Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Voice agents reached operations: what separates a working one from an expensive demo
ElevenLabs says its voice agents run more than 15 million conversations a week, and they now handle returns, insurance actions and bookings. Why voice became the first agentic AI segment with a real production use case, what separates a working agent from a polished demo, and what to settle about recordings and personal data before the pilot starts.
The FTC turns to autonomous agents: what you need to be able to prove about yours
A US regulator is asking whether existing law is enough to hold companies accountable for what autonomous AI systems do. While the legal frame is argued over, the engineering requirement is already clear: audit trail, permissions, sandboxing and human approval stop being optional.
Gemini 4 Argon: you now pick a model for a process, not for an answer
Google unveiled Gemini 4 Argon with a focus on coding, long multi-step tasks and cybersecurity, still in limited access. What changes in model selection once the unit of work is a process: long context, autonomy, checkpoints, and the questions worth asking before a pilot.
Local AI: running a model so your data never leaves the perimeter
An honest engineering and cost picture of local AI: what your own perimeter actually buys you, the ladder of options from Ollama to air-gapped, and when a hosted business tier is enough.
Can you delete your data from an already-trained AI model
Why removing yourself from a trained model is nothing like deleting a database row, what actually sits behind unlearning and the right to erasure, and where the real leverage is.
Can an AI hand my personal data to another user
An honest split: where a model can actually recall someone else's data from training, and where the real leak comes from your own system - RAG, logs, and access control.
How to stop AI from training on your prompts and chats
An honest walkthrough for an individual and for a company: where chatbots train on your conversations by default, how to switch it off, and why opt-out never erases what a model already learned.
Robotics after LLM: why the next question is not chat, but action in the world
LLMs changed the interface for interacting with computers. The next shift is physical systems that understand context and act in the real world.
AI does not fix bad data
A short note on why an AI rollout in a company starts not with the model, but with the quality of the data underneath.
The next evolution of Agents SDK: long tasks, sandbox, and a production-ready agent
Tools for building AI agents are maturing. What this means for companies thinking about real deployments rather than demos.
Long context in LLMs: what it changes for business tasks in 2026
Modern models support context windows of hundreds of thousands of tokens. What this practically changes for companies and where the real limits are.
EU AI system definition guidelines: how this affects real products
The European regulator published guidelines on defining an AI system. What this means for companies that build or use AI-powered products.