World's Fair 2026
Beyond the Harness: A Journey Towards Adaptative Engineering - Rajiv Chandegra, Annicha Labs
Overview
This talk introduces adaptive engineering as a new philosophy for AI development, moving beyond fixed, pre-defined harnesses. It argues that as AI models become more powerful and interact with the dynamic, messy real world, static harnesses will become brittle. Adaptive engineering allows the harness to emerge and adapt during the engineering process, with the engineer's role shifting to designing constraints rather than dictating every step.
Who should watch
- AI engineers and builders exploring new paradigms for AI development.
- Product Managers and engineers focused on shipping AI products in complex, real-world scenarios.
- Those interested in the future of multi-agent systems and collaboration.
- Individuals facing challenges with brittle AI systems in dynamic environments.
Key takeaways
- Current AI engineering relies on fixed harnesses that pre-define agent roles, tools, and sequencing, which are effective for predictable, complicated problems.
- The real world is dynamic and complex, characterized by multi-agent, multi-human, and cross-institutional interactions, making fixed harnesses insufficient.
- Adaptive engineering proposes that harnesses should emerge and adapt mid-process, allowing for greater flexibility and optimal positioning in response to changing environments.
- The engineer's role shifts from pre-defining the entire structure to designing initial constraints and then sensing and responding to the emergent harness.
- Complex systems, unlike complicated ones, are characterized by diverse, locally interacting, and recursively learning agents, leading to emergent properties not predictable from individual parts.
- Adaptive engineering emphasizes horizontal intelligence—how groups of agents coordinate—over vertical intelligence, which focuses on individual agent improvement.
- The engineer can influence adaptive systems by adjusting constraints such as the rate of coupling between agents, rewarding coherence, or setting guardrails.
- While adaptive engineering offers agility, potential failure modes include drift, monoculture, reduced legibility, and a lack of predictability.
Notable quotes
*The factory method is the right answer to a fixed problem and the wrong answer to a moving problem.*
*Adaptive engineering is the discipline of designing constraints to the extent that the harness emerges on its own, stabilizes, and adapts as needed.*
Unofficial community note. Prefer the recording for nuance.