"How do you develop real fluency in prompt engineering without a roadmap? At Mercor AI, I found myself inside an unusual opportunity:'Vibe Coding' 100+ high-fidelity experiments in 4 weeks. I knew the outcome was uncertain. I decided to focus on what I could control, the craft."

The Environment: Zero-Scrape Constraints

Working within a proprietary, air-gapped ecosystem meant no external AI assistance and no web-scraping. Every output was constructed from first principles, through manual prompt architecture. As constraints go, it was a good teacher.

The Approach

Architectural RLHF (Reinforcement Learning from Human Feedback)

"My process involved developing a Dynamic Prompting Framework that evolved through a rigorous voting and feedback loop. Whatever this work fed into, I was determined to come out of it with sharper skills than I went in with."

The A/B Benchmarking

For every prompt, I audited dual outcomes, voting on "Brief Proximity" vs. "Execution Quality." Repetition at this volume builds instinct.

The Feedback Loop

Writing detailed, granular critiques sharpened my ability to articulate what good looks like in AI-assisted design, a vocabulary I've carried forward.

Systemic Steering

Integrating SVG asset calling, Google Font mapping, and proportional grid-guides into prompt logic taught me how to embed design thinking directly into instructions.

Orchestration Layers

As the experiment progressed, I moved from single-screen prompts to multi-screen frameworks, managing interaction logic and animation states within the initial prompt load.

What I Took Away

Prompt engineering fluency, built under pressureHigh volume, tight constraints, and no shortcuts. The skills I developed here are ones I now apply independently.

A clearer eye for AI and design governance

Working this close to a model's decision-making gave me a more grounded understanding of how AI interprets design intent, and where it falls short.

My own tooling

The research led directly to developing specialised prompting tools for deep-application and interactive system framing, work that continues beyond this project.

Key Focus: Prompt Engineering, Generative Operations (GenOps), Design Governance, Systemic Steering, RLHF Methodology.