Blog
Posts about software architecture, development, infrastructure, and projects.
Can an Agent Understand a Real Spring Boot Codebase?
I connect my local coding agent to a real Spring Boot project to see whether it can understand the repository structure, recognize existing patterns, and find the right context without loading the entire codebase.
How to Set Rules and Permissions for a Coding Agent
Before giving my coding agent access to a real project, I define what it can do automatically, what requires approval, what it must never do, and where human review belongs.
Context Is the Real Interface to a Coding Agent
Before connecting my local coding agent to a real project, I define how it should discover the repository, select useful context, and learn the rules that cannot be inferred from code.
What Happens When You Give a Small Local Model Tools?
I give small local coding models access to files, Git, and terminal commands through Aider to see whether tools make them more useful or simply make their mistakes more dangerous.
Junie vs Aider: Which Agent Works Better with Local Models?
I connect Junie and Aider to the same local Ollama models to see how much the agent harness changes the coding experience.
Choosing the Agent Layer: IDE Plugin, Junie, Aider, or Something Else?
Before testing a coding agent, I need to decide where the agent should live, what tools it needs, and whether IDE integration should really be a requirement.
Qwen vs Ministral vs GPT-OSS on an M2 Mac
I tested four local coding models on an M2 Mac with 16 GB of memory, comparing coding quality, reasoning, speed, memory pressure, and practical development usability.
How to Benchmark Local Coding Models Without Fooling Yourself
A practical approach to benchmarking local coding models using repeatable tests for coding quality, reasoning, instruction following, speed, and resource usage.
Can a Mac Run a Useful Local Coding Model?
How I prepared my M2 Mac for local coding models, installed Ollama, selected a realistic model, and verified the first local LLM.
What Makes a Coding Agent Agentic?
What changes when an AI can inspect a codebase, modify files, run commands, observe failures, and continue working toward a solution?
What Does Productivity Mean in Agentic Coding?
What is productivity, and how do we measure whether agentic coding is actually improving our productivity?
LLMs, Agents, and Tools: What Is Actually Doing the Work?
Understanding the main pieces behind a local agentic coding environment: LLMs, parameters, quantization, runtimes, tools, and agents.
How to Pick a Project for Learning Agentic Coding
How I chose two existing portfolio projects as safe, useful sandboxes for testing agentic coding.
How to Learn Agentic Coding at Zero Cost
Can we learn the fundamentals of agentic coding locally before paying for tokens, cloud GPUs, or commercial coding agents?
Why Is Experimenting with Agentic Coding Challenging?
Why cost, motivation, learning challenges, and changing developer habits can make experimenting with agentic coding harder than it first appears.