Yevgen`s official homepage
Yevgen Somochkin
"ΞΣΣΘ"
CTO ESSO.DEV
Building the Future of AI & Automation | Full-stack & DevOps Architect | n8n | Scalable, Data-driven Web & Mobile Apps | Salesforce x2 certified
Hamburg, Germany
Yevgen`s official homepage
Yevgen Somochkin
"ΞΣΣΘ"
CTO ESSO.DEV
Building the Future of AI & Automation | Full-stack & DevOps Architect | n8n | Scalable, Data-driven Web & Mobile Apps | Salesforce x2 certified
Hamburg, Germany
Recent Blog Posts
Feb 1, 2026
Feb 1, 2026
Feb 1, 2026
Popular Questions
Best for: email routing, ticket categorization, lead scoring, data extraction, summarization. Avoid: complex reasoning, long-form generation.
BitNet exposes OpenAI-compatible API. In n8n, use HTTP Request node to localhost:8080. Deploy both via Docker Compose.
BitNet for efficiency (2-3x faster, 3-5x less memory). Ollama for simplicity. llama.cpp for model variety.
Yes! Pi 5 runs at 8-12 tok/s, suitable for voice commands, IoT classification, and chatbots. Pi 4 works at 3-5 tok/s.
Minimum 4GB RAM, 4-core CPU. No GPU needed. Works on Raspberry Pi 5, old laptops, mini PCs. Only ~1.2GB memory during inference.
| Sep |
|---|
| Oct |
|---|
| Nov |
|---|
| Dec |
|---|
| Jan |
|---|
| Feb |
|---|
| Mon |
| Wed |
| Fri |
Details
