We explore topics related to artificial intelligence, model evaluation, security, and the development directions of the field.
We are expanding the QFlexAI portfolio with two new products in active development: Qflex AgentGuard — real-time protection for AI agents that use tools, and Qflex Verum — RAG application for verifiable answers grounded in the organization's sources.
Read full articleQFlexFrame is a structured decision-making framework that makes the reasoning behind AI method choices transparent, surfaces uncertainty, and shortens the path from problem to working solution. Available for both mature AI teams and organisations building their AI capability.
Read full articleWe have launched the first private version of the QFlexTest platform — a controlled and reproducible environment for systematic testing and evaluation of AI models. The platform covers the entire testing lifecycle, from test generation to evaluation and control.
Read full articleWe explore the fundamental difference between producing correct outputs and understanding context. Inspired by Elkhonon Goldberg's work on the cognitive gradient, we analyze how Neuro-Symbolic AI can reconcile the flexibility of neural networks with the rigor of symbolic systems.
Read full articleIn a period of accelerated development, AI is approaching general competence. We analyze how open models become "cognitive cores" and how specialized extensions like LoRA adapters can build expertise in concrete domains.
Read full articleBeyond the noise and marketing, innovation in AI follows four main paths: inventing new architectures, refining existing ones, fine-tuning with proprietary data, and building multi-agent systems. We explore each direction and its strategic implications.
Read full articleFrom Cartesian geometry to the Transformer, we explore the mathematical foundations of AI, the birth of the field, the AI winters, and current challenges related to the complexity and interpretability of modern models.
Read full article