From Prompt Engineering to Context Engineering

Harness AI's ability with context engineering and bespoke tooling for optimal software development performance

Feb 19, 2026 6m 33s

Adam Ben-David

Expert Insights

Adam Ben-David steers the conversation towards the evolution of AI-assisted software development, emphasising a transition from model obsession to context comprehension. The potent blend of AI, domain knowledge, and specific action capabilities, he suggests, can be tuned to vastly improve a model's usefulness within a specialist environment.

Adam asserts that the focus should be on providing agents with high-leverage tools that enable them to manage tasks effectively. This shift to context engineering, he suggests, is more valuable than constantly iterating on different versions of the AI model itself.

Here's what Adam explains:

Why obsessing over model choices in an environment wherein models are converging can limit progress.
How the highest value lies in the quality of tools offered and the context management abilities of an agent.
Why innovation is leading towards agents having a remarkable level of access, such as querying a proprietary framework's source code.
How a new emphasis on context and tooling, rather than model selection, can improve the usefulness of a model within its specific environment.
Why the power of AI can be most effectively harnessed through the careful combination of domain knowledge, specific action capabilities, and a model-tailored context.

We are moving from obsessing over which model is best and testing all the different versions of the model to context engineering. So now, the really interesting thing, as I think from an engineering leadership perspective, is that you always try to focus on what is highest leverage, and that's now the tools we give to the agent and the context management.

Monterail Team Analysis

Here are actionable insights for teams seeking to optimise software development using AI:

Embrace Context Engineering: Move focus from improving AI models to comprehending and managing context. This involves developing tools and systems that make the most of your AI agent's abilities.
Invest in Agent Empowering Tools: Provide AI agents with tools that allow for efficient context management and problem-solving. The set of accessible tools could potentially determine the boundary of an agent's capability.
Capitalise on Bespoke Tooling: If you belong to a niche industry, don't settle for generalized tools. Build or adopt tools tailored for your unique requirements and challenges.
Nurture Domain Knowledge: Advanced tech, on its own, might not always translate to better products. Encourage teams to cultivate deep domain knowledge to extract the most value from AI.
Don't Over-Engineer Models: Constantly iterating on AI models might not be as beneficial as it seems. Spend those resources on tooling, structuring data and managing context.
Prioritize Tool Usability: Design interfaces to be easy to access, interact with and manage. The ease of use can significantly impact the effectiveness of your AI.
Leverage AI-INFUSED OS: For instance, revamp your IDE into an operating system. This can significantly improve the developer experience and productivity.