Europe 2026
Mergeable by default: Building the context engine to save time and tokens — Peter Werry, Unblocked
Overview
This talk introduces the concept of a "context engine" for AI agents, which aims to provide agents with the necessary information to perform tasks effectively without overwhelming them with irrelevant data. The core idea is to move beyond simple access to information and achieve true understanding, enabling agents to operate with the efficiency and insight of a seasoned team member. The presentation debunks common myths about context engines and shares lessons learned from building and implementing such systems.
Who should watch
- AI Engineers
- Product Managers
- Developers building AI-powered tools
- Anyone struggling with AI agent performance and context management
- Teams looking to improve developer productivity with AI
Key takeaways
- A context engine supplies AI agents with relevant information and excludes irrelevant data in an optimized manner, leading to more streamlined task execution.
- Naive Retrieval Augmented Generation (RAG) over documentation is insufficient as a context engine; it often leads to "satisfaction of search" where agents stop looking once they find a plausible answer, missing crucial information.
- Larger context windows alone do not solve the problem, as organizations often have more data than can fit, and reasoning across vast amounts of information remains a challenge.
- A context engine should understand organizational context, including team structures, expert knowledge, and historical decision-making, to resolve conflicts and provide accurate insights.
- Key features of a context engine include unified system context, conflict resolution, adherence to access controls, and delivering the right context at the right time for token efficiency.
- Implementing a context engine can dramatically reduce task completion time and token usage; one example showed a reduction from 2.5 hours and 21 million tokens to 25 minutes and 10 million tokens.
- Hard lessons learned include optimizing for understanding over mere access, surfacing rather than hiding conflicts, and avoiding caching answers due to the dynamic nature of code and documentation.
- Context engines are highly valuable during the planning phase of projects, for ticket enrichment, triage, incident management, and customer support.
Notable quotes
Context engineering is the art of supplying all the context that you need and most importantly, none of the context that you don't need in a highly optimized way.
Access doesn't equal understanding.
*Satisfaction of search is a real problem in radiology, and there's lots of protocols to prevent just stopping as soon as you find the first thing. This is what happens with agents.*
Unofficial community note. Prefer the recording for nuance.