From Failure to Guardrail: Building Project Memory for AI Work
An API compatibility example shows how to choose a guardrail, verify it catches the failure, and review it…
Latest articles
How to Synthesize Multiple AI Reviews Into a Decision
A security, performance, and compatibility review of one API change shows how to separate findings from fixes and…
Pass Evidence and Uncertainty Between AI Agents
A test passes alone but fails in the full CI suite. Pass its results, the relevant CI log,…
Choosing the Right AI Workflow Pattern
Which parts of your AI task need to wait for earlier results? Use that question to choose a…
Define Verification Before Delegating to an AI Agent
A disabled checkout button can pass a test while duplicate orders remain possible. Start with the failure you…
Verification Is Not “Looks Good”: How to Evaluate AI-Assisted Work
A practical framework for deciding whether AI-assisted work is verifiable, how much evidence it needs, and when AI…
AI-Generated Code Is Not AI-Assisted Engineering
Generated code is an output. AI-assisted engineering is the owned process that frames the work, verifies the result,…
MapReduce: One Large Problem, Smaller Repeatable Tasks
Learn when to partition work by comparable units—and when a perspective-based split is the better fit.
Red Team / Blue Team: Challenge Before You Trust
When an undetected weakness would matter, challenge the result—then let evidence decide whether to keep, modify, or reject…
Fan-out / Fan-in: One Problem, Multiple Perspectives
See how multiple perspectives can examine the same problem independently before their findings are combined into one result.
Prompt Chaining with Subagents in Codex and Claude Code
A practical guide to explicit AI handoffs, the difference between delegation and chaining, and when the additional overhead…
Prompt Chaining: When One LLM Call Is Not Enough
Prompt chaining turns a task into separate LLM calls. See when those boundaries help—and when one structured call…
I Use Design Patterns for More Than Writing Code
Reusable workflow patterns help me structure AI-assisted work, coding, and non-coding tasks. The real value comes from combining…
The Transition Is Part of the Architecture
A safe system migration requires more than a clean target architecture. Learn how small boundaries, controlled routing, transformation…