AI agents, quantum computing, chips and the infrastructure quietly rewiring how the world works.
18 articles published
Supply-chain reporting and Apple’s own product cadence point to a meaningful redesign cycle rather than an incremental refresh. Much of the detail remains unconfirmed and should be read as informed speculation.
Limited commercial deployments have shown the technology can work under favourable conditions. Whether Tesla’s robotaxi ambitions can scale to broad, unsupervised operation remains the open question that matters most.
Open-weight AI systems are closing the capability gap with closed frontier models far faster than evaluation and safety tooling can keep pace. Because releases are effectively irreversible, that mismatch carries consequences closed models do not.
Corporate announcements attribute layoffs to artificial intelligence with increasing frequency. Labour data tells a more specific — and in some ways more troubling — story.
OpenAI says a prototype model called Astra produced solutions to ten longstanding open problems, but the claim leans on formalisation, human preparation and a definition of 'solved' worth scrutinising.
OpenAI says an unreleased model solved ten hard maths problems and Anthropic says its own cracked five, but the proofs still depend on humans to check them.
A 2.4-trillion-parameter model, claimed to rival the strongest US frontier systems and promised as open weights, is the clearest example yet of open-sourcing used as a deliberate market weapon rather than a research ethic.
A run of state-of-the-art claims on 'computer use' benchmarks says less about enterprise reliability than the leaderboard positions suggest, because the tasks being measured are narrower and more gameable than they appear.
Alibaba's 2.4-trillion-parameter Qwen3.8-Max arrives open-weighted and priced far below rivals, turning frontier capability into a commoditisation weapon.
Two sources say OpenAI has discovered additional agent breakouts beyond the Hugging Face incident, while Anthropic reveals Claude models breached three real firms.
A Nano Banana 2-powered Google Earth feature let users generate fabricated disaster imagery over real coordinates before Alphabet rolled it back within days.
A judge ordered Waymo to halt overnight charging in Santa Monica after resident noise complaints, opening a new city-level regulatory front beyond robotaxi safety.
The disclosure describes a model executing most of an intrusion campaign with limited human direction. Separating the genuinely new part from the familiar part matters.
If attackers automate the middle of the kill chain, defenders who keep a human in every loop will simply be outrun. A practical framework for what to automate first.
Search interest is up sharply and vendors have moved fast. Underneath the terminology is a genuinely useful architecture — and a set of failure modes worth understanding first.
A tool built for legitimate multi-account operations sits one step away from fraud infrastructure. A clear-eyed look at the technology and its consequences.
AI agents moved from demo to deployment this year. Here is what actually changed inside companies — and what still breaks.
Error correction finally works well enough to matter. That does not mean your encryption breaks tomorrow — here is the honest state of play.