About This Tech Radar
What Is This?
This is a collection of architecture deep dives — analyses of how leading companies build internal AI developer productivity systems, and how frontier models work under the hood. Each entry dissects a real system: how it works, what architectural decisions were made, what patterns are worth borrowing, and what trade-offs exist.
The radar is inspired by the ThoughtWorks Technology Radar and built with AOE Technology Radar.
How to Read the Radar
The radar is divided into four quadrants, each covering a different category of engineering system:
| Quadrant | What It Covers |
|---|---|
| Autonomous Coding Agents | AI systems that write, test, and ship code with minimal human intervention |
| Agent Frameworks & Runtimes | Open-source frameworks for building, orchestrating, and running AI agents |
| Enterprise AI Dev Platforms | Developer portals, AI coding assistants, and agentic dev platforms |
| Model Architectures | How frontier models work — plus practical guidance on when to use each one |
Each item on the radar sits in one of four rings:
| Ring | Meaning |
|---|---|
| Adopt | Proven at scale — these patterns are mature enough to adopt in your own org |
| Trial | Worth pursuing — ready for real-world use with some adaptation |
| Assess | Worth studying — cutting-edge approaches with valuable lessons |
| Hold | Proceed with caution — too company-specific, immature, or superseded |
For Model Architectures, the rings also carry practical recommendations:
| Ring | Model Guidance |
|---|---|
| Adopt | Use in production today — well-understood, reliable, good ecosystem support |
| Trial | Use on real projects — strong capabilities, worth committing to |
| Assess | Evaluate for your use case — promising but understand the trade-offs first |
| Hold | Wait — too new, too niche, or superseded by something better |
Why Deep Dives?
The most valuable lessons in developer productivity don't come from product announcements — they come from understanding how systems actually work at scale. These deep dives aim to:
- Extract transferable patterns — What architectural decisions can you borrow?
- Expose trade-offs — What did they give up to gain speed/reliability?
- Cut through hype — What's genuinely novel vs. good marketing?
- Inspire pragmatism — Most orgs can't build Stripe's infrastructure, but they can adopt the thinking behind it.
- Guide model selection — For model architectures, help engineers understand when a model's strengths align with their workload.
Editorial Note
- Human decides — Which systems to analyze, ring placements, and editorial judgment come from the maintainer.
- AI researches — Claude Code gathers sources, drafts analysis, and summarizes findings. Every entry is reviewed and validated before merging.
- Living document — Systems evolve. Entries are updated as companies publish new details or open-source their tools.