AI-assisted architecture reviews
Testing how far a model can go reviewing event-driven designs before a human architect is required.
LLMEvent-DrivenDDD
Maverick · Jhones Maverick
Where architecture meets experimentation.
Maverick is my lab: an open space where I test AI-native development, distributed systems ideas and engineering practices in real conditions — then publish what actually worked.
8+
Years in production engineering
Staff
Current engineering level
Kafka
Event-driven systems at scale
Azure
Cloud and Kubernetes platforms
Principles
How the lab operates.
Think in systems
Every experiment starts from constraints, trade-offs and failure modes — not from hype.
Ship to learn
Small, disposable builds that answer one question each and get measured honestly.
AI as leverage
AI is tooling, not magic. I use it to compress feedback loops, never to skip engineering judgment.
Build in public
Decisions, dead ends and numbers published openly so other engineers can reuse them.
Maverick Lab
Experiments currently running.
Open work in progress. Nothing here is a product — these are questions I am answering in public.
Agent-driven delivery pipelines
Wiring agents into CI/CD to triage failures, propose fixes and keep the pipeline green without hiding risk.
CI/CDAgentsAzure DevOps
Event Sourcing without the pain
A minimal TypeScript kit that makes projections, replays and versioning boring instead of scary.
TypeScriptEvent SourcingKafka
Developer experience benchmarks
Measuring what actually slows teams down: build times, local setup, review latency, cognitive load.
DXPlatformMetrics
Coming soon
AI Dev Stack
A practical guide to the AI-native development workflow I use day to day — tools, prompts, guardrails and the parts that are still not worth it.
- Tooling map: what I use for code, review, tests and docs
- Guardrails that keep AI output production-safe
- Prompt patterns for architecture and refactoring work
- Honest limits: where AI still costs more than it saves
Get notified on WhatsApp
No product on sale. When it is ready, you get the link first.
Method
From code to field.
The loop every Maverick experiment goes through before it becomes advice.
01
Code
An idea becomes a small, isolated build with a clear question attached to it.
02
Field
It runs against real workloads, real teams and real constraints — not a demo dataset.
03
Report
Results, numbers and the decision I would repeat get published as a field report.
Community
A Tropa
A small group of engineers who get the experiments, the raw notes and the questions before anyone else.
- Early access to every Maverick Lab experiment
- Field reports with the numbers behind the decisions
- Direct line to ask questions about architecture calls
Writing
Field Reports
Long-form write-ups from the lab. The first ones are being written — meanwhile, my published articles live on DEV.to.
What AI actually changed in my architecture workflow
Draft
Event Sourcing decisions I would not repeat
Draft
The cost of a bad bounded context
Draft