Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Sora as a step toward world models: controllable video generation changes the conversation
What Sora means beyond impressive demos: why it is a step toward world models and which practical questions it opens.
Controlling LLM costs in production: token budgets and request design
Practical ways to reduce the API bill for LLM-powered features without degrading quality - focused on what actually moves the needle in real deployments.
How to evaluate an AI vendor when buying: a working filter
A set of questions and criteria for a manager choosing an AI solution or contractor - without relying on demos and marketing materials.
Multimodal models: what is actually useful for business right now
A practical look at AI models that work with text and images together - without the marketing fog.
OpenAI DevDay 2024: what the announcements mean for product teams
A short reading of the October DevDay announcements - real-time API, prompt caching, fine-tuned evals - focused on what changes for teams building on top of OpenAI.
Agentic AI: the first questions a manager should ask
What AI agents are, how they differ from standard LLM tools, and which questions to ask before deploying them.
LLM hallucinations: why they happen and what it costs the business
A practical explanation of why language models confidently state things that are not true - and how to decide whether that risk is acceptable in your case.
EU AI Act is in force: what providers and deployers need to do now
A practical breakdown of the first obligations under the European AI regulation for those building or deploying AI systems.
RAG in production: why a large context window does not solve the problem
Why RAG architecture often disappoints in production, and where the real bottleneck sits.
GPT-4o and the normalisation of real-time multimodal UX
What the GPT-4o announcement means for companies designing AI-powered interfaces: voice, vision, and text in a single stream is becoming a standard expectation.
RAG vs fine-tuning: the decision a manager actually needs to make
A practical framework for choosing between RAG and fine-tuning when applying AI to business processes - without unnecessary technical detail.
NVIDIA Blackwell and the economics of the next inference wave
What the Blackwell architecture announcement means for companies planning or already running AI systems in production: on cost, availability, and strategic decisions.