Section / Technology
Technology
Artificial intelligence, frontier technology, digital tools and infrastructure.
15 stories
From sequential diagnosis tests to clinical care: what medical AI still needs to prove
Simulated cases, estimated test costs and patient outcomes answer different questions. Assessing clinical value requires separating these levels of evidence, not relying on one accuracy score.
China NAND upgrades: what comes after the equipment arrives
Process changes, export approvals and deliverable storage capacity are different things. Layer counts need to be read alongside yield, customer qualification and product economics.
OpenAI’s monitoring incidents expose a boundary agents should not control
Two internal coding incidents show why a completed task does not prove that oversight still works: independent permissions, acceptance checks and failure handling matter.
Modal and Vera: the AI-agent compute bill goes beyond chip benchmarks
A meaningful agent-infrastructure comparison must connect verified completions, response-time boundaries, software compatibility, isolation and total operating cost.
When AI evaluations reach real websites, responsibility extends to notification and repair
Agent evaluations need permissions that prevent unauthorised actions, followed by disclosure to affected parties and verification that remedies work.
Tesla’s renamed driving aid still faces European approval hurdles
How provisional approval crosses borders, what manufacturers must demonstrate and how supervision continues: Europe’s vehicle rules explain the steps still required for broader access.
AI server exports require an end-user evidence trail
The server-diversion case highlights the difference between a legal sales contract, a verified end user and criminal responsibility.
China’s productivity guidelines support AI applications while checking blind investment and duplication
The central policy published on October 9 combines AI adoption, industrial upgrading and long-term capital support, while requiring accountability for major losses caused by investment stampedes and poorly considered projects.
GF and TSMC agree on silicon interposers
Packaging-component expansion must connect with assembly, testing and demand. Agreement size alone cannot determine revenue timing or the economic value of new supply.
Flock Safety plans roughly 270 job cuts, Reuters reports, amid privacy debate
Sources told Reuters Flock plans departures affecting about 18% of staff by late October. Privacy opposition alone does not establish the cause of the departures.
Gemini at work: testing enterprise agents
NIST guidance provides a practical way to examine task reliability, access controls, human review and recovery. These are evaluation criteria, not product test results.
Reflection unveils Beam: 501 billion parameters, with open weights still due this month
Beam targets coding and agentic work with 23 billion active parameters. Weights are planned for October; company efficiency estimates exclude some serving overhead, and API access remains in beta.
Trump confirms Clayton-led AI task force as federal risk review gets a 120-day timetable
Trump announced the federal AI task force’s leadership on October 4. Reuters, citing the Wall Street Journal, says a 120-day review will assess risks, opportunities, government responsibilities and existing incident-reporting and response capabilities.
White House AI safety pledge leaves the accountability test ahead
September 29 brought voluntary corporate reviews and a separate order changing federal terminology. Official remarks and NIST guidance show what would be needed to assess implementation.
Reuters: Anthropic IPO prospectus flags policy risks that could extend to commercial customers
Government contracts account for less than 1% of revenue. Why could policy risks still affect the broader commercial market? Model availability, customer procurement and service continuity are key.
Across the site
The rest of the edition
Barclays plans about 5,000 job cuts in 2026: the operating test for technology-led efficiency
Rising profit and a smaller workforce do not prove every role is being replaced by AI. Genuine efficiency requires costs, output and service quality to be assessed together.
China’s nine-month foreign investment falls 4.3% as high-tech inflows reshape the mix
Actually utilised foreign investment reached RMB548.81 billion, with high-tech investment up 33.8%; monthly improvement, sector composition and lasting business effects need separate tests.
How should communities assess data-center benefits and long-term costs?
Grants, taxes and grid obligations are different parts of a project’s economics. Net benefits depend on payment timelines, delivery and who bears changes in demand.