Gemini + Art of Prompts – AI workflow session

Session dedicated to mastering AI prompts and practical use of Gemini in a daily workflow.

🤖 Gemini vs GPT-4 Testing (1.5h)

I compared both models on identical tasks:

Code Generation:

• Gemini: better contextual understanding in Angular/TypeScript
• GPT-4: more precise on complex algorithms
• Winner: Tie — depends on the use case

Content Writing:

• Gemini: more natural Polish language
• GPT-4: better structuring for long texts
• Winner: Gemini for Polish content

📝 Prompt Library Development (1h)

I created a repository of best prompts:

# Code Review Agent
You are a senior developer with 10+ years of experience.
Analyze the code below for:
- Performance bottlenecks
- Security vulnerabilities
- Code maintainability
- Best practices adherence

[CODE_BLOCK]

Categories created:

• Development prompts (20)
• Content creation (15)
• Analysis & research (10)
• Translation & localization (8)

🔄 Content Automation (0.5h)

Experimented with automatic generation of:

• Post outlines with keyword research
• Social media adaptations
• Technical documentation templates

Result: 60% reduction in research phase time!

Results

• Prompt Library v1 – 53 tested prompts
• Post #5 (Gemini + Prompts) – completed in both languages
• AI Workflow – optimized for daily content creation

Related links

• 📝 Gemini and How to Write Prompts (EN)
• 📝 Gemini oraz jak pisać promty (PL)
• 🔗 Prompt Library Repository

Key Learnings

1. Context is king — more context = better results
2. Agent-based prompts work better than generic requests
3. Iterative prompting > single mega-prompt
4. Language matters — Polish content performs better in Gemini

Next: Integrate Gemini API into the Jekyll workflow + automated content suggestions.