SpaceX-Cursor Deal, SaaS Debt Bomb, New Apple CEO, SPLC Indictment, Colon Cancer

Key points

• Sacks visited Trump at the White House before recording. Reports Trump on Anthropic: 'brilliant guys, great product, very left-wing — something we can work through. They tried to tell the Pentagon what to do, which the Pentagon didn't like.' Trump also keeps saying 'I like high-IQ people' as a stated core conviction.
• Trump's behind-the-meter data-centre policy (proposed >12 months ago, before the issue got political) is held up as the reason the US AI build-out has any chance. Counterfactual: the alternative would have been Bernie-style shutdown.
• SpaceX-Cursor: same deal as covered on 20VC, but framed via M&A mechanics. Chamath's read: 'It looks like Elon got a 50% discount.' At a $2T valuation, $60B of SpaceX stock for Cursor is effectively $30B in tomorrow-dollars. Plus he gets a model service, RL patterns, and a cracked team.
• On Composer 2: Cursor's in-house model now ranks between GPT-5.4 and Opus 4.6 on coding leaderboards. Hard to imagine SpaceX paying $60B and not running on Grok in time, but Friedberg notes developer optionality (model toggles) is part of why Cursor wins UX.
• Friedberg's key thesis: **the IDE wins the AI arms race.** Reason: enterprises are spinning up millions of agents. Amazon alone has a million-agent sprawl story this week — redundant data, redundant API calls, money being wasted. The fix requires centralised software-engineering competency + a unified front-end (read: an IDE). Cursor sits at exactly that nexus.
• Token-spend explosion inside enterprises is real and not yet rationalised. Frontier models for frontier tasks; cheaper/open-source for mundane. Middleware to route between models is the next category to be built — explicit echo of Replit Masad's Society of Models.
• Cyber as the white-hot centre within the red-hot centre of coding. Mythos is the wake-up call but is too expensive (10T-parameter rumour, Anthropic doesn't have enough compute to serve it commercially). Expectation: every major lab will train a smaller, cheaper, dedicated cyber model in the next 3-6 months.
• Polymarket consensus: 74% chance the Cursor deal closes; 80% chance SpaceX IPO by end of August.
• Tim Cook → John Ternus succession at Apple. Same news as on 20VC.
• Side thread on Elon's 'XAI was not built right the first time around' tweet from 5 weeks back. Foundational rebuild now in progress. Compared to Tesla's similar arc.
• SaaS debt-bomb mention (alluded in title): the late-2024 / early-2025 cohort of SaaS companies that took on debt for buybacks and AI capex are now staring at refinancing into a higher-rate environment with no growth reacceleration.

Notable quotes

Themes

• The SpaceX-Cursor deal as the year's defining acquisition
• IDEs winning the AI arms race
• Cyber as the hottest sub-category of coding
• Trump White House as a tailwind for the AI build-out