Latest Research
Download our detailed analyses on AI architecture, team economics, and emerging technologies.
The Educated Guess
DeepSeek’s DSpark system speeds up AI inference by 60–85% through speculative decoding — and how IPNS’s own AnticiPro uses the same "speculate, then verify" pattern to predict caseworker questions before they’re asked.
The Off Switch
Examines how U.S. export-control law and national security policy could let the government abruptly cut off access to frontier AI models, using Anthropic’s 2025–2026 DoD dispute as a case study, and what companies should do to reduce that dependency risk.
Stop Prompting Better. Start Rejecting. The Real AI Skill
This white paper argues that the most defensible competitive advantage in the AI era is not the ability to prompt AI effectively, but the organizational capacity to reject AI output with precision and capture those rejections as institutional knowledge.
The Sixth Person Catastrophe: How AI Destroyed Team Economics
Companies are watching their per-person output climb, yet the calendar is fuller than ever. This piece examines the paradox of increased AI tool adoption leading to more meetings and coordination overhead, and how to rethink team structures.
Context Layers for LLMs
This paper explores "context engineering"—curating the optimal set of tokens during LLM inference. Architecting the right information environment is the biggest lever for AI performance to avoid "context rot".
The Intent Layer: A Missing Architectural Primitive for Autonomous Agents
Context alone is insufficient for autonomous agents. This paper proposes the "intent layer", sitting between raw context and action execution, allowing agents to act with genuine organizational alignment.
The Capability Gap: Why Both the Doomers and Boomers Are Wrong About AI
Challenging extreme AI narratives—from apocalyptic job displacement to instant utopian productivity—and revealing where the actual economic and operational opportunities lie in the next decade.
Ready to transform your IT strategy?
Let's align your technology initiatives with mission-critical outcomes.