Devices, software, AI, and the developments worth understanding.
The October 7 rollout introduced answers that can contain charts, controls, and small interactive tools. The Chat update did not change every OpenAI product in the same way.
Paid subscriptions and workspaces gain recorded-audio input for transcripts, summaries and questions, with a 512MB file limit and format restrictions.
An October 2 developer notice promised more explicit consent for an unusually broad Mac permission. Apple did not announce that every new control had already shipped.
Anthropic's September 28 release kept the headline API token prices while cutting cache-read pricing and claiming lower cost per task.
The institutional beta has separate team and project approvals. A September 30 clarification says its public form registers interest rather than completing an application.
Series 12 adds new sensing hardware, while watchOS 27 supports some older models. Later software features carry separate device and language requirements.
The main camera gains four physical aperture settings, while several headline camera specifications remain the same. Compare the change with the photographs you actually take.
OpenAI announced Astra on September 3, with broader access still to come and additional monitoring for agent actions.
Anthropic's proposed safeguards put monitoring records in customers' cloud infrastructure. Access controls, retention and cloud charges remain the customer's responsibility.
Google published a report on August 31 describing work by teams of AI agents in Antigravity, using Gemini 3.7 Flash.
Google introduced Gemini 3.5 Transcribe on August 26, giving developers separate routes for live speech and recorded audio.
Apple's M5 Ultra desktop promised up to 512GB of unified memory. Choosing a configuration meant identifying the work that needed that capacity.