Welcome to the AI Insights tribe. Join others who are receiving high-signal stories from the AI world. In case you missed it, Subscribe to our newsletter.Today’s newsletter is brought to you by The Rundown AI!
Get the latest AI news, understand why it matters, and learn how to apply it in your work. Join 1,000,000+ readers from companies like Apple, OpenAI, NASA.
Context Engineering is quickly becoming the next big thing in AI development. If prompt engineering was about writing the perfect input, context engineering is about designing the world your agent operates in — the data it sees, the tools it can use, and the state it remembers. If you’ve been hearing the term and wondering why it’s gaining traction, this is your guide.
Basics of → Context Engineering
Context engineering is about managing and optimising everything that shapes an AI agent’s output — from system prompts and conversation history to user preferences, retrieved knowledge, and long-term memory.
For most AI engineers, this isn’t a brand-new concept. It’s really the next step after prompt engineering — but with a more structured approach.
When you get context right, your agents respond more accurately, stay consistent across interactions, and make better decisions based on all the information available to them.
Context Engineering vs. Prompt Engineering: What's the Difference?
You may be wondering: "Isn't this just prompt engineering with a different name?" Not exactly! Here's the main difference:
Prompt Engineering focuses mainly on crafting the correct input text to get desired outputs from an LLM
Context Engineering encompasses the entire ecosystem of information that influences AI behaviour, prompts, memory, retrieved data, user state, tools, and more.
Think of it this way: prompt engineering is about perfecting the question you ask, while context engineering is about creating the entire environment in which that question is asked and answered.
In short, prompt engineering is a subset of context engineering.
You can also watch this talk by Chroma CTO Hammad Bashir to learn more about Context Engineering!
Why Context Engineering Matters?
Context engineering is crucial for building AI agents that can:
Make smarter, personalised decisions based on individual user history and preferences
Maintain consistency across conversations and sessions
Learn and adapt from past interactions without starting from scratch every time
Provide relevant responses by understanding the whole picture, not just the current query
Scale effectively as your user base and use cases grow
Without proper context and memory, AI agents treat every user like a stranger every single time. Imagine having to reintroduce yourself and explain your preferences to a human assistant every day; that's what happens with contextless AI.
Types of Context in Modern Agentic Systems
Understanding the different types of context is crucial for practical context engineering.
Here are the seven key categories:
1. Instructions/System Prompt: These are your agent's core personality and behaviour guidelines. It defines how the AI should act, including the tone to use, its capabilities and limitations, and its primary objectives.
2. User Prompt: The immediate input from the user, the question, request, or command that triggered the current interaction.
3. Retrieved Context/RAG: Information is pulled from external sources like vector databases, APIs, knowledge bases, or search results that are relevant to the current query.
4. State (Short-term Memory): Recent conversation history and temporary context that help maintain coherence within a session or recent interactions.
5. Long-term Memory: Persistent information about the user, their preferences, past decisions, and learned behaviours that should influence future interactions.
6. Tools: Available functions, APIs, or external services the agent can use to complete tasks or gather information.
7. Structured Output: Predefined formats or schemas that guide how the agent should structure its responses for consistency and integration with other systems.
We’re going to explore a new open-source Memory layer in upcoming section: https://memori.gibsonai.com
How to Manage Context Effectively
As context engineering continues to evolve, the big question becomes: how do we properly provide the proper context to AI agents and ensure they remember what matters?
A new framework called Memori has been developed to address the challenges of memory and context engineering in AI agents. Memori gives your AI agents human-like memory, helping them remember what matters, promote what's essential, and intelligently inject structured context into your LLM conversations.
Memori uses a multi-agent system with three specialised AI agents working together. The Memory Agent processes and structures every conversation intelligently, the Conscious Agent analyses patterns and promotes essential information, and the Retrieval Agent intelligently selects relevant context for seamless injection.
It implements four specialised memory systems working in harmony: Short-term Memory that lasts about 7 days and stores recent context and promoted essentials, Long-term Memory for permanent insights, preferences, and knowledge, Rules Memory for user-defined guidelines, policies, and constraints, and Entity Memory that dynamically tracks people, technologies, projects, and relationships.
The infrastructure is powered by GibsonAI and built on enterprise-grade systems that scale with your needs. Zero configuration required, just plug and play.
Tools and Frameworks
Building context-aware AI doesn’t always mean starting from scratch. Several frameworks and tools come with built-in capabilities or integrations that make context engineering easier:
LangChain – Provides integrations for RAG, memory systems, and agent orchestration, helping you manage both short-term and long-term context.
LlamaIndex – Makes it easy to connect your AI to structured data and knowledge bases for enhanced retrieval-augmented workflows.
Other Agent Frameworks – Offer pipelines and memory integrations that simplify storing and retrieving contextual information across multi-agent setups.
Chroma / Weaviate / Qdrant / Pinecone – Vector databases that help maintain semantic memory and retrieved context for AI agents.
OpenAI Functions & JSON Mode – Let you enforce structured outputs, track interactions, and integrate external APIs directly into your AI workflows.
These tools help abstract away much of the heavy lifting, letting you focus on designing intelligent, context-aware applications rather than building everything from scratch.
Wrap-Up
Context engineering is becoming a must-have skill for building truly intelligent AI apps. In short, you can:
✅ Move beyond stateless prompts and add memory to your AI
✅ Build context-aware agents that feel less robotic
✅ Use frameworks to manage state, history, and external knowledge
✅ Ship smarter, more human-like AI experiences
👉 Follow our Newsletter on X: Daily AI Insights
This Week in AI
NVIDIA Grace Hopper - NVIDIA introduced JUPITER, world’s most energy-efficient supercomputer
Kimi K2-0905 - Moonshot launched updated Kimi-K2 with enhanced coding and tool-call capabilities and 256k tokens.
EmbeddingGemma - Google dropped new open 308M parameters embedding model designed for on-device AI.
Hacktoberfest - Hacktoberfest countdown has started, landing page is live and registration will open soon.
VibeVoice Large - Microsoft released an upgraded VibeVoice Large ~10B Text to Speech model (MIT licensed).
Sukriya🙏🏼 See You Again Next Week! Till Then Keep Learning and Sharing Knowledge with Your Network