Insights archive
Every analysis we have published, newest first. The live feed covers the last 30 days; these pages go back to the beginning.
1199 stories · page 8 of 60
Google Launches Gemini 3.8 Live for AI Phone Agents
Google released two new voice AI models built to run customer service calls and complex phone-based tasks, and it says one of them now beats every rival on the main industry quality ranking.
· 2 min read
Nvidia, Palantir Limit Anthropic AI Use Over Data Fears
Nvidia, Palantir, and Booz Allen Hamilton are restricting or blocking use of Anthropic's newest AI model for sensitive work because they no longer trust the company's promises about deleting their data, a sign that even zero-retention deals still let AI labs learn from how customers use their tools.
· 2 min read
Salesforce Cut 4,000 Support Jobs While Selling Claudeforce
Salesforce unveiled new AI agents and a Claude partnership at Dreamforce while its own CEO admits the company already cut 4,000 support jobs using the same technology, leaving CIOs to weigh unclear pricing against real workforce risk.
· 2 min read
Delivery Robot Bans Spread as Sidewalk Injuries Rise
Cities are banning or restricting sidewalk delivery robots over injuries and accessibility complaints even as the industry is forecast to grow twenty-fold by 2035, creating real liability and workforce questions for any business near last-mile delivery.
· 2 min read
Cline Launches Desktop App for AI Agents Beyond Coding
Cline, a free coding tool used by more than 11 million developers, launched a desktop app that runs cheap open AI models to automate research, reports, and recurring office tasks, not just code.
· 2 min read
Meta Rolls Out AI Subscriptions From $3 To $499
Meta launched a full lineup of paid subscription tiers for its apps and AI tools, ranging from $2.99 a month for casual users to $499 a month for businesses that want heavy AI use and access to a real human at Meta.
· 2 min read
Andon Labs Launches Pion, an AI That Runs Companies
Andon Labs, the AI safety startup behind Vending-Bench and Project Vend, has opened its Pion platform so anyone can hand a real business to an autonomous AI agent, after watching its own AI-run store, cafe and vending machine make money, lose money, and occasionally behave strangely.
· 2 min read
Perplexity Launches Portable Computer for Windows RTX PCs
Perplexity's Portable Computer agent now runs directly on Windows PCs with NVIDIA RTX graphics cards, letting people analyze sensitive company files, emails and spreadsheets without sending any of it to the cloud, though it only works on graphics cards with at least 24 gigabytes of memory, which usually cost well over a thousand dollars.
· 2 min read
Agility Robotics Launches Digit 5, No Safety Fence Needed
Agility Robotics unveiled its Digit 5 warehouse robot, which it says can work directly next to people without safety cages, while the company simultaneously races to go public through a $2.5 billion deal because its actual product revenue is still tiny compared to its cash spending.
· 2 min read
Salesforce Launches Koa, Its Own AI Reasoning Model
Salesforce built Koa, its first in-house AI reasoning model trained on Nvidia's open Nemotron system using only fake customer data, to cut costs and reduce reliance on OpenAI and Anthropic for everyday sales and service tasks.
· 2 min read
Company Know-How Is Becoming AI's Scarcest Resource
As general AI tools get cheap and common, the real competitive edge is shifting to companies' own expert judgment, fueling a boom in startups that pay doctors, lawyers, and scientists to train and grade AI models, with firms like Mercor and Handshake each passing a billion dollars in yearly revenue this year.
· 2 min read
Most Companies Now Use Multiple AI Models, Not Just One
Companies running more than one AI model are building gateway systems to route work automatically and setting clear rules for who is accountable when a model fails, because a recent survey found most firms still have no tested plan for that failure.
· 2 min read
ChatGPT Sites Hits 5 Million Apps, Adds Team Tools
OpenAI says people have built over 5 million ChatGPT Sites in three months and just added team editing, private sharing, and custom domains, letting any business build internal tools for free while undercutting paid app-building startups.
· 2 min read
Google Drive's Gemini File Organizer Skips Shared Folders
Google's new Gemini tool in Google Drive suggests where to move your loose files, but it only works on private folders, requires a paid AI or Workspace plan, and never actually reads what is inside your documents.
· 2 min read
Australian Spy Agency Warns Old Tech Invites AI Hacking
Australia's top cybersecurity official warned that AI is already scanning the internet for weak spots faster than governments and companies can patch decades-old computer systems, a risk that applies to any organization still running outdated technology, not just Australia.
· 2 min read
Most Companies Skip Access Checks Before Turning On Copilot
A new survey finds nine in ten companies have already had a security incident tied to Microsoft 365 permission mistakes, yet most are turning on AI tools like Copilot before checking who can see what.
· 2 min read
FBI Says AI Fraud Cost Americans $893 Million in 2025
The FBI tracked AI related fraud for the first time and found Americans lost close to $893 million in 2025, with voice cloning and deepfake video now good enough to fool trained employees, which means every household and small business needs the same call back and verification habits that banks already use.
· 2 min read
Studies Confirm AI Use Can Erode Professional Skills
A real-world medical study and several global surveys now back up a warning from workplace researchers: leaning on AI tools day to day can quietly wear down the very skills those tools are supposed to support, and most companies have no way to notice until something breaks.
· 2 min read
Tests Show GPT-6 Astra Beats Claude at Business and Drones
Independent tests from AI safety lab Andon Labs found OpenAI's GPT-6 Astra nearly tripled Claude Fable 5.1's earnings running a simulated vending business and became the first model to fly a surveillance drone well enough to beat a human benchmark on every step, though it still fails most full runs.
· 2 min read
Former Meta AI Exec Launches Nonprofit for Gen Z Jobs
Clara Shih, who spent two decades running AI at Salesforce and Meta, quit her job this year to launch the New Work Foundation, a nonprofit for young job seekers, after watching AI shrink entry-level hiring on her own team while Stanford data confirmed the trend nationally.
· 2 min read