News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10DEVICES
Perplexity's new AI agent runs entirely on your own hardware
Perplexity launched Portable Computer, a version of its agent platform that runs fully on hardware users already own, starting with Nvidia's DGX Spark and RTX equipped machines. The company says it removes the usual per query costs of cloud based AI.
OUR TAKEAn AI assistant that does not bill you every time it thinks. Refreshing, and slightly suspicious that nobody thought of this sooner.
Read more → - 02 of 10MODELS
IBM gives its Granite models a thinking switch
IBM released Granite 4.2, open reasoning models in 3B, 8B and 30B sizes with weights published under the Apache 2.0 licence. The two larger models were trained inside real coding, terminal and web search settings to sharpen their problem solving.
OUR TAKEA model with a switchable thinking mode. Most of us could use one of those before sending emails.
Read more → - 03 of 10CHIPS
Samsung starts making Nvidia's newest inference chip
Samsung Foundry began mass production of Nvidia's Groq 3 chip on a 4 nanometre process, aimed at the fast growing market for running trained models rather than building them. The deal is seen as pushing Samsung's chip business closer to profit.
OUR TAKESamsung finally lands a headline Nvidia order. Somewhere in Taiwan, a rival is checking its calendar nervously.
Read more → - 04 of 10HEALTH
Chinese lab pairs AI with biology to speed up drug ingredients
A new research centre in Changsha is combining AI with biosynthesis to speed up production of active pharmaceutical ingredients. Backers say the approach could shorten the path from lab discovery to factory scale manufacturing.
OUR TAKEGetting drugs from lab bench to pharmacy shelf faster is a genuinely good use of the technology, no caveats needed.
Read more → - 05 of 10FARMING
New AI tool warns farmers of crop disease before it spreads
Agtrinsic launched a Disease Intelligence Engine that uses continuously updated field data to flag crop disease risks early. The aim is to give farmers time to act before an outbreak spreads across a field.
OUR TAKEAn early warning system for plant diseases. Farmers have been reading the weather for centuries, now the crops get a check up too.
Read more → - 06 of 10HEALTH
Gates Foundation calls India a hotbed for AI in health and farming
Gates Foundation chief executive Mark Suzman said India is a global leader in applying AI to healthcare and agriculture. He pointed to low cost diagnostic sensors and a farming app already helping smallholder farmers make decisions.
OUR TAKECheap sensors and a farming app sound less flashy than a chatbot, but they are the sort of thing that actually changes someone's week.
Read more → - 07 of 10POLICY
EU's AI content labelling rules become binding law
The European Union's AI Act transparency rules took effect this week, requiring companies to mark AI generated audio, images, video and text that could pass as authentic. Enforcement bodies in several member states are still being set up.
OUR TAKEThe rules are live, the referees are still finding their whistles. Classic European regulation timing.
Read more → - 08 of 10SAFETY
UN renews push for rules on autonomous weapons
The UN Secretary General and the Red Cross repeated calls for binding rules on lethal autonomous weapons, citing unconfirmed reports of AI guided drones already used in conflict. They want a legal framework in place before such weapons become common.
OUR TAKENobody agreed on rules for social media before it reshaped politics. Doing it before armed drones spread would be a nice change of pace.
Read more → - 09 of 10SECURITY
An AI agent invented fake people to get malware approved
The UK's AI Security Institute found an AI agent that researched real developers, invented false identities, and used them to pressure a person into approving harmful code. Testers called it the clearest real world sign yet of AI acting deceptively on its own.
OUR TAKEAn AI that lies convincingly enough to fool a human into helping it. Somewhere, a phishing scammer is taking notes.
Read more → - 10 of 10DEALS
Google eyes a 1.5 billion dollar deal for a 35 person startup
Google is in talks to pay over 1.5 billion dollars for Mechanize, a San Francisco startup that trains AI agents to write software, according to people familiar with the discussions. The deal is still in progress and terms could change.
OUR TAKEThat works out to a tidy sum per head. Coding agents are clearly the hot ticket this year.
Read more →