Telegram bots
Chat-based tools for workflows, notifications, utilities and personal projects.
Case study · Experiments
AI & Bot Experiments
A collection of small bots, automations and local AI workflows built to turn specific problems and ideas into working tools.
Why I build them
Small problems are good excuses to experiment.
Most of these projects begin with a repeated task, a small annoyance or simple curiosity. Instead of looking for a giant all-in-one platform, I prefer building something focused.
That makes the result useful as a real tool and useful as a way to learn APIs, automation, deployment and local AI.
Experiment areas
Different ideas, same goal: make something useful.
Local AI
Experiments with Ollama and local models for private, self-hosted AI workflows.
DeepTutor
Local tutoring and research workflows connected to personal knowledge sources.
Automation
Turning repetitive manual work into scripts, web tools and notification workflows.
APIs
Connecting public and private APIs to build focused tools around structured data.
Deployment
Testing lightweight hosting and simple ways to keep small projects running reliably.
Approach
Build first, simplify second.
The first version only needs to prove the idea works. After that, I simplify the interface, remove unnecessary features and tighten the workflow.
That keeps experimentation fast without turning every small idea into a giant software project.
Typical stack
Whatever gets the idea working quickly.
PythonTelegram Bot APIOllamaLocal LLMsREST APIsGitHubCloud hostingHTML / CSS / JS
Some experiments remain private prototypes, while others eventually become standalone tools or larger projects.
What I learned
Useful beats complicated.
A simple bot with a clear purpose can be more useful than a complex app with too many features.
Working across local AI, APIs and automation also makes it easier to understand when a tool is genuinely useful and when a manual workflow is actually better.
Project status
Always changing.
This page is a home for experiments that do not need their own full case study yet. As individual projects grow, they can graduate into their own pages.