Descriptions:
Nate B Jones makes the case that computing has undergone a categorical shift — from the instruction as the unit of work to the token — and backs that claim with a dense set of financial data that most AI coverage has overlooked. He opens with StrongDM CTO Justin McCarthy’s disclosure that his three-person engineering team deliberately targets $1,000 per day in token spending with zero handwritten code, then layers in the broader picture: Cursor’s AWS costs doubled from $6M to $12M in a single month following Anthropic’s priority service tier launch, Anthropic itself spent $2.66B on AWS through September 2025 against only $2.55B in cumulative revenue, and Perplexity burned through 164% of its entire 2024 revenue across AWS, Anthropic, and OpenAI combined.
From that economic foundation, Jones constructs a framework for three diverging developer archetypes emerging in 2026. The Orchestrator specifies outcomes and manages AI budgets rather than writing code, with compensation likely to correlate with token spend rather than lines of code. The Systems Builder does deep infrastructure work — building the agent frameworks, eval pipelines, and routing layers that orchestrators rely on — a smaller, more specialized, and highly compensated track. The third type is the person who doesn’t yet know they’re a developer: someone using AI to produce functional software without any traditional engineering background.
The video also covers the sharp drop in per-token inference costs — somewhere between 10x and 200x per year depending on benchmark — and argues that token economics is now a core business competency. Companies that fail to master it, Jones warns, are one supplier pricing change away from a crisis, pointing to Cursor’s decision to build its own model as a direct response to that vulnerability.
📺 Source: Nate B Jones · Published February 20, 2026
🏷️ Format: Deep Dive
