Since ChatGPT and other LLMs have taken the world by storm, Avraham Poupko has been reading, speculating, writing, and talking about how the LLMs might change the way we design software. We all know that LLMs are changing the world of software development dramatically.
Developers are increasingly using LLMs when creating software, leveraging more intelligent and more innovative tools. The growth in that area is likely to continue. Meanwhile, the use of popular resources such as Stack Overflow is dropping as developers are no longer searching for communities or examples; instead, they are just asking the LLMs to write the code. Many of these LLMs, such as GitHub Copilot, are embedded directly into IDEs.
To a large extent, the code being created is syntactically and functionally correct, often clear, and conforms to accepted standards. However, there are other aspects to software development beyond coding, such as design and architecture. We know that architecture is often the art of trade-offs. As such, when it comes to architecture decisions, there is no one correct answer; instead, a multitude of possible answers needs to be considered, and choices (often tough ones) must be made. Considering these trade-offs, arriving at the right decision is often considered to be the hard part of building software.
To explore how LLMs can be applied to architecture and design, Poupko has been experimenting with the concept of a "Virtual Think Tank". The complete article describes what a Virtual Think Tank is, and how to make use of it, along with an example that highlights some interesting results observed.
This content is an excerpt from a recent InfoQ article by Avraham Poupko, "The Virtual Think Tank: Using LLMs to Get a Multitude of Perspectives".
To get notifications when InfoQ publishes content on these topics, follow "Architecture & Design", "Culture and Methods", and "Sociotechnical Architecture" on InfoQ.
Missed a newsletter? You can find all of the previous issues
on InfoQ.
Sponsored
|
|
The rise of agentic AI is raising new demands on modern systems, requiring architectures that can adapt, scale, and recover under uncertainty. This guide explores proven principles and patterns for distributed applications across cloud, multi-cloud, and edge environments. Learn how to apply event sourcing, CQRS, and state partitioning to handle elasticity, latency, and async processing—building resilient, event-first systems ready for the next generation of AI-driven workloads.
Download the eBook “Principles and Patterns for Distributed Application Architecture (By O'Reilly) sponsored by Akka
|
|
