AI
AI technologies that enhance my productivity and accelerate processes in areas like coding, research, content checking, and translation.
APPROACH
CONTEXT = 'ai'; MODE = 'practitioner'
Summary
AI technologies are driving significant transformation in business and production processes today. For developers and makers, a 20–50% increase in production speed is reported, along with significant time savings on repetitive tasks. These tools automate routine work while facilitating focus and freeing up time for more strategic or creative areas. However, the view that the human factor remains critical in areas like creative thinking, analysis, and decision-making is prominent. Therefore, on this site, in publications and training, processes are conducted in a human-centered manner; AI is used as a supportive tool.
Where I Use AI
Coding:
Debugging, code reviews, optimization suggestions
Research:
Finding resources, fact-checking, alternative perspectives
Content Checking:
Grammar, consistency, flow
Skeptical Analysis:
Questioning claims, verifying sources, generating counterarguments
Turkish Spelling:
TDK spelling checks, detecting awkward phrasing
Translations:
Technical documentation, multilingual content
Blog Posts:
My own voice, my own experiences
Image Generation and Diversification:
AI-powered image creation, variation generation, concept visualization
What I Build
My Claude Code plugins, MCP servers, ComfyUI nodes and apps are on the Lab page.
TOOLS I USE
QUERY = 'tools'; TYPE = 'external' // 10 rows
Coding & Development
Local Models
Qwen3-Coder-Next
Qwen's MoE coding model. I run it locally with MLX for code review and offline fixes.
Qwen
Alibaba's open source multilingual model family. With Qwen3.8 I delegate everyday agentic tasks to it locally.
Gemma
Google's open source model family. I run Gemma 4 locally for Turkish and English content work.
Nimble and Tev1
Decision models served through Ollama's /v1/systemone endpoint. I am piloting them for structured decisions such as choice, yes/no and scoring.
GLiNER2.5
Fastino's encoder-based decision and classification model. I am testing it as an alternative to an LLM for labeling and classification.
mmBERT and Laya
Multilingual encoder models: mmBERT and Laya, which is built on it. I am testing them for multilingual classification, Turkish included.
BGE-M3
Multilingual embedding model. It powers the vector layer of the semantic search in my own MCP servers.
Nomic Embed
Lightweight embedding model. I use it for local search work that needs fast indexing.