The Week the AI Story Shifted

What It Covers

A weekly recap analyzing how the AI narrative shifted across economics, Wall Street, and infrastructure during one week in 2025, covering the Anthropic-SpaceX partnership, enterprise deployment challenges, job market data from a16z and Ezra Klein, and OpenAI's new voice models in the Realtime API.

Key Questions Answered

• •AI Job Displacement Data: A16z's David George shows that on public market earnings calls, companies mention AI augmentation over substitution at an 8-to-1 ratio. Historical data since 1850 shows labor markets diversify rather than collapse during technological shifts — nail salons, pet care, and exam prep each grew from under 100,000 to 150,000–350,000 workers post-1990 productivity gains.
• •Enterprise Deployment Gap: Both OpenAI and Anthropic launched separate enterprise deployment joint ventures — valued at $10B and backed by $4B, and $1.5B respectively — with partners including Blackstone and Goldman Sachs. This signals that closing the capability-to-deployment gap requires dedicated infrastructure investment, not just model advancement, and timelines likely span decades rather than years.
• •Compute Supply vs. Demand Reality: BlackRock CEO Larry Fink stated AI faces supply shortages, not a bubble, with demand growing faster than anticipated. Carmen Lee's framework explains why: capital moves fast, but GPUs, power substations, cooling, and fiber each carry independent lead times, meaning a compute bubble requires every physical bottleneck to clear simultaneously — a near-impossible condition.
• •Anthropic-SpaceX Infrastructure Logic: XAI, folded into SpaceX, holds substantial compute capacity but lacks competitive frontier models, while Anthropic holds strong models but limited compute. The partnership — giving Anthropic full capacity of Colossus One — follows a clear resource-exchange logic. Elon's TerraFAB chip manufacturing project in Texas is now projected at $55B–$119B, far exceeding earlier $20–25B estimates.
• •Codex Slash Goal Meta-Prompting Technique: OpenAI's Codex slash-goal feature enables persistent, multi-hour autonomous coding sessions. To maximize output, avoid writing the slash-goal prompt manually. Instead, ask a separate AI model to research the slash-goal feature, review your project context, then generate three detailed slash-goal prompts optimized for your specific use case before running them in the Codex CLI.