AI News: 24 Things You Need To Know This Week
Matt Wolfe showed 17 AI tools in this video on . 6 are verified available in the EU. Each row below jumps to the exact second. Part of week 2026-43.
Tools shown in this video, with timestamps
Anthropic's proactive personal assistant agent that pulls from connected business apps, shows inline dashboards and drafts messages. Waitlist for Pro/Max users; the creator doesn't have access yet and couldn't demo it.
OpenAI used GPT-5.6 to identify a 20-year-old PHP vulnerability before attackers did and shared a tool to check vulnerable PHP versions. Exploitation attempts surged after announcement and some sites went down to patch.
Platform for hiring AI agents that work like employees; you can now sign in with an existing ChatGPT plan. Sponsored segment.
Recap of days 12-17: improved effective usage via context reuse across sub-agents, ability to edit previous messages in the desktop app, 12-hour API cache and permanent cache for logged-in users, plus several reset days.
ChatGPT feature that lets you lock chats behind a PIN, rolling out gradually.
Now generally available, built with GPT-6 Luna; returns bounded answers such as probabilities, choices from a set, or scores rather than free text.
Ideogram's new model generating clean, editable SVG vector graphics. The creator tested it with early access; some outputs had extra limbs but about half looked solid, and elements can be edited and downloaded as SVG.
Live demoTool combining Seedance 2.5 and ElevenLabs V4 to lip-sync an image to an audio file. The creator found lip sync decent but the output ignored the original image and the UI had bugs, including a 30-second audio limit error.
Live demoPerplexity open-sourced two decision models, 4B and 27B parameters, with the 4B shown leading on a benchmark chart.
Perplexity's agent can now run fully locally on a Mac, with 36 GB or more of unified memory recommended, instead of only on Perplexity's cloud computers.
Yelp can now call restaurants on your behalf to check wait times and ask about menu items, even as Yelp blocks third-party agents.
Mistral's voice agent model that handles multiple speakers with real-time diarization, interruptions, and tool calls, shown in a demo video.
Canva's first image generation model, built with the Leonardo AI research team. Available in Canva's Dream Lab tab and the Canva API.
Rumor that OpenAI is testing a new image model; Aurora-named models have appeared on leaderboard sites and in OpenAI's API. Not confirmed.
Zoom introduced an enterprise tool for visibility over AI activity, including deepfake risk detection to tell whether you're talking to a real human on a call.
Microsoft's medical diagnosis model, claimed to be 1.7 times more accurate than physicians and more than twice as accurate as physicians using textbooks and the internet.
Viral Unitree demo video of the H2 humanoid doing flips, kicks, jumps and lifting a car. The creator doubts it's autonomous and questions the car-lifting clip.
- Sign into Hyperagent with your existing ChatGPT plan, or choose a paid plan and get $100 in bonus creditshyperagent.com
- Explore AI Tools & Newsfuturetools.io
- Claude Duetclaude.com
- PHP Security Vulnerabilitiestheverge.com
- OpenAI Patch the Planetopenai.com
- Railway Security Breachblog.railway.com
- ChatGPT Locked Chatstechcrunch.com
- OpenAI Decisions APIopenai.com
- Perplexity Computer on Macperplexity.ai
- Websites Blocking AI Agentstheverge.com
- Yelp AI Restaurant Callstechcrunch.com
- Mistral Voxtral 3mistral.ai