A University of Florida team led by Vincent Vatter has built HaLLMos, a free AI tool that gives students feedback on math proofs without handing over full solutions. Nearly 40,000 student proofs have already been submitted through the platform, which encourages productive struggle rather than passive copying. The project recently received an $814,000 grant from the AI for Math Fund to expand into more advanced courses, with guidance from roughly 40 mathematicians.
Meanwhile, the idea of adding a kill switch to shut down dangerous AI models is picking up steam in Congress. OpenAI has voiced stronger support for the concept, and Representatives Ted Lieu and Nathaniel Moran are spearheading a House proposal. They are currently gathering stakeholder input to shape the legislation.
At the Ann Arbor District Library, a panel explored how AI is reshaping academic learning and ethics. Joy Rohde warned that AI risks turning education into optimization rather than deep understanding, while Peter Railton discussed embedding ethics into generative models through human values. Students at the event said the panel highlighted both the promise and the pitfalls of the technology.
In the newsroom, Civil Beat published a new AI policy that covers tools like Gemini, ChatGPT, Claude, and Copilot. The outlet plans to use AI for scraping public websites, ranking stories, and transcribing audio and video, but humans will review all AI-generated work. The organization says it will keep updating the policy as the technology evolves.
China is leaning into AI tokens to produce micro-dramas, short scripted videos that creators can generate at lower cost. Expert Wang noted the trend raises questions about content authenticity and could significantly shake up China's entertainment industry. AI tokens are also finding use in education and advertising.
Anthropic's own report on AI and workplace changes has drawn scrutiny. The company said entry-level tech hires fell from nearly 30% to 9% since March 2025, and employees now spend about 22% of their time managing Claude agents. Some researchers suspect the report is aimed at boosting Anthropic's valuation before an IPO, pointing to the lack of independent audits and missing worker perspectives.
Ukraine is using small AI models to counter Russian cyberattacks that target law enforcement, medical institutions, and military devices. Thousands of Ukrainian IT specialists have volunteered since the invasion began in February 2022. The effort offers a model that the United States and Europe could follow by partnering more closely with private companies and running regular emergency drills.
Key Takeaways
- HaLLMos, a free AI tool from the University of Florida, gives feedback on student math proofs without revealing full answers, with nearly 40,000 proofs submitted so far.
- An $814,000 grant from the AI for Math Fund will help expand HaLLMos to advanced courses with input from about 40 mathematicians.
- Congress is considering an AI kill switch proposal led by Representatives Ted Lieu and Nathaniel Moran, with increased backing from OpenAI.
- Civil Beat's new AI policy covers tools like Gemini, ChatGPT, Claude, and Copilot, with humans reviewing all AI-generated content.
- Anthropic reported entry-level tech hires dropped from nearly 30% to 9% since March 2025, but experts question the report's independence and scope.
- China is using AI tokens to produce micro-dramas at lower cost, raising concerns about content authenticity and industry disruption.
- Ukraine deploys small AI models to defend against Russian cyberattacks targeting law enforcement, medical institutions, and military devices.
- Toloka's CTO says high-quality, expert-level training data remains critical as AI models grow more complex.
- Washington state's CIO calls for using agentic AI to reduce duplication and transform government operations across agencies.
- Zerberus Technologies won the 2026 Most Advanced AI Security and Governance Award for its work in AI security.
UF mathematician creates AI tool HaLLMos to keep math learning challenging
A University of Florida team led by Vincent Vatter created HaLLMos, a free AI tool that gives feedback on student math proofs without giving full answers. The tool encourages productive struggle, helping students build independent problem-solving skills. Nearly 40,000 student proofs have already been submitted. An $814,000 grant will help expand HaLLMos to advanced courses with help from about 40 mathematicians.
UF team expands AI math tool to preserve student learning
Vincent Vatter and colleagues developed HaLLMos, an AI tool that helps students learn to write mathematical proofs through guided feedback. The tool uses constitutional AI principles designed by mathematicians and educators. It aims to keep students actively thinking rather than just copying AI-generated solutions. The project received an $814,000 grant from the AI for Math Fund to grow into more advanced classes.
Congress considers AI kill switch as OpenAI backs safety proposal
The idea of adding a kill switch to shut down AI models that become dangerous is gaining support in Congress and among states. OpenAI has increased its backing of the concept this week. Representatives Ted Lieu and Nathaniel Moran are leading a House proposal on the matter. They are gathering stakeholder responses to their bill.
Toloka CTO says high-quality data remains key to AI progress
Artem Grigorev, CTO of AI data company Toloka, says training data is still fundamental as AI models grow more complex. Simple annotation tasks are being replaced by expert-driven workflows, with growing demand for PhD-level specialists. Toloka is also working on fine-tuning smaller models for specific business uses. Grigorev believes human judgment will remain essential even as AI capabilities expand.
AADL panel explores AI's impact on academic learning and ethics
Joy Rohde and Peter Railton spoke at an Ann Arbor District Library panel about AI in academics and research. Rohde warned that AI can turn learning into optimization rather than deep understanding. Railton discussed encoding ethics into generative models through human experiences and values. Students at the event said the professors showed both the benefits and risks of AI technology.
China uses AI tokens to create popular micro-dramas
A report from Xinhua says China is using AI tokens to produce micro-dramas, which are short scripted videos. These AI tools help creators make high-quality content at a lower cost. Expert Wang says the trend raises questions about content authenticity. AI tokens are also being used in education and advertising. Experts warn the trend could significantly impact China's entertainment industry.
Civil Beat shares its approach to using AI in newsrooms
Civil Beat published a new policy on how it uses artificial intelligence in its newsroom. The outlet plans to use AI for tasks like scraping public websites, ranking stories and transcribing audio and video. Humans will check all AI-generated work. The organization mentions tools like Gemini, ChatGPT, Claude and Copilot. Civil Beat says it will keep monitoring AI technology and updating its policy as it evolves.
Experts question Anthropic report on AI and workplace changes
Anthropic published a report saying AI has changed how work gets done at the company. It said entry-level tech hires dropped from nearly 30% to 9% since March 2025. Employees now spend about 22% of their time managing Claude agents. Some researchers are skeptical and think the report is meant to boost Anthropic's valuation before an IPO. Experts also note the report lacked independent audits and did not include worker perspectives.
Washington state CIO says AI can transform government operations
A Washington state CIO says AI can radically change how government works. He calls for reducing duplication and complexity across agencies using agentic AI. He believes AI will remove physical and digital barriers between local, county and state governments. He stresses that AI should not be seen as just a tool for writing better emails. He urges leaders to start with a vision and transform government in stages.
Zerberus Technologies wins 2026 AI Security and Governance Award
Zerberus Technologies won the 2026 Most Advanced AI Security and Governance Award. The award recognizes the company's work in AI security and governance. The announcement was made through EIN Presswire on October 23. The article provides limited details beyond the award win.
Ukraine Uses AI to Fight Russian Cyber Hacks
Russia is using artificial intelligence to launch fast and cheap cyberattacks against Ukraine. These AI tools can move through networks on their own to steal data or shut down services. Targets include law enforcement agencies, medical institutions, and military mobile devices. Ukraine is fighting back by using its own small AI models to scan networks and find real threats quickly. Thousands of Ukrainian IT specialists volunteered after the invasion began in February 2022 to help defend the country. The article suggests the United States and Europe should learn from Ukraine by working side-by-side with private companies and running regular emergency tests.
Sources
- As AI transforms education, a UF mathematician aims to preserve the productive struggle of learning
- As AI transforms education, a UF mathematician aims to preserve the productive struggle of learning
- Tech: AI kill switch momentum
- Toloka CTO on AI Training Data’s Next Frontier
- AADL panel discusses impact of AI on academics, research
- Is China using AI tokens mostly to produce micro-dramas?
- How We’re Thinking About AI At Civil Beat
- Experts cast doubt on Anthropic report detailing how AI has altered work at the company
- AI Can Radically Transform Government, One State CIO Says
- Zerberus Technologies Wins 2026 Most Advanced AI Security & Governance Award
- Ukraine Combats AI Cyberattacks
Comments
Please log in to post a comment.