🌟 Editor's Note
Welcome to the first issue of Cool Architecture, a newsletter covering the cold cool world of software architecture. You might be wondering, why another newsletter?
Vibrant communities have matured around things like DevOps, frontend development, and AI. Yet, architecture has largely been forgotten outside of small circles. Why is this the case? Architects are pretty cool and smart, but communities take time and effort to build. Most architects, being senior in their organizations, have little time to spend outside work, family, and hobbies.
We believe having a community (even a small one) can be valuable for architects. We’re also pretty stubborn and like to build, whether that’s a product or something more abstract like a newsletter.
Thanks for being a part of the beginning. Expect a healthy dose of opinions, interesting reads, upcoming events, and interviews with architects. And more importantly, if you want to be a part of the community we’re building, please reach out to us. It’ll take some time, but we’re excited to create something together.
As a small token of appreciation for the early readers, we’re giving away one free item from our merch store to the first 100 subscribers. Without ado, let’s dive in.
🤔 LLMs for architecture diagrams — Are they useful?
AI coding tools are all the rage today, but what about using LLMs for software architecture? We tested a few popular models (GPT-4o, Claude Sonnet, Sonar, and Grok 3) and got them to create C4 diagrams for a hypothetical product recommendation service at a bank.
Diagram generated from ChatGPT for a product recommendation service at a bank
We noticed that the responses from the four LLMs tend to fixate on generating the Code diagram, while struggling to produce Context, Container, and Component diagrams. The models did provide key concepts with some explanations such as the banking gateways, data ingestion pipelines, model inferencing, etc. However, they fell short in explaining the reasoning behind their architectural choices. The models also describe the same technology choices for implementing the key concepts like AWS API gateway, SageMaker, PostreSQL, Redis, and S3 in their Container/Component diagrams. It seemed like the LLMs were selecting popular technologies based on patterns seen in other designs, rather than applying contextual logic or reasoning.
We also observed that the LLMs often operated in a “perfect world” setup, designing architecture without any real-world constraints. The designs lacked external factors like technology restrictions, budget limitations, or change of requirements. It seemed easy to corner the LLM and cause it to get stuck when we asked follow-up questions or changed the requirements. 🪦
While the LLMs did identify system actors for the banking system, they failed to account for the engineering team maintaining that system. How should they deploy their code to production, how to roll out newer versions of the recommendation service, or how data should be exported for further analytics.
Based on what we have, here’s our take on LLMs:
They design software diagrams like junior or mid-level programmers
They lack pragmatic thinking in architectural decisions
They assume ideal scenarios and fail to cover edge cases or evolving requirements
TL;DR: we should only trust software architects in their output, not LLMs. For now… 🙅♂️
🧊 Ice-breaker with Fardjad Davari
A running series of interviews with architects and tech leads.
Fardjad Davari
💼 Lead software engineer at BIMcollab
🎮 Video games and curiosity got him into coding in 1997
🤝 Passion for socio-technical systems and collaborative software design
We spoke with Fardjad late last year, a talented software engineer and architect. Our chat focused on documenting architecture in a large organization.
Want to break the ice? Drop us a line — [email protected].
🤓 Reads
From Restaurants to Retail: Scaling Uber Eats for Everything — Learn how the Uber team handled massive scale by building INCA (INventory and CAtalog) to manage product data and inventory. Read more
Building Production-Ready Agentic Systems: Lessons from Shopify Sidekick — Shipping production-level agents isn’t as simple as plugging into an LLM. Learn how Shopify dealt with the complexity with just-in-time instructions and rigorous testing. Read more
MAAD: Automate Software Architecture Design through Knowledge-Driven Multi-Agent Collaboration — Researchers from China and Singapore developed a multi-agent framework for software architecture design. It outperformed a comparable baseline against MetaGPT. Read the paper
vFunction 2025 architecture report — Nearly two in three organizations (63%) reported that architecture is integrated in development, but more than half (56%) have out-of-date documentation that doesn’t reflect the architecture in production. Download report
The Architect’s Paradox on LeanPub — How do we make decisions as architects? What drives our thinking? Barry O’Reilly’s introduction to philosophy for software engineers and architects is an interesting read. Lean more
🗓️ Events
We’re hitting the road in September and October. If you still have some conference budget to spend, check out these out. We’d love to see you there and say hi! 👋
Prairie Con 🇨🇦
Prairie Dev Con started in 2010 to bring a software conference experience to the Canadian prairies.
Date: September 22-23, 2025
Location: Winnipeg, MB
GSAS 🇪🇸
3-day event to connect software architecture experts worldwide and those interested in building working software.
Date: October 13-15, 2025
Location: Barcelona, Spain
Stay chill till next time,
