GitHub Copilot’s computer use and dynamic workflow previews, alongside Claude Code’s mods, warrant updates to our advanced agent lessons. ChatGPT’s iOS scanner and Guided Vision in Gemini Live add practical document and accessibility demonstrations, subject to roll-out and input limitations. No new Microsoft Copilot change met today’s teaching threshold.
New releases
ChatGPT. ChatGPT’s camera can capture consecutive document pages and automatically combine them into a single PDF ready to upload to a conversation. More
Claude Code. Mods package JavaScript or TypeScript event handlers inside plugins. They can rewrite prompts and tool calls, approve or deny permission requests, and add or replace interface elements. Claude can write them. 'You should know' uses a side agent to flag things the user or Claude might overlook. More
Gemini Live. Guided Vision describes the live camera view and gives spoken cues to improve framing, such as centring an object or changing the camera angle. It can help read text, identify objects and describe immediate surroundings. More
GitHub Copilot. GitHub Copilot can read accessible application content and screenshots, click, type, scroll and drag across desktop applications. This expands automation to legacy software and other applications without a suitable API or integration. More
GitHub Copilot. Dynamic workflows define a process in code inside a Copilot extension. They combine automated steps and agents, pass structured results between stages, run sequentially or in parallel, and support review checkpoints, pausing and resuming. More
Training your team on ChatGPT, Claude or Gemini? See our ChatGPT training, Claude training and Gemini training.
Elsewhere in AI
California’s attorney general issued an investigative subpoena to OpenAI as part of an inquiry into potential cybersecurity vulnerabilities involving its AI models. Read
Amazon is seeking to offload $8bn of Nvidia chips to investors to improve its balance sheet as spending on artificial intelligence rises. Read
OpenAI fired workers for mishandling sensitive information; the former employees were investigated for sharing data with an outside AI evaluation group, the BBC reports. Read
Anthropic urged Australia to consider conditional approval for AI training on copyrighted works under an opt-out system, while broadcasters ABC and SBS sought media regulation. Read
Bank of England governor Andrew Bailey warned that the AI investment boom could trigger market shocks, saying the central bank is watching investment carefully. Read