Tech Radar
My Tech Radar
What I run in production, what I'm trialing, what I'm studying — and what I've set aside. Inspired by the Thoughtworks Tech Radar.
Updated August 2026
This edition: Next.js and Tailwind moved up to Trial; RAG, LLM as judge and agentic coding entered as Trial; Jenkins stays on Hold.
- Adopt
- In production with confidence — my default choice.
- Trial
- In use on real projects while I build conviction.
- Assess
- Studying and prototyping; not yet proven in my work.
- Hold
- Deliberately set aside — I wouldn't pick it for anything new.
New on the radarEstablished entry
Hover, click or tap a blip to see details.
- Adopt
- 21
- Trial
- 17
- Assess
- 4
- Hold
- 4
Techniques
Adopt
Retry patterns, DLQs and multi-tenant messaging in production on RabbitMQ and Pub/Sub.
My services — and this portfolio — isolate the domain behind ports; swapping a mock adapter for a real one costs one file.
- New on the radar
Lighthouse and bundle budgets enforced by a dedicated CI runner — this site holds Performance 100 because regressions fail the pipeline.
Tests drive my service design; integration tests hit real dependencies via Testcontainers, with over 80% coverage.
Trial
- New on the radar
AI agent loops with automated gates drive this portfolio's design, performance and content — Lighthouse 100 came out of one.
Entities and bounded contexts shape my modeling; deepening the strategic patterns before calling it Adopt.
Decoupling deploy from release; built NOP_FEATURE_FLAG to explore flag patterns in Java.
- New on the radar
Automated evaluation gates where a model scores the work before it ships — the design loop only closed after two consecutive 4+ scores from the judge.
- New on the radar
Built VectorSeek — semantic search feeding LLM answers; validating chunking and evaluation before promoting.
- New on the radar
Versioned specs drive the agents' work — this portfolio's design, performance and content loops run from prompt specs with explicit acceptance gates.
Languages & Frameworks
Adopt
Three years in production, v8 through 17 — reusable component libraries with Angular Material.
Core backend language — Java 8 to 21 in production across fintech and agribusiness systems. Still my default for services.
My pick for new JVM code; published an open-source integration lib (cbenef-integration) on JitPack.
Default backend stack — REST APIs and microservices on Spring Boot 3 with clean architecture in production.
Persistence by default with JPA/Hibernate; I know where the abstraction leaks and when to drop to SQL.
JWT, OAuth2 and mTLS in production — security configuration I own rather than copy.
Every line of frontend I write is typed; it is also spreading into my tooling and scripts.
Trial
- New on the radar
This portfolio runs on it — App Router, RSC and a Lighthouse-100 performance budget. Heading to Adopt as I ship more.
Light services, adapters and tooling — what I reach for when the JVM is overkill.
Automation scripts and the RAG backend of VectorSeek. A utility language, not an identity.
Hooks and context in production apps; the foundation of my Next.js work.
Eureka, Config Server and Gateway in microservice setups; watching which parts survive Kubernetes-native alternatives.
- New on the radar
Utility-first styling on this portfolio and new frontends — fast iteration with a design system on top.
Hold
Superseded by Spring and Next API routes in my new work; kept only for legacy services.
Shipped NF-e mobile apps with it, but mobile is no longer my focus — paused deliberately.
Tools
Adopt
Everything I ship runs in a container — Compose for local environments, lean images for CI.
Five years of clean branching, rebases and code-review discipline.
My default CI/CD — lint, test and deploy on every project, including this portfolio.
Multi-module JVM builds; the build tool behind my open-source lib.
Unit tests with ts-jest and coverage on my TypeScript projects, including this site.
JUnit, Mockito and Testcontainers with over 80% coverage on services I own.
Trial
- New on the radar
Building MCP servers to plug LLMs into real tooling — this repo ships browser-automation and reasoning servers.
- New on the radar
End-to-end and performance verification of this portfolio, with a Lighthouse budget enforced in CI.
Hold
Ran my pipelines for years; migrated to GitHub Actions and GitLab CI and haven't looked back.
Platforms
Adopt
Production workhorse — indexing and query tuning that cut key queries by 60%.
My default relational database for new projects — transactions, procedures and advanced queries in production.
Async processing with retry patterns and DLQs in production; my default broker for service integration.
Where this portfolio deploys — edge-ready with a preview per branch. Default for frontend projects.
Trial
App Engine, Cloud Run, Cloud SQL and Pub/Sub in real projects — my main cloud today.
Event-driven integrations with multitenancy running in production on GCP.
Distributed cache in production; expanding into patterns beyond caching.
Assess
- New on the radar
S3 and core services under study — mapping equivalents to what I already run on GCP.
- New on the radar
PoCs with producers, consumers and offset management; assessing where it beats RabbitMQ or Pub/Sub for my cases.
- New on the radar
Building cluster literacy before recommending it over managed runtimes like Cloud Run.
- New on the radar
Aggregations and indexing in PoCs; I still default to relational unless the model truly demands documents.
Hold
Corporate legacy I operate well — procedures and tuning — but I wouldn't choose it for anything new.