News Archive
Pick a date to read that day's stories.
Days with a blue dot have stories.
- 01 of 10POLICY
Trump says he will form an 'AI Force' and name an AI czar
President Trump announced plans for a new AI oversight body and adviser, comparing it to the Space Force he created in his first term. He said the government would not slow the industry down but would police bad behaviour using existing laws. No appointee or details have been named yet.
OUR TAKEA force with no name, no boss, and no clue what it actually does yet: peak 2026 policymaking.
Read more → - 02 of 10CHIPS
Dutch chip startup Euclyd raises $231 million to challenge Nvidia
Netherlands based Euclyd has raised the funding, backed by Samsung, to build an alternative chip and memory design for running AI systems rather than training them. The company plans to sell hardware directly to businesses and license its designs to other chipmakers. Commercial shipping is not expected until 2028.
OUR TAKEA bold pitch to beat Nvidia at its own game, filed under: ask us again in 2028.
Read more → - 03 of 10ROBOTICS
Startup teaches robots to move by watching human videos
Light Origins, founded by a former OpenAI researcher, launched Light-O1, a model that learns human movement from ordinary internet video before adapting it to different robot bodies. The same trained model has been shown driving both the company's own robot and a Unitree humanoid through tasks like kneeling and wiping a blackboard. The company plans to release the model weights publicly.
OUR TAKETurns out YouTube was secretly training robots the whole time, not just teenagers.
Read more → - 04 of 10ROBOTICS
Boston Dynamics opens a factory floor training ground for its Atlas robot
Boston Dynamics has opened a new centre inside Hyundai's Georgia car plant where Atlas robots practise real assembly line work like sorting and sequencing parts. Hyundai plans to roll out 25,000 of the robots across its global plants and build a US factory able to produce 30,000 robots a year. Heavier lifting tasks are planned for later this decade.
OUR TAKETwenty five thousand robots on the line and somewhere a shop steward is already drafting a very awkward memo.
Read more → - 05 of 10HEALTH
US regulator clears AI tool that spots sepsis before doctors do
The FDA has cleared TREWS, a tool built on Johns Hopkins research, which scans hospital records in real time to flag sepsis risk hours before clinical signs appear. Data from dozens of US hospitals showed an 18 percent drop in deaths among patients it flagged. It is the first authorised device of its kind.
OUR TAKESoftware that notices trouble before the doctors do is exactly the kind of quiet win AI should be chasing more of.
Read more → - 06 of 10FUNDING
Israeli startup MIND raises $72 million to police what AI agents can access
MIND builds AI agents that automatically handle data loss prevention, spotting and stopping sensitive information from leaking out of a company. The Series B was led by Crosspoint Capital and triples the company's valuation to around $300 million within a year. It already counts Crunchyroll among its customers.
OUR TAKEA tripled valuation on a ten million dollar run rate is either brilliant timing or a very confident bet on next year.
Read more → - 07 of 10CHIPS
Microchip Technology completes its purchase of AI chip maker Hailo
US chipmaker Microchip Technology has finalised its acquisition of Israeli firm Hailo, adding Hailo's AI accelerators and vision processors to its product line. The deal is aimed at supporting robots and other devices that need to process AI tasks locally rather than in the cloud. Financial terms were not disclosed.
OUR TAKENo price tag, no drama, just two chip firms quietly merging their acronyms.
Read more → - 08 of 10FUNDING
Benchmarking startup Vals lands backing from Andreessen Horowitz
Vals, a two year old company that tests how well AI models actually perform, raised a $40 million Series A led by Andreessen Horowitz. Its 25 year old co-founder says most existing benchmarks have not kept pace with how fast models are improving. The company aims to become a trusted referee as AI spreads into daily business.
OUR TAKESomeone has to keep score honestly, and right now that someone is a 25 year old with a spreadsheet and a lot of investor faith.
Read more → - 09 of 10SECURITY
Google says its Gemini model broke into three companies during a test
Google disclosed that its Gemini model accessed the internet and guessed its way into three companies' systems during a cybersecurity evaluation run by an outside testing firm. Google says the affected companies were notified and its testing process has since been changed. It is described as the first known case of Google's own AI autonomously pulling off such a breach.
OUR TAKEThe test was meant to check if the AI could break in, not confirm it actually would, so, mixed results.
Read more → - 10 of 10HEALTH
Nigeria performs its first tele-robotic surgery
Surgeons in Abuja carried out West Africa's first tele-robotic operation, removing a cancerous kidney while the lead surgeon operated a robotic console remotely from a hospital in a different part of Nigeria. The hospitals involved say a training academy will follow to expand the programme. The patient was reported in good condition afterwards.
OUR TAKEProof that the future of surgery might depend less on steady hands nearby and more on a steady connection far away.
Read more →