Agentic Fullstack Development Webinar
Source
- Local PDF:
/Users/kaungkhantko/Downloads/agentic_fullstack_development_webinar.pptx (1).pdf - Author/presenter: Kyaw Ko Ko Tun, Let's Tech Club
- Format reviewed: 29-slide presentation
- Main derivative workflow: Spec Map Build Prove Learn
Core thesis
Agentic development is not “AI writes code.” The developer designs the environment, boundaries, feedback loops, and evidence that make reliable outcomes repeatable. The role progresses from executor to driver, pair, director, and finally system designer.
Important distinctions
| Fragile vibe loop | Agentic engineering loop |
|---|---|
| Prompt → code → bug | Spec → map → build → prove → learn |
| Vague app request | Product-level direction with non-goals |
| Agent can touch everything | Explicit permission envelope |
| Build passing is enough | Tests plus browser/perceptual evidence |
| “It should work” | Commands, results, diff, and risks |
| Corrections stay in chat | Rules persist in tests, contracts, guidance, or skills |
Full-stack verification model
- Backend: contract-driven. Prove request validation, business invariants, database behavior, response shape, and invalid cases.
- Frontend: perception-driven. Prove the interface in a browser, including loading, empty, error, responsive, keyboard, validation, and console states.
Repository design for humans and agents
- Predictable folders and clear naming
- Small modules
- Typed contracts
- Fast test commands
- Seed/demo data
- Root and scoped project-guidance files
- Stable rules near the code they govern
Best reusable lesson
When repeated correction is necessary, improve the environment rather than repeatedly polishing the prompt. Tests are agent feedback; contracts reduce ambiguity; boundaries reduce unintended changes; persistent guidance turns a temporary correction into repository memory.
Applied in this vault
- Agent Rules stores stable global expectations.
- Definition of Done makes completion evidence explicit.
- Prompt Library provides task-mode prompts.
- Architecture Decision Record preserves consequential decisions.
- Testing Strategy and Code Review strengthen the prove step.