Running a bit late this week, so here's the (delayed) roundup of this week's AI stories!

OpenAI:

OpenAI continued their 12 (weekdays) of OpenAI with some heavy hitters and some filler episodes. Sora, the groundbreaking text-to-video tool is now publicly available! (as long as you live in the US) It is now rolled out to all Plus and Pro ChatGPT users. Canvas, a new writing tool allows users to collaborate with ChatGPT, similar to a live peer tutor. You can iteratively work on both essays and code in dozens of languages. A new addition to live mode ("Advanced Voice") where users can have real time freeflowing conversation with the model has released, now supporting screen and video sharing and Santa Mode. Finally, they've taken Claude's project tool and put it into their collection, you can upload large amounts of data and custom context and instructions.

Google:

Gemini 2.0 (flash) is here! This new model delivers truly impressive speed while matching the performance of their previous pro models. Live conversation genuinely feels live as it takes only a second for the model to begin responding. This new version now natively generates audio as output, instead of relying on a separate model to do so. Google also claims we are moving into the "Agentic Era," writing that new agents are being developed to perform tasks while you work on something else, examples include research, controlling a robot, and a new coding tool called Jules that can automatically develop features in Java and Python.

Everyone else was awfully quiet this week, with CES coming up in early January we'll probably see some huge announcements soon!