jared

Massive GPU overbuild creates cheap compute abundance like 1990s fiber glut enabled YouTube; startups are the YouTube equivalent benefiting from infrastructure overinvestment by incumbents.

The addressable market for dev tools expands from ~20M human developers to hundreds of millions of 'vibe coders' plus their agent swarms; winning requires optimizing documentation (LLM.txt, structured code snippets) for agent consumption, not human readability.

Regulatory barriers (CEQA), power shortages, and land scarcity force AI infrastructure into orbit; space fusion (Zephyr) and space data centers (Star Cloud, Google, Elon) form a new infrastructure stack.

Boom Supersonic repurposes jet engines for data center power; Helion pursues fusion; Zephyr Fusion targets space-based fusion; 2-3 year lead times for power equipment create structural energy bottleneck for AI buildout.

Harvey locks up VC capital to block competitors; Gamma achieves $100M ARR with 50 people; post-Series A hiring remains robust as customer expectations rise faster than AI productivity gains.

General-purpose models are commodities; durable value lies in domain-specific 'harnesses' that imbue models with expert knowledge, taste, and workflows — as Anthropic did for coding with Claude Code, and Ploy is doing for marketing.

AI coding tools exploded as a category in 2025 but cannot yet ship 100% production code; the behavior of multi-model arbitration (Claude, Gemini, GPT-4) is becoming standard for developers and startups alike.

90% enterprise AI project failure stems from 90% of enterprises not knowing IT; this organizational drag acts as a brake on fast takeoff, giving society time to adapt to log-linear scaling improvements.