Chatbot privacy fails by default, not by users
A survey finds people confide intimate details in chatbots while distrusting the providers behind them. My own project shows why that gap persists.
Sections
Artificial Intelligence
Models, tooling and the question of what survives contact with production. We cover language models, agents and infrastructure from the perspective of people who have to run this stuff.
TypeSafe AI launches Jev, a model that answers in numbers
Instead of text, Jev returns probabilities for yes/no questions, choices and ratings. Only the input is billed.
OpenAI halves API prices with GPT-6 Sol and Luna
Anthropic also cuts Claude Opus 5.5 prices by 20 percent, while GPT-5.6 is scheduled for a 25 percent increase in November.