks — and people are getting frustrated trying to learn. Steinberger: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents." Boris Cherny on CNBC, six months ago: "I write loops, not code." From last week: "I have loops that are running. They're the ones that are prompting Claude. My job is to write loops." See Loops as Core Primitive. The gap is wide: Shanu Matthew's "non-technical idiot guy here — what does this mean?" post got 300K views; Cloudflare CTO Dane Catch got 250 replies on "how are people using loops." Whittemore's line on the difficulty: Jake retweets "both evidently the future and somehow gatekept… like pulling a neutron star out of a magic hat."
7. Vanguard → power user → casual user is the rollout shape. Vanguard (lab-internal) → loops; power user (Plus/Pro) → agents; casual user (free) → chat. Each cohort is opening up more distance, not closing it. Inflection: ~Nov 25–Jan 26 when coding tools went non-developer.
8. Seat → usage shift is now structural. $3M run rate → $47B run rate inside Anthropic. The chatgpt overhaul + Codex Sites + role-bundled plugins are all part of a UX-rooted bet that teaching people to want more agents is the route to defensible revenue.
9. What's missing in everyone's training material: how non-developers should actually use loops. Whittemore notes OpenAI's published list of Codex use cases is one strategy; the deeper play is changing the interface itself to lead the behavior. Open thread for the vault — the Cherny / Steinberger meme is repeated all over the AI internet but no one has the learn-to-write-loops playbook for non-engineers.

Connects to your work

• Advantage Gap is the load-bearing addition this episode makes; it's the conceptual frame that ties together your existing Code Is Free, Token Maxing, Default Shift, and the How Loops Are Improving Work — Sunil's Research Brief thread. Use it as the headline argument for why IT teams have to upskill on agent patterns now.
• Loops as Core Primitive is the missing concept for Turning This Mac Into an AI Operating System — your process-raw is exactly Felix Ryberg's responsibility pattern (loop watching crash reports → loop watching the raw folder).
• Engagement-multiple data (3× / 11×) is brand fodder for any internal AI literacy pitch — concrete, sourced, and matches the lived experience.
• The government-equity thread is downstream noise for now from a vault-utility standpoint — flagged but not pursued in concept pages.

Source

• source — YouTube, The AI Daily Brief, 2026-06-11
• Companion episodes (same week): Fable 5 Raises the Bar for AI Ambition, The Next Wave of Enterprise AI