AI/LLM/agentic coding guru Scott Werner reveals some intriguing "#prompt patterns" (my term...? Did I just invent a term???) when using #Claude Code (CC) to do Test-Driven Development (#TDD)
Notes & caveats:
(a) We're planning a "sequel" where we look deeper into the results of this session, because I'm not entirely confident that CC got everything right. And, it will give us the opportunity to showcase *how* one can investigate the results of a bigger bite (bigger than what us test-driven nerds are used to, anyway). We also hope to ask CC to make more architecturally-significant changes, e.g., persistence.
(b) The video style is--how shall I put this?--"rough & ready with a touch of retro." My fault. I really do need to hire a video editor/producer. I recommend you watch at 1.25 speed, listen closely to everything Scott says, and ignore my awkward interruptions. 🤣
(c) Scott's interactions with CC seemed very natural, well-considered, and emergent. In one sense, it did feel like a mini-waterfall, because we asked CC to read the "PRD" and to build out a set of user stories all at once. OTOH, since progress was so rapid and feedback so easy to obtain, it also felt a lot like continuous flow, or what us old-timers thought of as #Agile (😱 "language...!")
#AI #LLM #Agentic
Okay, enough of that. On with the show...!
