5 AI agents, 5 contradictory bets, 3 questions that tell you which one fits — and the prompts to pressure-test your answer

Source: Nate’s Newsletter Date: 2026-03-23 URL: https://natesnewsletter.substack.com/p/5-ai-agents-5-contradictory-bets

Summary

Nate maps five distinct agent strategies across the current market — sovereignty/open-source plays, delegation-first products, distribution-through-scale bets, safety-positioned deployments, and generalist expansion from a vertical wedge — and argues the divergence is the signal, not a flaw. Rather than asking “which agent is best,” he offers three diagnostic questions — where the agent runs, who controls the model selection, and what the interface assumes about the user’s technical sophistication — as the actual selection criteria for teams choosing an approach.

Implications

• The framing that agent architectures are “contradictory bets” rather than a converging standard is useful for evaluating vendor positioning claims; a vendor who says their approach is obviously correct is hiding a bet.
• The three diagnostic questions (runtime locus, model ownership, interface assumptions) map directly onto build-vs-buy decisions for agent infrastructure: teams that can’t answer all three cleanly are not ready to commit to a platform.
• Feeds the agent patterns thread — the framework is a lightweight taxonomy for categorizing new entrants as they appear in release tracking.