March 2025
31 published posts filed under March 2025. Each article shows its actual publication date.
-
Agentic coding: Case for suggest-only agents in safety-critical code
A field-tested take on the case for suggest-only agents in safety-critical code with Agentic coding: what it rewards, where it breaks, and how to keep the workflow honest.
-
MCP servers: First sandboxing policy we wished we had started with
A field-tested take on the first sandboxing policy we wished we had started with with MCP servers: what it rewards, where it breaks, and how to keep the workflow honest.
-
AI coding workflows: Shutdown routine that made AI-heavy days survivable
A field-tested take on the shutdown routine that made AI-heavy days survivable with AI coding workflows: what it rewards, where it breaks, and how to keep the workflow honest.
-
Prompt engineering for code: Diffing prompts without a system, and the system that followed
A field-tested take on diffing prompts without a system, and the system that followed with Prompt engineering for code: what it rewards, where it breaks, and how to keep the workflow honest.
-
Cline and Roo Code: Picking models per task with almost no data, and learning fast
A field-tested take on picking models per task with almost no data, and learning fast with Cline and Roo Code: what it rewards, where it breaks, and how to keep the workflow honest.
-
JetBrains AI Assistant: AI Assistant versus plain inspections on the same codebase
A field-tested take on AI Assistant versus plain inspections on the same codebase with JetBrains AI Assistant: what it rewards, where it breaks, and how to keep the workflow honest.
-
Windsurf: First atomic commit authored entirely by Cascade
A field-tested take on the first atomic commit authored entirely by Cascade with Windsurf: what it rewards, where it breaks, and how to keep the workflow honest.
-
Cursor: First week we stopped fighting Cursor and let it drive
A field-tested take on the first week we stopped fighting Cursor and let it drive with Cursor: what it rewards, where it breaks, and how to keep the workflow honest.
-
Google Gemini: Mixing AI Studio and Code Assist without losing context
A field-tested take on mixing AI Studio and Code Assist without losing context with Google Gemini: what it rewards, where it breaks, and how to keep the workflow honest.
-
ChatGPT Codex: When Codex rewrote code we did not ask it to
A field-tested take on when Codex rewrote code we did not ask it to with ChatGPT Codex: what it rewards, where it breaks, and how to keep the workflow honest.
-
Claude Code: Running Claude Code on a legacy repo nobody wanted to touch
A field-tested take on running Claude Code on a legacy repo nobody wanted to touch with Claude Code: what it rewards, where it breaks, and how to keep the workflow honest.
-
GitHub Copilot: First .github/copilot-instructions.md that actually shipped
A field-tested take on the first .github/copilot-instructions.md that actually shipped with GitHub Copilot: what it rewards, where it breaks, and how to keep the workflow honest.
-
Local LLMs with Ollama: Local evaluations that predicted our first cloud regression
A field-tested take on local evaluations that predicted our first cloud regression with Local LLMs with Ollama: what it rewards, where it breaks, and how to keep the workflow honest.
-
Agentic coding: Short loops as the winning rhythm, discovered through long ones
A field-tested take on short loops as the winning rhythm, discovered through long ones with Agentic coding: what it rewards, where it breaks, and how to keep the workflow honest.
-
MCP servers: Why we promoted an internal wiki to an MCP resource
A field-tested take on why we promoted an internal wiki to an MCP resource with MCP servers: what it rewards, where it breaks, and how to keep the workflow honest.
-
AI coding workflows: Time-boxing agent loops before they ate the afternoon
A field-tested take on time-boxing agent loops before they ate the afternoon with AI coding workflows: what it rewards, where it breaks, and how to keep the workflow honest.
-
Prompt engineering for code: Prompt library we started in a shared Markdown file
A field-tested take on the prompt library we started in a shared Markdown file with Prompt engineering for code: what it rewards, where it breaks, and how to keep the workflow honest.
-
Cline and Roo Code: First audit trail that made a security lead smile
A field-tested take on the first audit trail that made a security lead smile with Cline and Roo Code: what it rewards, where it breaks, and how to keep the workflow honest.
-
JetBrains AI Assistant: Beginner workflow with Junie, ADRs, and module scope
A field-tested take on a beginner workflow with Junie, ADRs, and module scope with JetBrains AI Assistant: what it rewards, where it breaks, and how to keep the workflow honest.
-
Windsurf: Writing our first Cascade-friendly task spec
A field-tested take on writing our first Cascade-friendly task spec with Windsurf: what it rewards, where it breaks, and how to keep the workflow honest.
-
Cursor: Indexing hiccups on a monorepo larger than the demo
A field-tested take on indexing hiccups on a monorepo larger than the demo with Cursor: what it rewards, where it breaks, and how to keep the workflow honest.
-
Google Gemini: IAM audits with Gemini: the day it noticed what we missed
A field-tested take on IAM audits with Gemini: the day it noticed what we missed with Google Gemini: what it rewards, where it breaks, and how to keep the workflow honest.
-
ChatGPT Codex: Scoping tasks to finish in a single pass, a beginner lesson
A field-tested take on scoping tasks to finish in a single pass, a beginner lesson with ChatGPT Codex: what it rewards, where it breaks, and how to keep the workflow honest.
-
Claude Code: Hooks as a cheap gateway drug to workflow automation
A field-tested take on hooks as a cheap gateway drug to workflow automation with Claude Code: what it rewards, where it breaks, and how to keep the workflow honest.
-
GitHub Copilot: Why @workspace changed how we wrote issues
A field-tested take on why @workspace changed how we wrote issues with GitHub Copilot: what it rewards, where it breaks, and how to keep the workflow honest.
-
Local LLMs with Ollama: Quantisation choices with incomplete information
A field-tested take on quantisation choices with incomplete information with Local LLMs with Ollama: what it rewards, where it breaks, and how to keep the workflow honest.
-
Agentic coding: First rollback plan we actually used
A field-tested take on the first rollback plan we actually used with Agentic coding: what it rewards, where it breaks, and how to keep the workflow honest.
-
MCP servers: Composing agents over a tiny catalogue of shared tools
A field-tested take on composing agents over a tiny catalogue of shared tools with MCP servers: what it rewards, where it breaks, and how to keep the workflow honest.
-
AI coding workflows: Rescue patterns discovered after the first agentic disaster
A field-tested take on rescue patterns discovered after the first agentic disaster with AI coding workflows: what it rewards, where it breaks, and how to keep the workflow honest.
-
Prompt engineering for code: Few-shot examples drawn from our own first PRs
A field-tested take on few-shot examples drawn from our own first PRs with Prompt engineering for code: what it rewards, where it breaks, and how to keep the workflow honest.
-
Cline and Roo Code: Cline tool manifests we shared with colleagues over lunch
A field-tested take on Cline tool manifests we shared with colleagues over lunch with Cline and Roo Code: what it rewards, where it breaks, and how to keep the workflow honest.