This edition focuses on what directly touches a small team's budget, security and planning: the cost and permissions of AI agents, shifting search traffic, rising hardware prices and realistic expectations.
Usage-based AI services badly need hard spending limits by default
Agents produce code and API calls fast enough that a bill can balloon before anyone notices, and a warning email is not enough — you need an automatic shut-off. Why it matters: in a small company where nobody watches cloud spend daily, one runaway agent can eat a month's IT budget; set limits the day you sign up. simonwillison.net
Apple tightens full-disk access to limit how much AI agents can see
The change affects how broadly desktop agents can read user files, and has already sparked a disagreement with Meta. Why it matters: if staff install AI assistants on work machines, review what files and messages those tools can actually reach. arstechnica.com
Corporate AI spending is surging faster than organisations can absorb it
The overview notes model capability is rising and costs are falling quicker than most companies manage to adopt. Why it matters: a small firm's advantage is speed — properly automating one real process beats a long strategy document. technologyreview.com
A judge threw out lawsuits claiming Google's AI search unfairly drains publisher traffic
The court accepted that AI answers hurt content creators but found no antitrust violation. Why it matters: falling organic search traffic won't be fixed in court — invest in direct channels, email lists and referrals instead. arstechnica.com
AI demand is pushing up prices even for old hardware
Nvidia's seven-year-old media player got hundreds of dollars more expensive simply because chips are going elsewhere. Why it matters: expect sustained pressure on servers, GPUs and ordinary electronics — plan hardware refreshes earlier and budget with a margin. arstechnica.com
A reminder that language models don't reason the way people do
One of AlphaGo's builders explains why impressive answers are not evidence of actual logical thought. Why it matters: in critical processes — contracts, accounting, health — keep a human decision-maker and build verification steps into the workflow. technologyreview.com
Amazon's billion-dollar effort to ease data centre opposition drew fresh criticism
The company was praised for dropping NDAs but accused of understating environmental impact. Why it matters: the energy and emissions footprint of AI infrastructure will soon show up in tenders and client questions — know where your services physically run. arstechnica.com