Orrin Francis, Principal Consultant
So Pat built an app with Claude, and now he thinks he’s
a developer. Cute.
While the business side of ITAaaS is busy celebrating
their newfound Python powers, the engineers are over
here quietly automating CI/CD pipelines, enforcing
security policies, and managing codebases with tools
that speak fluent architecture.
In this post, I offer a slightly more engineered take on
AI-assisted development—one built for scale, rigour,
and long-term maintainability.
Spoiler: it's not just about code, it's about systems.
An Engineer’s Take: Augmented Development Workflows
Pat recently shared his experience in our blog with using Claude to build a production-grade app—no coding background required. It’s a great read (and worth checking out if you haven’t already). But as the resident engineer, I have a slightly different take.
While tools like Claude are empowering business leaders to get hands-on with tech in new ways, engineers are experiencing their own quiet revolution. The tooling we now have access to—especially IDE-native platforms like Cursor—offers a radically more efficient and robust path to building scalable, secure applications.
So here’s how someone like me might tackle the same kind of project, just with an engineering lens.
Step 1: Problem Definition & Architecture
Business Owner Approach
Pat outlined his goal in plain English, and Claude interpreted and generated code accordingly.
Engineering Workflow with Cursor
I’d use Cursor’s /spec command to turn requirements into a technical blueprint—complete with architecture diagrams via Mermaid.js, and a scaffolded FastAPI project with PostgreSQL models and a job scheduler built in.
# /spec: "Build a market data tracker with API polling, PostgreSQL storage, and error handling"
Cursor returns:
- FastAPI scaffold
- SQLAlchemy models
- APScheduler config
Step 2: Code Implementation
Business Approach
Claude helped Pat build the project step-by-step through chat, iterating as he went.
Engineering Workflow
Cursor’s code generation is aware of the entire codebase, not just one file. If I want to add retry logic to an API client, Cursor applies the right decorators based on the project’s existing patterns. It also supports real-time collaboration with other devs or AI reviewers, right in the IDE.
# Type: "Add retry logic to API client"
→ Injects @retry decorators with Tenacity, consistent with codebase
Step 3: Debugging & Optimisation
Business Approach
When errors popped up, Claude helped interpret them and offered fixes.
Engineering Workflow
With Cursor, I get error-specific suggestions and performance profiling built-in.
Example:
ERROR: API rate limit exceeded (HTTP 429)
→ Suggestion: "Implement exponential backoff? [Apply Fix]"
Or, if performance is an issue:
# Highlight code → "Benchmark this"
→ Identifies bottleneck in pandas.DataFrame
→ Recommends Polars rewrite with parallelism
Step 4: Security & Maintainability
Business Approach
Security and resilience weren’t a core part of the build, yet. And that’s totally fine for a quick prototype.
Engineering Workflow
We build with security-first principles baked in. Examples:
Secret detection flags exposed API keys
`cursor audit --cve` checks for vulnerable dependencies
Auto-generated documentation and typed interfaces
