Build better AI agents with primitives

In a recent keynote at the Cognitive Human-Agent Interaction (CHAI) conference, Ahmad Awais articulated a compelling vision for AI agent development that challenges the prevailing wisdom of framework-first approaches. His presentation cut through the hype that surrounds many AI conversations today, focusing instead on a foundational principle that may reshape how developers build the next generation of AI systems. At its core, his message was deceptively simple yet profound: primitives trump frameworks when creating truly effective AI agents.

The essence of Awais's argument lies in recognizing that AI development faces a critical inflection point. As he meticulously explained through his own journey building AI agents, the most robust systems emerge not from rigid frameworks but from flexible, purpose-built primitives that allow developers to construct exactly what they need. This approach addresses a fundamental tension in software development—the trade-off between convenient abstractions and the control necessary to build truly innovative solutions.

• Primitives provide essential flexibility that frameworks often sacrifice for convenience, allowing developers to construct purpose-built solutions rather than forcing their requirements into predetermined patterns
• Current AI frameworks create unnecessary constraints by imposing opinions and structures that may not align with the specific problems developers are trying to solve
• The "primitives-first" approach enables more resilient systems that can evolve with emerging technologies rather than becoming obsolete when frameworks fall out of favor
• Real innovation happens at the primitive level, as demonstrated by Awais's work on DevOps.ai, where building from first principles led to more powerful and adaptable solutions

The most compelling insight from Awais's presentation was his distinction between "playing house" with AI frameworks versus doing the substantive work of building meaningful solutions from primitives. This perspective matters tremendously in our current AI landscape, where flashy demos and framework-dependent solutions often mask fundamental limitations. By focusing on primitives—the basic building blocks from which more complex systems can emerge—developers gain the ability to create solutions that actually solve real problems rather than merely demonstrating capabilities within artificially constrained environments.

When examining this approach in context, it's clear that Awais's philosophy aligns with broader shifts in software development. The industry has repeatedly witnessed cycles where frameworks rise to prominence before eventually constraining innovation, only to be replaced by new approaches built on