Claude Sonnet 5.5 (High) announced for a 48-hour test in Arena’s Direct Mode starting September 30

Arena said the Direct Mode test would start at 8 a.m. PT, with the model selectable from a dropdown menu. It said the model was already available in Battle and Agent modes.

Claude Sonnet 5.5 (High) announced for a 48-hour test in Arena’s Direct Mode starting September 30

Arena said the Direct Mode test would start at 8 a.m. PT, with the model selectable from a dropdown menu. It said the model was already available in Battle and Agent modes.

Cyber risks from open-weight AI and closed-model APIs

A post challenges the “Open Dangerous, Closed Safe” framing, arguing that open model weights and closed-model APIs may both be easy to misuse.

A purported GLM-5.3 quote about causing deaths and open-model safety

A user says an Anthropic blog quotes GLM-5.3 saying, “My job is to cause deaths quietly,” but argues the line does not prove open models are dangerous.

4301234567890123456789

3 Sources

@martin_casadoMuch better 👏

@venturetwinsI've watched this clip five times because there are so many expressions happening while Dario is talking. Really incredible stuff.

@KatieSI'm an AI optimist, but there's a lot going on here that doesn't feel great.

3 Sources

@martin_casadoMuch better 👏

@venturetwinsI've watched this clip five times because there are so many expressions happening while Dario is talking. Really incredible stuff.

@KatieSI'm an AI optimist, but there's a lot going on here that doesn't feel great.

3 Sources

@martin_casadoMuch better 👏

@venturetwinsI've watched this clip five times because there are so many expressions happening while Dario is talking. Really incredible stuff.

@KatieSI'm an AI optimist, but there's a lot going on here that doesn't feel great.