Java · Spring · AI Engineering
Articles on Java, Spring Boot, AI Engineering, MCP, LLMs, RAG, software architecture, and study notes.
Pagefind — offline search after the production build.
I tried Matt Pocock's engineering skills harness on a Java CRUD refactor. First run skipped architecture; second run asked 17 questions—and this is what came out.
A Python FastAPI POC that combines MCP tools for CRM data with Azure AI Search RAG—orchestrated by Semantic Kernel for executive sales briefings.
Expose an existing Spring Boot REST API as MCP tools with Spring AI—without changing business logic.
I keep asking agents to remember more things. Promises in chat do not survive the next session. dont-forget makes the agent propose executable enforcement and wait for your OK before touching the repo.
I run Superpowers on Grok until the plan is closed, then implement with Composer. Cursor stays because of the harness — and because the plan has to last.
I wake up, fire questions at my AIs, go to the gym and pilates, come back with no idea where I left off. /recap turns the session into a dark HTML page — requested vs delivered, mark topics as seen.
After skipping /to-spec in faruk-base2, I ran the full Matt Pocock skills path on a greenfield idea. Six GitHub issues, a label system, and a dependency chain — no implementation yet.
I tried Matt Pocock's engineering skills harness on a Java CRUD refactor. First run skipped architecture; second run asked 17 questions—and this is what came out.
A Python FastAPI POC that combines MCP tools for CRM data with Azure AI Search RAG—orchestrated by Semantic Kernel for executive sales briefings.