Insights
Engineering concepts explained the way we wish someone had explained them to us.
Practical thinking on AI-assisted development, data systems and architecture ; written to help engineers understand the technology and help leaders make better decisions about how their teams use it.
Day 8 : Eval-Driven AI Engineering: Stop Judging Copilot by ‘Looks Good’
- AI-assisted Dev, GitHub Copilot
How golden task sets, objective scoring and hidden evals turn Copilot configuration changes into measurable engineering experiments instead of prompt opinions.
read moreInvariant Discovery and Property-Based Testing: The 5-Minute Mental Model
- Testing, Property-Based Testing
A short practical explanation of invariants and property-based testing, with one software example and one data-engineering example.
read moreDay 7 : Make GPT Discover What Must Always Be True
- AI-assisted Dev, Testing
How AI-assisted invariant discovery and property-based testing can reveal behavioral guarantees that example-based unit tests often miss.
read moreDay 6 : Make a Second GPT Try to Break the First GPT’s Code
- AI-assisted Dev, GitHub Copilot
How adversarial AI verification uses fresh context, evidence-driven findings and reproducible counterexamples to challenge AI-generated code.
read moreDay 5 : Plan–Execute–Verify: Stop Letting Copilot Agent Improvise
- AI-assisted Dev, GitHub Copilot
How bounded agency turns AI coding from open-ended improvisation into a controlled engineering workflow with explicit plans, scope gates and verification.
read moreDay 4 : Control What Copilot Reads Before It Reasons
- AI-assisted Dev, GitHub Copilot
Why context budgeting, which constructs the smallest sufficient evidence set, improves repository-aware AI decisions more than simply giving a model more files.
read moreDay 3 : Turn Good Prompts Into Version-Controlled Engineering Workflows
- AI-assisted Dev, GitHub Copilot
How prompt files turn repeatable AI engineering tasks into version-controlled workflows that teams can review, reuse and improve.
read moreI Thought GitHub Copilot and GPT Were the Same Thing
- AI-assisted Dev, GitHub Copilot
A practical mental model for software engineers and engineering leaders adopting LLMs, coding assistants and agentic development inside real engineering teams.
read moreDay 2 : Turn Your Repository Into Part of the Prompt
- AI-assisted Dev, GitHub Copilot
How GitHub Copilot repository instructions turn stable engineering knowledge into persistent, version-controlled context for developers and coding agents.
read moreDay 1 : Context Engineering: Stop Prompting, Start Controlling Context
- AI-assisted Dev, GitHub Copilot
Why better AI-assisted engineering starts by controlling what GitHub Copilot can see, separating repository discovery from implementation, and verifying its architectural model before code changes begin.
read more