Agentic AI in SW Development: Evolving Patterns & Protocols

Item: Agentic AI in SW Development: Evolving Patterns & Protocols
Rating: 4
Author: PostSingularity Institute

by Bhuvaneswari Subramani

Applied Sciences & Engineering AI — News & Applications UYComputer science UYQArtificial intelligence (AI)

🎙 Bhuvaneswari Subramani 👥 1.1M 📅 December 29, 2025 ⏱ 23 min 👁 2K 📄 expert opinion 🧭 2026-08-06

Available in: English (current) Français

Keywords

Agentic AIMCPRAGSelf-correcting RAGConversational clients

Summary

In this GOTO Serverless 2025 talk, Bhuvaneswari Subramani presents a mental model for understanding the evolution of agentic AI in software development. She outlines seven foundational patterns, starting from simple conversational clients to sophisticated agent-to-agent interactions. The talk begins with a live demo using the Nova Act model to automate a web search and add an item to a cart, illustrating the power of natural language commands. She then traces the progression: conversational clients lack memory and context; contextual conversational clients incorporate context but are limited to small data; RAG (Retrieval-Augmented Generation) enables querying large indexed datasets; self-correcting RAG adds validation and grading using LLMs as judges; function calling agents allow interaction with external tools; agents with Model Context Protocol (MCP) standardize tool integration; and finally, agent-to-agent interaction enables collaboration between agents. She emphasizes the importance of validation and grounding in AI responses, and provides code examples and references to her blog series ‘The Agentic Shift’ for further exploration.

159 words

Critical Evaluation

The talk provides a clear and structured overview of the evolution of agentic AI patterns, which is valuable for practitioners seeking to understand the landscape. The speaker’s practical experience is evident, and the live demo effectively illustrates the potential of natural language-driven automation. The progression of patterns is logical, and each step addresses limitations of the previous one, offering a coherent framework. However, the content is largely based on the speaker’s own experience and blog series, lacking rigorous academic or industry validation. The talk is more of an expert opinion and tutorial than a critical analysis of the field. The sources cited are primarily the speaker’s own blog posts and LinkedIn, which limits the diversity of perspectives. The technical depth is moderate, suitable for a general developer audience, but not for those seeking deep implementation details. The adéquation between title and content is good, as the talk indeed covers evolving patterns and protocols. The presence of a short sponsorship segment (likely at the beginning) does not detract from the content. Overall, the talk is informative and well-structured, but it would benefit from more external references and a more critical examination of the challenges and limitations of agentic AI.

198 words

Title / Content Match

The title accurately reflects the content, which focuses on evolving patterns and protocols in agentic AI for software development.

Quality & Reliability

7/10

The speaker is a Chief Cloud Evangelist with practical experience, and the talk is based on her ongoing blog series. The content is well-structured and grounded in real-world examples, but it is largely opinion and experience-based rather than peer-reviewed research. The demo and code references add credibility, but the talk is not a formal study.

Chapters

Cited Sources

The Agentic Shift: Mapping the Landscape of AI System Design — Speaker's blog post that forms the basis of the talk
Speaker's LinkedIn post about the talk — Promotional post for the talk
Speaker's personal website — Personal website with more resources
Speaker's Dev.to profile — Speaker's articles on Dev.to
Speaker's Medium profile — Speaker's articles on Medium
GOTO Serverless 2025 conference page — Conference where the talk was given
Session abstract — Detailed abstract of the talk

Concurring Sources

Andrej Karpathy's talk on software development and AI — Referenced in the talk as inspiration for the evolution of AI in software development

External References

Contribution & Novelties

The talk provides a clear, structured framework for understanding the evolution of agentic AI patterns, which is useful for developers and architects. It emphasizes the importance of validation and grounding in AI systems, and introduces the concept of self-correcting RAG. The live demo with Nova Act showcases practical applications.

Pour aller plus loin :

Model Context Protocol (MCP) — Official documentation for MCP, a key protocol discussed in the talk.
Retrieval-Augmented Generation (RAG) — The original paper on RAG by Lewis et al., providing foundational knowledge.
LangGraph — Framework used for implementing the self-correcting RAG workflow, allowing for stateful agent workflows.

101 words

Radar Profile

The radar profile shows a balanced performance across all dimensions, with slightly higher scores in information quality and reliability, indicating a solid but not exceptional presentation. The technical level is moderate, making it accessible to a broad audience.

Reliability 7/10

💬 Sur les 0 commentaires analysés, aucune tendance n'est disponible.