Agent LabGame LabProjectsNotesKnowledgeToolsContact CZDE
I'm Dave, AI Enthusiast & Experimenter

I love AI,
so I iterate.

I iterate. I break prompts. I build agents. I ask questions. I commit and push.

↓ My projects Write me
daveai.cz ~ zsh
120+Experiments
340+Prompts
10M+Tokens
15+Live prototypes
19+AI agents
Interactive Demo

Agent Lab

See how AI agents work step by step

🔍
GEO Audit
Analyzes website for AI search engines
📝
LinkedIn Post
Generates a short LinkedIn post
🏔️
eMTB Advisor
Recommends trail based on conditions
🤖
Multi-Agent
4 agents collaborate on a blog post
Multi-Agent Orchestration
Waiting to start
TASK "Write a blog post about AI trends 2026"
🧠 PLANNER 🔍 RESEARCHER ✍️ WRITER 🔎 REVIEWER
🧠
Planner
waiting
↓
🔍
Researcher
waiting
↓
✍️
Writer
waiting
↓
🔎
Reviewer
waiting
Planner waiting
Researcher waiting
Writer waiting
Reviewer waiting
Agent workflow
Game Lab

Games from the AI workshop

Experimental games built with AI assistance — from idea to code. Each one is a creative experiment in AI code generation.

Projects

What I build

Active

Multi-Agent Blog Writer

4 AI agents collaborate on blog post creation — planner, researcher, writer, reviewer with feedback loop.

4 agents · 12 steps · feedback loop
Claude APILangChain
Active

Prompt Engineering Lab

Systematic testing of prompt techniques — few-shot, chain-of-thought, role prompting. Measuring output quality across models.

348 prompts · there's no bad output, only a bad prompt
Claude APIPythonJSON logs
Active

AI Personal Assistant

RAG-based assistant with access to company documents, emails and calendar. Answers queries in context.

4 months · ~25 queries/day · saves 2h of work every day
LangChainSupabasepgvector
Planning

AI Code Reviewer

GitHub Action that automatically reviews pull requests using Claude — checks code quality, security and consistency.

GitHub ActionsClaude API
Active

GEO Prototyping

Experiments with Generative Engine Optimization — how to optimize content for visibility in AI models.

10,000 queries · 18 categories · 5 AI platforms · visibility +100%
ClaudeChatGPTGemini
Active

eMTB Community

Platform for processing Bosch motor data — route sharing, ride analysis and performance stats for eMTB riders.

800 routes · 1,250 rides · 50 testers
Next.jsSupabaseBosch eBike API
Active

EV Cars Management

Dashboard for tracking potential savings in the network by managing charging power.

data from 1,485 vehicles · European markets
Next.jsACEA dataRecharts
Active

LinkedIn Agent

Automated LinkedIn post creation — from topic selection through generation to publication. Full pipeline.

90 posts · hundreds of hours saved
Claude APIn8nLinkedIn API
Active

Daily News Agent

Automatic daily briefing — scans 50+ sources, filters relevant AI news and compiles a personalized summary.

6 months · 50+ sources · 180 briefings
Claude APIRSSPerplexity
In development

Fee Management

Digital portal for managing municipal fees — waste, dogs, bio-waste. Citizens pay online, officials manage via admin panel.

Next.jsSupabaseResend
In development

Event Management

Platform for event registration and management with AI assistant for organizers — from attendee registration to automated communications.

planning, tracking and evaluating events · 15 events in testing
Next.jsStripeClaude API
In development

Digitalization for Municipalities

Comprehensive solution for the digital agenda of small municipalities — forms, payments, citizen communication.

Next.jsSupabaseClaude API
Active

Dealer Monitor

Automated dealer network monitoring — data collection, performance analysis and reporting across European markets.

5 months · 34 countries · 5,300 dealers
PythonSupabaseClaude API
Active

PID live

Real-time visualization of Prague integrated transport — public transit vehicle positions on a live map.

2,500–3,500 vehicles live · metro, trams, buses, trains, ferries
Next.jsGTFS RealtimeMapbox
Active

FinSignal

Financial market monitoring — earnings, analyses and signals from relevant sources for better investment decisions.

1 month · financial markets · earnings · analyses
PythonClaude APIFinancial APIs
Field Notes

What I learned along the way

Mar 14
GEO is not SEO — different game, different rules
Optimizing for AI models means thinking in citations and authority, not keywords.
Mar 10
Agents fail where a prompt succeeds
Not every task needs an agent. A simple prompt with good context often beats a complex workflow.
Feb 28
Vibe coding changes speed but not responsibility
With Claude Code a prototype takes an hour. But debugging AI code without understanding — that slows you down more.
Feb 10
Data quality beats model quality
Poorly structured data from Google Places ruins results more than model choice.
Feb 25
RAG is not magic, it's architecture
Chunking strategy changes results more than the model choice itself.
Currently studying
Prompt engineering
Agentic AI frameworks
GEO & AI visibility
Vibe coding & Claude Code
Output evaluation
Knowledge

Learn AI

67 terms

AI Glossary

From basics to advanced concepts — LLM, RAG, prompt engineering, agents, GEO and more. In Czech, clearly explained.

8 categories · live search · filters
LLMRAGAgentsGEO
 Guide

First Steps with AI

From zero to first results — for beginners and marketers. What AI can do, which tool for what, and how to write good prompts.

10 tools · prompt templates · practical tips
BeginnersMarketersPrompts
 Handbook

Prompt Engineering

Techniques that work — zero-shot, few-shot, Chain-of-Thought, role prompting and system prompts. With examples and templates from practice.

5 techniques · 4 templates · prompt anatomy
TechniquesTemplatesExamples
Toolstack

What I work with

// AI / LLM
Claude APIOpenAI APIGeminiHugging FaceOllama
// Agents & Automation
LangChainn8nMakeClaude Code CLI
// Backend & Data
SupabasePostgreSQLpgvectorPythonSQLDocker
// Frontend & Deploy
Next.jsVercelGitHub ActionsVS CodeTailwind CSS
Contact

Get in touch