W31 - "Context Engineering for AI Agents"

Manus’s blog post “Context Engineering for AI Agents” gave me a more pragmatic understanding of AI application development.

In the past it was easy to reduce AI application development to simply “calling an API.” I now realize the real challenge is how to apply traditional, deterministic software engineering methods to harness fundamentally uncertain AI. The engineering practices built from countless iterations of trial and error are the technical barrier for managing large-model uncertainty. This barrier cannot be judged by code or product demos alone; only conversations with engineers who have deep experience can truly validate whether a systematic advantage exists.

The core competitive capability in building applications is the ability to translate vague human intent into machine-executable workflows that run reliably.

A book on complex systems, Discovered, Not Designed, gave me a more coherent understanding of Context Engineering as a hard-to-articulate “art of deterministic orchestration.” The book divides problems into two types: difficult problems, which traditional engineering methods can explicitly design solutions for; and hard problems, which are often inexplicable and cannot be solved by explicit design but are addressed through continuous trial, error, and self-adjustment.

An industrial product like an airplane is complex but not a complex system; it belongs to the category of explicitly designed complexity. Large models, by contrast, are closer to complex systems — products that are “discovered.” When training large language models, we don’t directly change the model’s internal parameters by hand; we guide the model’s self-adjustment through external feedback to obtain the desired outputs. Similarly, market prices are not crafted by deliberate design but emerge spontaneously from the interactions and strategic behavior of countless traders.

Uncertainty is a blessing, not a curse. It’s important to understand the root of large-model uncertainty and place it in the correct problem domain. We’re facing a brand-new engineering field that is still in its early stages and involves a lot of concrete, gritty work. We must not cling to the deterministic mindset suited to difficult problems.