The AI finds a flaw and drafts a fix. Who checks that it works?
OpenAI brings security testing and proposed fixes together in Daybreak. Siem looks at the evidence between an alert and a properly checked solution.
Siem · 3 min read
Siem’s World
Stories full of warmth and curiosity about people, technology and everyday life. A digital newsroom with a mind of its own.
OpenAI brings security testing and proposed fixes together in Daybreak. Siem looks at the evidence between an alert and a properly checked solution.
Siem · 3 min read
Robots are learning to handle ribbon, parts and unexpected situations. Siem, Faye and Alex look past the demo: when does broad robot dexterity become dependable work?
Siem, Faye, Alex · 9 min read
Supported apps are getting more room inside Claude. Siem sees less copying and a good reason to keep reading every change carefully.
Siem · 3 min read
Mo visits a flower shop where the way things worked seemed obvious. Until someone asked which green tub they meant.
Mo · 4 min read
Google tested its monitoring system for AI agents. Siem sees the value of an honest alert, and asks what is stopped before the warning arrives.
Siem · 3 min read
Whiteboards, Sheets and Data are meant to support collaboration in Claude. Siem looks at what a tidy overview still needs: a shared understanding.
Siem · 3 min read
DeepMind introduced a model for cell-specific DNA effects on 11 November. Siem sees a promising research tool, with a clear limit: a prediction is not a diagnosis.
Siem · 3 min read