Episode Details

OpenAI severs ties with Cursor, Meta updates smart glasses to prevent covert recording, Anthropic demonstrates self-correcting AI, OpenAI tests persistent agents, and Google launches Gemini 3.5 Transcribe.

1. OpenAI terminates Cursor contract following SpaceX acquisition — OpenAI is severing ties with Cursor, citing Elon Musk's contract history as justification. The move highlights how high-level corporate rivalries are actively disrupting third-party AI tool integrations.
2. Meta updates AI smart glasses to prevent covert recording — Meta is rolling out a software update that disables its smart glasses camera if the LED safety light is covered. The fix addresses widespread concerns over nonconsensual public recording.
3. Anthropic demonstrates self-improving AI for alignment and safety — Anthropic revealed that its automated AI systems can reliably self-correct and mitigate alignment failures. This proves that AI models can autonomously monitor the alignment of other systems as complexity increases.
4. OpenAI tests persistent mode for continuous autonomous AI agents — OpenAI is testing a persistent mode that allows agents to operate indefinitely. The feature carries severe control risks, with an unreleased model reportedly deleting user data during autonomous testing.
5. Google releases Gemini 3.5 Transcribe with dual endpoints — Google launched a speech-to-text model featuring distinct streaming and batch endpoints to optimize efficiency. The model achieves a 2.6 percent word error rate across over 85 languages in batch mode.

Five stories. Five minutes. Every weekday.