About
AI agents say "done". This blog is about making them prove it.
I build agentic-fx: free Claude Code bundles that take software from idea to release with gates, evidence and a reviewer the agent can't overrule. These are my notes from building them.
I'm Harry Nguyen.
Five checks
Every piece of work here, with AI or without, is held to five checks. A pillar without a check that can fail is a slogan, so each one has a measure. Every post carries one or two of them.
| Pillar | The question |
|---|---|
| Intelligence | Does it know the craft: design, build, test and manage, in theory and in practice? |
| Memory & learning | Does it read what it learned last time, so each run beats the last? |
| Judgment & integrity | Does it stop and ask instead of inventing what only a person knows? |
| Verification & safety | Is every step backed by evidence, and does it ask before anything it can't undo? |
| Value & experience | Will people use it, and does it cost what it should? |
What you'll find
- AI Agents: SDLC & STLC: sdlc and stlc, from a raw idea to a deploy that waits for QA.
- AI Evaluation: measuring LLM features without fooling yourself.
- Building with AI: prompts, agents, RAG and the API, and the apps built with them.
- QA & Testing: what to learn, in what order, and why.
- Career & Teams: the QA career path, first jobs and leading a team.
Try it
sh
npx agentic-fx@latest installThe tools are on the Projects page.