Ömer Asım
software, research and observation

I work on artificial intelligence, software, and how knowledge is organised and used. This site brings together the tools I develop, the questions I research, and the notes I make along the way.
From İnovapark to the present
İnovapark is an important part of my work: it began with software development in 2005 and gradually extended to organisational structures, decision processes and ways of working. Its consulting period ended in 2026. Today, the İnovapark Open Archive is gradually sharing the documents and methods from that work. Questions of knowledge, responsibility and decision-making encountered there also run through my current projects.
Tools and experiments
KLM — An open-source record contract and validation tools for reading a model answer alongside its sources, claims and evaluation trail. The KLM paper, co-authored with Cihan Aydıner and available as an arXiv preprint, discusses the record model, separate internal experiments and negative results.
nage — A platform project addressing the knowledge lifecycle, model execution and access boundaries together.
a2 — A Python agent that checks solutions externally and investigates the reuse of validated skills. Its first open-source technical preview has been released.
Muhit — Design work on interface consistency, preserving context and keeping people in control of AI suggestions.
Circle Visual Engine — A visual prototype exploring what size, light and motion can express within a fixed field of circles.
Research
Autorite asks what a system preserves as it changes. ERM and RGM address distinct questions about relational memory and retrieving evidence through context.
Evidrel and BMK explore relations, supporting evidence and local rules. ADL investigates what learning interventions contribute under the same budget. In this work, I try to keep experiments, observations and unanswered questions distinct. Current status is recorded on each project page and in Now.
Around the work
Photography is a way of looking at Istanbul and the details of everyday life. The Library, language, literature and music also belong here. The writing follows developing thoughts, the Discovery Notebook holds small finds and experiments, and the Atlas traces their conceptual connections.
Selected work
All projectsa2
A Python agent that checks solutions externally and investigates reuse of verified executable skills. Development alpha.
KLM — Knowledge Layers Model
An open-source record contract and validation toolkit for knowledge provenance, evidence relationships and assessment traces in language-model outputs.
nage — knowledge and model operations
A platform project connecting model execution, the knowledge lifecycle and access boundaries.