The Software Development Monthly is a monthly blog post that shares interesting or useful content which I consumed during the previous month. This blog post is always published on the eighth day of the month.
Let's begin!
AI
Continuous Development with Claude Code on GitHub demonstrates how we can use the Claude GitHub App and Claude Code for implementing features and reviewing PRs.
AI Coding will Prevent Expertise warns that heavy reliance on AI creates an illusion of competence and argues that true expertise is built through failing and solving hard problems yourself.
The current AI pricing was always going to go away explains why subsidized flat-rate AI pricing will eventually be replaced by usage-based billing.
Better tools made Copilot code review worse. Here’s how we actually improved it explains why replacing the code exploration tools of the GitHub Copilot code review agent led to worse reviews and describes how GitHub fixed this problem.
Viability of local models for coding identifies the key factors that determine if running local LLMs for programming is worth it.
Experiences with local models for coding describes what happened when the author used local LLMs for completing real programming tasks.
Reviewing AI Code Is Not A Viable Argument challenges the belief that humans can effectively review AI-generated code and ensure that only "good code" is merged to main.
The AI hype is a mass psychosis echo chamber of incompetent individuals rants that one of the main reasons behind the AI hype is that AI gives incompetent individuals a false sense of competence.
The AI Productivity Illusion explains why generating more things in less time doesn't necessarily increase economic productivity.
The Economic Benefit of Refactoring documents an experiment where the author demonstrated that refactoring AI-generated (and messy) code reduces the future token consumption of AI agents.
Software Development
Waterfall Was Never the Problem argues that waterfall is actually a good choice in certain situations and points out that critics often attack Royce's initial version instead of the final product. Finally, the author explains how AI fixes waterfall's biggest problem.
Nulling Out References Won’t Help Your Garbage Collector busts the myth that setting local variables to null helps the JVM garbage collector do its job more efficiently.
When NOT TO USE Event-Driven Architecture (EDA) identifies seven situations where we shouldn't use event-driven architecture because it adds unnecessary complexity and causes issues in production.
OWASP Top 10 2025 identifies the new OWASP Top 10 for 2025, compares it to the 2021 list, and analyzes what has changed in web security over the past four years.
The cost of saying yes has changed argues that even though AI has dramatically reduced the cost of writing code, the cost of owning, reviewing, and maintaining code hasn't changed.
The Archaeologist’s Copilot explores how software developers can use AI for investigating and modernizing legacy codebases.
The Conductor Developer describes how we are shifting from solo coding to "orchestrating" AI agents and argues that because AI agents write more code, human attention has become the new bottleneck.