OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze

Summary

This All-In episode treats OpenAI’s missed user and revenue targets as a test of whether demand, compute supply, product momentum, and capital commitments can align. The hosts connect Codex, Claude Code, GPT 5.5 Cyber, AI Cyber-Defense Utility, Data Center Power Bottleneck, and hyperscaler capex into one AI infrastructure cycle, then move through OpenAI Nonprofit Conversion Dispute, Retatrutide, and Product Liability Federal Preemption in the Monsanto/Roundup litigation.

Key Claims

• OpenAI’s missed ChatGPT user and revenue targets may reflect both consumer competition from Google/Gemini and supply constraints from power, data-center construction, and compute commitments.
• David Sacks argues Codex regained coding momentum through a GPT 5.5 base-model upgrade while Anthropic and Claude Code faced user complaints about rationing and regressions.
• GPT 5.5 Cyber and Anthropic’s Mythos discussion strengthen the wiki’s dual-use cyber frame: models can discover existing vulnerabilities for defenders or attackers, so access and supervision matter.
• Hyperscaler capex from Amazon, Microsoft, Google, and Meta is framed as an asset-heavy AI infrastructure buildout where power, turbines, transformers, debt, and utilization matter as much as software margins.
• OpenAI Nonprofit Conversion Dispute is presented as a material governance and IPO-timing risk because Elon Musk’s lawsuit challenges OpenAI’s nonprofit-to-for-profit transition and names Sam Altman and Greg Brockman.
• Retatrutide is framed as a next-generation metabolic drug candidate that may intensify GLP-1 competition, body-composition use, and peptide enthusiasm while remaining dependent on clinical approval and monitoring.
• The Monsanto/Roundup Supreme Court segment frames Product Liability Federal Preemption as a conflict between federal EPA labeling authority and state failure-to-warn claims.

Key Quotes

“AI agents are valuable but must be supervised.” - Sacks on accountability for agentic software work.

“They don’t create vulnerabilities; they find vulnerabilities.” - Sacks on cyber-capable models.

Connections

• OpenAI, ChatGPT, Sarah Friar, Codex, Google, Gemini, Anthropic, Claude Code, and OpenAI Compute Strategy - model-company demand, revenue, and coding-product momentum.
• GPT 5.5 Cyber, AI Cyber-Defense Utility, Frontier Model Cyber Misuse, Cybersecurity AI Supervision, Palo Alto Networks, and CrowdStrike - cyber model capability and defender/attacker symmetry.
• Data Center Power Bottleneck, AI Infrastructure Debt Financing, Data Center Debt Risk, AI Compute Continuity, Oracle, Amazon, Microsoft, Google, and Meta - hyperscaler capex, power, and financing branch.
• OpenAI Nonprofit Conversion Dispute, Elon Musk, Sam Altman, Greg Brockman, Polymarket, and AI IPO Valuation - governance litigation and public-market timing risk.
• Retatrutide, Eli Lilly, GLP-1 Agonists, GLP-1 Body Composition Stack, Peptide Evidence Hierarchy, and Medical Risk Management - metabolic drug, peptide, and body-composition branch.
• Monsanto, Bayer, Roundup, Environmental Protection Agency, Supreme Court, Product Liability Federal Preemption, and Chevron Doctrine - pesticide labeling, administrative law, and state-law failure-to-warn branch.

Contradictions

• No settled contradiction found. The episode qualifies existing OpenAI compute optimism by adding missed-target and CFO-concern claims, but the hosts also argue that coding-product momentum and power scarcity may explain the gap. User counts, revenue targets, compute commitments, Polymarket odds, hyperscaler capex totals, GDP-growth claims, clinical metrics, and legal outcome predictions remain source-scoped.