🧪 AI Development, Part 5: RIPER in Action — Real Examples from Production

“When you do it right, Cursor does the right work.”

So far in this series, we’ve covered the theory:

• Why AI development is now viable
• How Cursor works as an IDE
• What RIPER brings to the table
• How memory banks and rule files give structure

But none of that means much without real-world results.

This post is all about showing what happens when theory meets execution — when Cursor and RIPER help ship actual features. No demos. No Hello World. These are real stories from GrowthSpace production systems.

📡 Example 1: Rolling Out Telemetry Across Microservices

Challenge:
We had partial telemetry coverage across our NestJS and ExpressJS microservices — some had the GSRequestContext module wired in, others didn’t. Manually auditing and adding it was going to be tedious.

Approach with RIPER:

1. /research → Asked Cursor to identify which microservices already had telemetry and which didn’t.
2. It scanned the repo and generated a breakdown, identifying modules with missing telemetry.
3. /plan → Asked it to prepare a checklist of changes, one microservice per commit.
4. Adjusted the plan manually (skipped deprecated or unused services).
5. /execute → Cursor added imports, initialized middleware, and validated builds.
6. /review → We checked that changes matched the checklist and passed all tests.

🧠 Cursor even flagged microservices that were technically using the telemetry module, but weren’t initializing it correctly.

Result:
A clean, commit-per-service migration with build/test validation — 90% of the work was done by Cursor.

🗂️ Example 2: Migrating Angular Material to MDC Components

Challenge:
Angular 16 dropped support for legacy Material imports. We had to migrate hundreds of components across shared libraries and the main app.

Approach with RIPER:

1. Supplied Cursor with:
   • Legacy and new import formats
   • Angular’s migration doc
   • A single example of a legacy component
2. /research → Identified all non-legacy Material imports.
3. /plan → Created a multi-phase checklist:
   • Migrate buttons, form fields, selects, etc.
   • Update shared library first
   • Run manual QA after each phase
4. /execute → Migrated each group of components
5. Also generated a QA test checklist:
   • Included specific components
   • Mapped to UI pages where changes occurred

🧠 Cursor tried skipping QA focus at first — we corrected it, and it regenerated the plan to include test cases and target screens.

Result:
A hybrid migration strategy + QA plan, with Cursor handling the heavy lifting and tracking which pages needed testing.

💬 Example 3: Building a Shared Chat History Viewer

Challenge:
We wanted to share successful Cursor chat histories with the team. The idea: a GitHub Pages-based viewer that could display .md logs in a readable UI.

Approach with Cursor:

1. Started a new project folder.
2. /plan → Described a GitHub Pages viewer using a basic UI.
3. Cursor generated:
   • A React frontend to display Markdown chats
   • Routing with shareable query params
   • UI that highlights specific messages
4. Added Markdown enhancements like rendering tables correctly
5. /execute → Built and tested with Cursor
6. /review → Adjusted UI responsiveness and added permalinks to messages

Result:
A fully working shared chat viewer written almost entirely by Cursor — now used by the entire R&D org to share good examples.

🔍 One More: Improving Error Reporting in Our Tests

Mini-win:
When synthetic data failed due to a wrong HTTP status, the logs were unclear.

Cursor Fix:

• Located error-prone sections
• Proposed improving printed error messages
• Added readable diffs between expected/actual status codes

Now our test failures actually explain why they failed.

🧠 What These Examples Show

• Cursor isn't just a code monkey — with RIPER, it becomes a feature-level contributor.
• We still need to correct and review, especially when hallucinations creep in.
• But the productivity gain is undeniable — more done, less tedium.
• And it works best when you guide it like a teammate — not just prompt it like a chatbot.

Coming Next: How We Write AI-Generated Tests (That Actually Work)

In the next post, Romario’s going to walk you through how we:

• Generate E2E tests with step-by-step precision
• Use Playwright’s MCP to explore UIs
• Write API tests that follow internal standards
• Iterate safely with Cursor across dev/test cycles

Until then, try this:
Open Cursor and ask:
/research which parts of this codebase don't use telemetry yet?
It might just find something you missed.

🧪 AI Development, Part 6: How We Write E2E and API Tests with Cursor

“It’s like talking to a child — the more specific you are, the better the result.” — Romario Nijim Testing is where most AI-powered dev tools fall short. Sure, they can generate some expect() statements — but useful, maintainable, standards-compliant tests? That’s another level. At GrowthSpace, we’ve built a workflow

Behind the BitRomario Nijim