The ai models & agents stories our desk judged worth your time, written so anyone can follow them.
Nvidia, Anthropic and OpenAI each set a lobbying spending record this summer, as Washington decides rules on AI, kids' safety and data center power costs.
Read more: Nvidia, Anthropic and OpenAI each spent a record amount on lobbyingAnthropic sells no product called a Claude bot. If you see one, it is either an official tool under another name or something an outside company built.
Read more: Anthropic Makes No 'Claude Bot,' So Know Which One You MeanDeepSeek has made one of its AI models openly available, while the US and China limit each other's tech. That could shape which AI tools you can use.
Read more: DeepSeek Releases Its AI Model for Anyone to Use and ModifyGoogle's new Agent Compute Engine puts each AI agent in its own locked-down virtual computer, which it says could cut costs and make agents easier to track.
Read more: Google gives each AI agent its own sealed-off virtual computerA founder says winners in each tech shift are not those with the newest software, but those who notice when customers' problems quietly change.
Read more: Staying close to customers beats having the newest tech, says one founderDiscord's latest update shows uploaded Markdown files as clean pages instead of messy symbols, and Nitro members can now send files up to 1GB.
Read more: Discord now shows Markdown files as clean, readable pagesClaude's free plan is a real product, but Anthropic does not say how many tokens Pro or Max include. It only gives multipliers.
Read more: Claude has a real free plan, but paid plans don't list token countsAndroid Authority has a guide to making a one-of-a-kind ringtone with AI, and it says the way you word your request changes the result.
Read more: You can now make custom ringtones for your Android phone with AIAnthropic said AI could copy itself across smart TVs. Security experts call the claim overblown and say the real risk is old, poorly protected home devices.
Read more: Security experts push back on Anthropic's warning about AI worms in TVsAnthropic, OpenAI and Elon Musk have each pledged donations to public causes, and the details matter as public criticism of AI grows.
Read more: Big AI companies pledge money to public causes as criticism growsAnthropic and UK investment firm Hg are backing a new European company that will advise businesses on using AI, which could make expert help easier to find.
Read more: Anthropic and Hg launch a European firm to help companies use AIOpenAI says its AI agents found two critical flaws in PHP, which 70% of websites use, and tech leaders say more trouble is coming. Here is what to do now.
Read more: Tech leaders warn of a wave of internet security flawsA Nasscom and Krisp survey finds many Indian firms are testing voice AI on customer calls, but only 13% use it widely, and most that do report big savings.
Read more: Voice AI is answering more customer calls in India, but few firms use it widelyBernstein gave Kevin O'Leary's data center firm a 0.35 out of 5 credibility score. Its CEO says the score misses key work, a reminder to look past big AI numbers.
Read more: Kevin O'Leary's AI data center firm disputes a harsh credibility scoreGitHub's $10 Copilot Pro plan now includes $15 of AI credits, but long agent sessions can use them fast. Here is how the plans compare.
Read more: GitHub's $10 Copilot plan now comes with 1,500 usage creditsLegal software makers are connecting their tools so AI can handle more tasks, which raises new questions about legal bills and how far to trust AI answers.
Read more: Legal software makers are linking their tools so AI can do more workGitHub is rolling out a default rule for Copilot business plans through October 26, so features nobody configured will follow it. Admins should check now.
Read more: GitHub will switch on unset Copilot features for business accountsPalmer Luckey and Ben Lamm raised $200 million for Warhound, a startup that wants to use gene editing to breed better military and police dogs.
Read more: Palmer Luckey's new startup plans to engineer smarter military dogsGoogle Cloud now offers Claude AI two ways. One gets new features first. The other keeps your data under Google's own agreements.
Read more: Google Cloud now lets companies run Claude AI two different waysAmazon Connect managers can now build custom screens so each call center agent sees only the customer details their job needs.
Read more: Amazon lets call center managers choose what customer info agents seeDavid's Bridal wants to take its AI wedding planner, Pearl, public as a separate company. A deal is far from done, but it shows how a retailer sees its tech.
Read more: David's Bridal plans to spin off its AI wedding planner, PearlNvidia and Reflection released open AI models that companies can run on their own servers, and Chinese rivals answered with price cuts.
Read more: Nvidia and Reflection Release Free AI Models Built for BusinessAnthropic is handing its AI skills standard to an independent foundation, which could make it easier for businesses to switch AI tools later.
Read more: Anthropic gives up control of its AI skills standardA writer for Entrepreneur says the same AI model can score 62.7% or 99.9% on one test, depending on the setup around it. That matters before you buy any AI tool.
Read more: Same AI model, wildly different results depending on its setup
