Author(s): Veera RS

Originally published on Towards AI.

AI can write the code. But these five engineering concepts determine whether it actually works.

Imagine being a software developer and not having to write any code. That sounds nice, doesn’t it? On teams at popular search engine and music streaming companies, that’s already the reality. AI and new AI coding assistants write 80% or more of their code. It’s incredible, and it’s happening because of AI-assisted development practices and what’s known as agentic engineering.

After the introduction, the article breaks down five core terms for building useful AI agents: prompt engineering focuses on communication and domain expertise (including few-shot and chain-of-thought examples for smaller models); context engineering is about managing what the model “sees,” using system prompts, tools/RAG inputs, codebase context, and techniques like compacting plus sub-agents; harness engineering provides the runtime and sandboxing that let agents act safely, verify results, and iterate using feedback from tests, merges, and other signals; loop and graph engineering describe how agents run repeatedly via triggers (often with a human-in-the-loop) and how multiple agents can work in parallel as a workflow graph; finally, the software factory applies these ideas to automate the end-to-end development lifecycle with isolated agents, clear identity/access, and strong observability.

Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI