Systems that do the repetitive part
The experiments here are practical: content pipelines, agents and workflows built to remove repetitive work. Each is documented by objective, architecture, tools, challenges and what actually happened.
The question behind most of these experiments is not "can AI do this?" but "which part of this work is repetition that should not be done by a person?"
That framing keeps the work useful. A pipeline that writes something nobody wants is still failure — it is just faster failure.
Self-hosted n8n agents
Workflow automation running on self-hosted n8n, connecting content, data and messaging tools.
Telegram agent — @KR8DigitalsBot
A Telegram bot used as an interface to KR8 automation flows.
KR8 Onboarding Director
An automation concept to handle parts of the periodic KR8 onboarding process. Internal growth targets and schedules are never published.
Multi-agent content system
A daily multi-agent workflow producing Facebook content — research, draft, review, format.
Faceless video pipeline
A production pipeline for faceless YouTube-style video using AI generation and stock media services.
Quote and mantra workflow
A small daily automation producing quote and mantra content.
AI research systems
AI-assisted research and summarisation workflows for learning and content.
Automated distribution
Workflows that move finished content into the platforms it belongs on.
How a pipeline is built
A typical pipeline in this lab has five stages:
- Trigger — a schedule or an incoming message.
- Research — a model gathers or summarises source material.
- Draft — a second step produces the asset in the required structure.
- Review — a check step, human or automated, catches obvious failures.
- Distribution — the finished asset is delivered where it belongs and the run is logged.
Tools across these experiments include Groq, Gemini, ElevenLabs, Pexels, Pixabay, Kling and other AI video and content services.