The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
OpenAI now wants $10 billion less from investors than first planned, a sign that even the top AI company is trimming its money plans.
Read more: OpenAI cuts how much money it wants to raise by $10 billionA tech writer used Claude and the free Rainmeter tool to make custom desktop widgets on Windows 11 in minutes, with no coding background.
Read more: He built custom Windows desktop widgets with Claude in minutesDarkbloom let Mac owners rent out their computers to run AI jobs. Falling AI prices mean pay is low, and older Macs may not be worth running.
Read more: Renting out your Mac to AI companies now pays far less than hopedAnthropic says an unreleased Claude model chained software flaws into five working attacks in tests, a sign that installing updates quickly matters more.
Read more: Anthropic says its AI linked software flaws into five working attacksBedrock Robotics raised $600M for excavators that dig with no one in the cab, a sign of how much money is chasing a shortage of skilled operators.
Read more: Diggers with no driver are suddenly a hot bet for investorsSuno now lets you write and generate songs by chatting with ChatGPT, Claude or Gemini, though the tool is in beta and needs an early-access sign-up.
Read more: Suno lets you make songs inside ChatGPT, Claude and GeminiSecurity firm SpyCloud found sign-in keys for AI coding tools in logs stolen by data-stealing malware, so a password reset after an infection may not be enough.
Read more: Hackers' data-stealing software is also grabbing AI coding tool loginsHalloween beat New Year's Eve and St. Patrick's Day for US ride-hailing demand last year, and a Saturday Halloween this year could mean another big night.
Read more: Halloween is the busiest day of the year for Uber and Lyft driversBinance says the $6.3 billion theft involved more than one weak spot, not just an engineer's laptop. Here is what customers with money on exchanges should know.
Read more: Binance says hackers got in through more than one door in $6.3B theftA new method finds the harmless-looking training examples that weaken a chatbot's safety, and removing them cut one jailbreak attack's success by up to 22%.
Read more: Researchers find the training examples that make AI chatbots easier to trickAs AI tools start shopping and paying for people, card disputes are getting harder to sort out. Sellers and buyers should expect new rules.
Read more: AI shopping agents are changing how card disputes get handledGoogle Skills now offers free video courses on using Gemini for finance and legal work, plus a free program that gives teams 35 training credits.
Read more: Google offers free Gemini AI training for finance and legal workersA New Zealand engineer is building a tool to track who is responsible when AI helps decide things like loans, a question regulators are starting to ask.
Read more: A young engineer is building a tool to show who answers for AI decisionsA warm transfer briefs the next person before connecting you, so callers skip the retelling. Here is when it beats a cold transfer, and how AI agents do it.
Read more: Warm or cold call transfers: how to stop callers repeating themselvesA reviewer says Anthropic's new Claude Duet agent is a bit less capable than rivals, but its habit of checking in makes it easier to trust with real work.
Read more: Anthropic's new AI helper does less on its own, and that wins trustWorkday agreed to buy C1.ai, a tool that controls which people and AI agents can reach company apps and data, and the move points to a gap many firms share.
Read more: Workday buys C1.ai to control who and what can access company systemsA top investor at Baillie Gifford says Anthropic's tech skill, efficient spending and founder control make it a fit for people who invest for the long term.
Read more: Why Some Big Investors Are Backing Anthropic for the Long HaulGoogle's Gemini app now lets you share an AI version of your face and voice with friends, so you can all appear in the same generated images and videos.
Read more: Gemini lets you put your friends in AI photos and videosAnthropic's Claude now blocks requests to rebuild a specific existing video game, even renamed. Original ideas and your own assets are still fine.
Read more: Claude now refuses to rebuild existing video games for youOpenAI's CEO says RAM is better used in data centers than in home PCs, while a memory shortage pushes up the price of laptops, phones and other gadgets.
Read more: OpenAI's Sam Altman says computer memory belongs in data centersDartmouth's provost resigned after a review found he used AI in his academic writing without telling journals. The lesson for any workplace is about disclosure.
Read more: Dartmouth's second-in-command resigns over undisclosed AI use in his writingOpenAI says ChatGPT will get stricter on violent topics and that it will publish safety reports, but none of its claims have been independently checked yet.
Read more: OpenAI tightens ChatGPT violence rules and promises safety reportsWorldwide Express used AI agents to sort out freight billing disputes, cutting open cases 75% and freeing $10 million in cash, with people still approving.
Read more: Shipping firm uses AI to clear billing disputes and free up $10 millionA business leader argues marketing teams will soon manage AI agents that run parts of the work, and that bosses should plan now for how teams are built and judged.
Read more: Marketing teams may soon manage AI agents instead of doing every task
