Tag: software-engineering
-
Writing acceptance tests first, agents second
The most reliable way to use a coding agent isn't a better prompt. It's writing the acceptance test first, so the agent has a target it can't argue with and you have a stopping condition.
-
Diffing prompts the way you diff code
Prompts are the least version-controlled artefact in most engineering teams. Moving them into files, with a fixture set and a pinned model, turns prompt work from vibes into something you can actually review.
-
Cline and Roo Code: open-source agents teach you more than closed ones
Cline and Roo Code let you read the exact prompt, the tool definitions and the token bill behind every agent turn. Closed tools hide all three, which is why they're worse to learn from.
-
JetBrains AI Assistant: not every green inspection is an agent task
IntelliJ's inspections come with quick-fixes that are correct by construction. Knowing which warnings deserve an agent, and which deserve Alt+Enter, is most of the value of AI Assistant.
-
Windsurf: the value of a pause button in an otherwise relentless agent
Windsurf's Cascade agent is built to run long. Interrupting it costs less than letting a bad run finish. Field notes on the three signals worth stopping on, and what state an interrupt leaves behind.
-
Cheap wins: the keybindings that turn Cursor back into a proper editor
Cursor's default shortcuts collide with a decade of VS Code muscle memory. Here are the remaps that stop the AI from hijacking keys you already own, and the two bindings worth adding.
-
Google Gemini: Using long context to write migration scripts that respect every edge
A field-tested take on using long context to write migration scripts that respect every edge with Google Gemini: what it rewards, where it breaks, and how to keep the workflow honest.
-
ChatGPT Codex: Splitting epics into Codex-sized tasks on purpose
A field-tested take on splitting epics into Codex-sized tasks on purpose with ChatGPT Codex: what it rewards, where it breaks, and how to keep the workflow honest.
-
Claude Code: Quiet art of pausing the agent instead of letting it churn
A field-tested take on the quiet art of pausing the agent instead of letting it churn with Claude Code: what it rewards, where it breaks, and how to keep the workflow honest.
-
GitHub Copilot: Underrated power of /explain on legacy code no one dares to touch
A field-tested take on the underrated power of /explain on legacy code no one dares to touch with GitHub Copilot: what it rewards, where it breaks, and how to keep the workflow honest.
-
Local LLMs with Ollama: When a local 7B beats a cloud 70B in latency-sensitive loops
A field-tested take on when a local 7B beats a cloud 70B in latency-sensitive loops with Local LLMs with Ollama: what it rewards, where it breaks, and how to keep the workflow honest.
-
Agentic coding: How to review a PR an agent opened in the middle of the night
A field-tested take on how to review a PR an agent opened in the middle of the night with Agentic coding: what it rewards, where it breaks, and how to keep the workflow honest.