Most read articles
The write-ups people come back to — architecture, AI tooling and the practical side of shipping software.
Tech insights & development tips
Practical write-ups on Java, Spring, React, AI tooling and the day-to-day work of shipping software.
When Self-Hosting a Small LLM Makes Sense for Internal Automations
Self-hosting a compact language model can be justified for narrow, repeatable internal workflows, but only when workload fit, operational ownership, and measured…
How Java Teams Should Choose Restate or Temporal for Long-Running AI
A practical decision guide for Java teams evaluating Restate and Temporal for reliable, long-running AI-assisted and business workflows.
A Practical Sandbox Architecture for Internal AI Agents That Can Act
Internal agents become risky when model reasoning, tool permissions, credentials, and execution all share one trusted process. This reference design separates those…
How to Measure AI Coding Agent Productivity Without Misleading Metrics
A practical scorecard for evaluating AI coding agents without relying on token counts, lines of code, or misleading pooled metrics.
Optimistic React Forms That Recover Cleanly When the Server Says No
A practical React 19 pattern for fast forms that preserve user input, distinguish provisional UI from confirmed server data, and recover clearly…
When Your RAG Copilot Needs Exact Numbers, Route Calculations to Code
RAG can retrieve useful context, but it should not be the authority for arithmetic, date logic, or business-rule outcomes. This guide shows…
Skills vs MCP for Internal AI Tools: A Decision Framework for Teams
MCP provides a standard interface to external capabilities, while skills define the repeatable procedure for using them. This framework helps Java and…
How to Stop AI-Hallucinated Dependencies Reaching Java and React CI
Treat every package, Maven coordinate, npm module, and install command proposed by an AI coding assistant as untrusted input until registry, provenance,…
How to Pilot Cross-Cloud AI Agents Without Unmanageable Sprawl
A cross-cloud AI agent pilot should prove one workflow, not create a new platform. Use narrow contracts, local identity, failure testing, and…
Screen Recordings to Reliable Internal Automations: A Practical Guide
Screen recordings can reveal repetitive work, but raw video, audio, OCR, and UI metadata should remain restricted evidence. This blueprint shows how…