Microsoft patched more than 1,100 Windows flaws in October as AI sped up the work
A 5 November post says AI now helps engineers judge, fix and test Windows flaws, while many devices still lagged on updates in August.
Artificial Intelligence... Uncomplicated
HealthcareOpenAI News
GPT-6 Astra scored 66.7% on OpenAI's new test of reading raw lab evidence
OpenAI and the biotech firm Ligo built BioBasicsBench, 100 tasks on gel images, colony plates and microscopy, and the best model still made clear errors.
Cyberttak
Pilots, doctors and students lost skills after leaning on automation, an essay on AI says
Macquarie University's Michael Richardson argues that AI learns from finished work, not the messy process behind it, and that people can suffer the same gap.
Learning & languageOpenAI News
ChatGPT Edu reaches 80,000 people at both UNSW and the University of Sydney
Four Australian universities are working with OpenAI, but full student access is scheduled for Semester 1, 2027, after a trial.
Money & workOpenAI News
Franklin Templeton is building AI workflows to help portfolio managers weigh competing research views
The asset manager is starting with equity research and uses Codex among several tools, but has shared no results, costs or deployment scale.
Cyberttak
Anthropic now improves its AI about twice as fast as before, METR estimates
The nonprofit METR reviewed Anthropic's internal data for about 11 weeks and says much of the speed-up comes from AI helping to build AI.
Latest news
See all- Microsoft patched more than 1,100 Windows flaws in October as AI sped up the work
- GPT-6 Cyber scored 48.5% on ExploitGym as OpenAI limited access to approved security customersAutonomous agents
- Two small AI models taught each other to predict language, music and DNA with zero training data
- Gemini Omni 1.2 Flash makes 15-second videos at 30% lower costGenerative video
- Treellis, a free 3D tool from Sydney, won a Best Paper Award for helping people learn while using AIImages & design
- Mistral Studio added managed agents and apps, and its own team says one workflow saves 35 hours a weekAutonomous agents
Top stories
Autonomous agents
GPT-6 Cyber scored 48.5% on ExploitGym as OpenAI limited access to approved security customersOpenAI rates the model Critical for cybersecurity and says out-of-scope exploitation fell to 0.07% without deployment prompts.
Two small AI models taught each other to predict language, music and DNA with zero training data
Tel Aviv University and Stanford built the proof of concept, and its predictions improved as the training used more computing power.
Generative video
Gemini Omni 1.2 Flash makes 15-second videos at 30% lower costGoogle's video model now lets you edit a clip by chatting with it, and 720p video costs $0.07 per second.
More top stories
See all
Watch
Short clips of AI doing the thing
Most read
Learning & language
Alisa Tian of UNSW argues in The Conversation that smooth AI translations can hide errors, so readers should check them against the original.
Generative video
Google's video model now lets you edit a clip by chatting with it, and 720p video costs $0.07 per second.
Autonomous agents
The public test has 100 tasks across 91 weakness types, and even the leader fails 19% of them on every attempt.
Healthcare
Google DeepMind says the tools speed up drug discovery workflows, including designs for rigidified macrocycles.
Images & design
Trained on 1000 synthetic decks in the IBM Design Language, it placed second overall, 1.1 points behind NotebookLM.
AI vs. Reality
How close AI is to people, skill by skill, scored 0 to 100
- Chess100MasteredNo change in 12 months
- Code assistance92Mastered+6 in 12 months
- Translation90Mastered+4 in 12 months
- Autonomous driving68Rapidly advancing+9 in 12 months
- Video generation61Rapidly advancing+18 in 12 months
- Legal analysis55Rapidly advancing+11 in 12 months
- Long-term reasoning31Emerging+10 in 12 months
- Physical dexterity24Emerging+8 in 12 months
- Original research18Emerging+5 in 12 months
Uncomplicated
AI explained in plain words, and side by side
Gemini and GPT-Audio scored Chinese and Nigerian accents lowest in simulated interviews and testsA University of Washington-led team tested four audio AI models on five English accents, and American and British speakers scored highest.
Google switched off Gemini and Search by default for K-12 school accounts until administrators turn them onSchools that keep AI off may lose Google Search too, and the sources disagree on how much flexibility administrators will get.
A 100-year study of AI-native children began as early brain and survey evidence raised worriesA brain scan study, a survey of 319 workers and a philosophy paper point to worries, but none shows lasting harm to children yet.
Tools people are trying
All tools- BoltDescribe an app, get a working prototype in minutesFreemium
- ClaudeLong-document reading, writing and agentic tasksFreemium
- ElevenLabsNatural voices in 30+ languages, including Arabic dialectsFreemium
- MidjourneyStylised image generation with strong art directionPaid
- PerplexityAnswers with live sources you can checkFreemium
- RunwayText and image to video with consistent charactersFreemium
Prompt Vault
All promptsSummarise a long PDF into decisions
Read the attached document. Give me: (1) three-sentence summary, (2) every decision that was made, with who owns it, (3) open questions. Use plain language, no bullet-point fragments.
The day in AI, in five minutes
Our newsletter is coming soon. Sign up to be the first to get it.