Same AI. Same requirements. One extra sentence. Completely different results.
| Naive app | This app (allo_dept_gl) |
|
|---|---|---|
| Step 1 | Ask AI for the app | Run genai-logic create |
| Step 2 | Done | Load ALS context into AI |
| Step 3 | — | Paste requirements — done |
The naive developer gave the AI a blank page.
The ALS developer gave the AI a scaffold, a rulebook, and a running start.
Two architects. Same client brief. One works alone from scratch. The other works inside a firm with standard plans, building codes, and a crew standing by. Both deliver — but only one is move-in ready.
| Question | Naive | ALS |
|---|---|---|
| Production-ready? | No — months more work | Yes — included |
| Admin UI? | No | Yes |
| Add a new rule? | Risky rewrite | One line |
| Auditable? | Partially | Yes |
It's not a black box. Everything ALS generates is standard Python — plain files that engineers read, edit, and version-control with their normal tools. The scaffold is a starting point, not a cage.
The bottom line: Neither developer wrote more than a paragraph of requirements.
The tool — not the effort — is what made one system production-ready.