AI Tools: most effective in bold
- Usage: Generate PlantUML-based C4 system context, container, and component diagrams for Insurance Hub Java-based project
- AI Tools: Claude, Gemini, ChatGPT
- Comments: Gemini produced the most accurate diagrams with minimal edits.
- Usage: Generate explanations and flows for the above-mentioned C4 diagrams
- AI Tools: Gemini
- Usage: Generate PlantUML-based business flows for Insurance Hub Java-based project
- AI Tools: Claude, Gemini, ChatGPT
- Comments: Gemini produced the most accurate diagrams with minimal edits.
- Usage: Outline and refine the structure of the "Insurance Hub System & Go Migration Analysis" document
- AI Tools: Claude, Gemini, ChatGPT, Perplexity
- Comments: Took many follow-ups to polish and finalize
- Usage: Research and validation of possible target state, i.e., component migration choices, reasoning for replacing components, various system design options comparison, etc.
- AI Tools: Claude, Gemini, ChatGPT, Perplexity
- Comments: Very time-consuming, used primarily as a research assistant
- Usage: Writing of blog posts—content expansion based on my rough drafts
- AI Tools: Claude, Gemini, Perplexity
- Comments:
- Claude—very wordy, Gemini—less wordy, too dry
- Used primarily as "writing amplifier" rather than a content creator
- Prompt: Should include the writing style directive, e.g., "Use a straightforward, but not too simplistic, and reader-focused style."
- Usage: Writing of blog posts—final polish
- AI Tools: Grammarly (AI-powered)
- Usage: Publishing of the first blog post on kiroule.com was blocked by failing builds on Netlify due to outdated scripts for uploading index data to Algolia SaaS.
- AI Tools: Junie
- Comments: Algolia client was successfully migrated from v4 to v5, updating the
data-upload.jsscript for new API syntax and correcting the shell script package installation.
- Usage: What medium-size models (that were trained on Go latest versions) to use with Ollama local server?
- AI Tools: Perplexity, Gemini, ChatGPT, Claude
- Comments: Pulled qwen2.5-coder:7b, codegemma:7b, gemma:7b, deepseek-coder:6.7b, llama3.1:8B, codellama:7b, starcoder2:7b
| Model | Size | Code Training Cutoff | Go Support | Best For |
|---|---|---|---|---|
| Deepseek Coder | 6.7B | Early 2024 | Excellent | Go 1.21+, 1.24.x |
| Qwen2.5 Coder | 7B | Early 2024 | Excellent | Up-to-date Go |
| StarCoder2 | 7B | Early 2024 | Very Good | Accurate completions |
| Phi-3 Mini | 4B | Early 2024 | Good | Lightweight |
| Llama 3.1 | 8B | Early 2024 | Good | General plus code |
| Gemma | 7B/9B | 2024 | Good | Fast inference |
- Usage: Come up with a test tasks to verify if an Ollama model has been trained on the latest Go version(≥1.18 through 1.22+)
- AI Tools: Claude, Gemini, ChatGPT, Perplexity
- Usage: Generate incremental, detailed GitHub issues(tickets) aligned with the finalized Phase 1 scope, using a consistent ticket template for clarity.
- AI Tools: Junie(ChatGPT)
- Usage: Automate Makefile targets for cluster provisioning, infrastructure deployment, observability stack management, and lifecycle operations.
- AI Tools: JetBrains(Gemini 2.5 Flash), Perplexity
- Usage: Create Kubernetes manifests, Kustomize overlays, and resource configurations for both local (Kind) and production-like QA (K3s) clusters.
- AI Tools: JetBrains(Gemini 2.5 Flash), Perplexity
- Usage: Draft structured documentation for cluster provisioning, observability setup (Prometheus, Grafana, Zipkin), and troubleshooting tricky issues.
- AI Tools: JetBrains(Gemini 2.5 Flash), Perplexity
- Usage: Facilitate connectivity validation and health checks for stateful services using port-forwarding and test pods to ensure production readiness.
- AI Tools: JetBrains(Gemini 2.5 Flash), Perplexity
- Usage: Use as a proofreader to do the final polishing of the "From Docker Compose to Kubernetes: Lifting the Insurance Hub into the Cloud" article.
- AI Tools: JetBrains(Gemini 2.5 Flash), Perplexity
- Comments: Given the large size of the article(~5000 words or ~44000 chars - counted with the
wcutility), Perplexity was not able complete the task in one shot; I had to be explicit in the prompt to provide the revised version in parts. Also, for my established writing style, the prompt included a link to the previous article in the series, but Perplexity continued revising the already published article. - Prompt: Should include "I want you to preserve the original length and depth, including the headers, and improve flow, clarity, and consistency while applying my established writing style".
- Usage: Finalize technical "recipes" for K8s manifests (Kustomize bases/overlays) by replicating validated patterns across the remaining services.
- AI Tools: JetBrains(Gemini 3 Flash)
- Comments: Highly efficient for generating boilerplate manifests once the initial service convention was established.
- Usage: Refactor and implement cross-cutting service improvements:
LoggingFilterfor the gateway, AOP-basedEventPublisherLoggingInterceptorfor Kafka visibility, andJsReportTemplateProvisionerfor automated PDF template setup. - AI Tools: JetBrains(OpenAI Codex)
- Comments: Used the autonomous agent mode for these refactorings.
- Usage: Troubleshooting and resolving "stubborn" cluster validation issues (NGINX path matching, Micronaut 2 discovery client RBAC restrictions).
- AI Tools: JetBrains(Gemini 3 Flash), Perplexity
- Comments: Perplexity was effective for deep-diving into specific documentation for Kubernetes and NGINX resolver behaviors.
- Usage: Use as a proofreader to do the final polishing of the "Lift Completed, Now Shift: Making the Insurance Hub Kubernetes-Native (Enough)" article.
- AI Tools: JetBrains(Gemini 3 Flash), Perplexity
- Usage:
- AI Tools: JetBrains(Gemini 3 Flash), Perplexity
- Comments: