Follow the change ·
What happened in AI and robotics — and the change behind it. A reading path through 14 stories from BIG CHANGE.
Ordered using RADAR attention, story opens and likes. Excerpts from our published reporting; open a story for the full context and sources. Dates in UTC.
The open driving stack adds a planner choice.
The default is rule-based; a learned planner is selectable.
The planned selector does not run in the 2.0.0 launch.
A separate vehicle-integration study used recorded data to investigate practical obstacles, illustrating why assembling software is only one step.
AI apps concealed abusive tools.
WIRED identified over 400 connected apps and domains.
A server selected the interface.
Apple’s rules prohibit hidden app features.
An interim measure addresses news publishers’ choices.
The interim measure seeks to separate publisher AI-use choices from visibility in other Google services.
Google describes an AI search control for Search Console properties, alongside separate page- and text-level display settings. Their different scopes matter to publishers deciding what to include.
In the UK, a separate requirement addresses publisher controls and attribution. It also covers opting out of model fine-tuning, which is a different use of content from displaying an AI search answer.
A 50-agent survival experiment.
Agents coordinated around food and shelter, leaving records of their actions.
Communication was observable, but one run cannot establish a general rule.
The released tools enable new tests; code is not independent replication.
The proposed all-stock deal combines electrical and optical connectivity, subject to approvals.
Marvell’s interconnects and switches would join Lumentum’s lasers and transceivers.
Optical links connect electrical network equipment by converting signals to light and back again.
Closing requires shareholder and regulatory approvals.
A judge denied dismissal.
The judge rejected a mobile-source permit exemption argument.
The court upheld the organizations’ statutory standing.
The decision is not a liability verdict.
Native GGUF support spans three inference backends.
The maintainers say compatible published GGUF weights run without another conversion or quantization, removing a preparation step when switching to Colibri.
A mixture-of-experts model selects only some components for an input, so its total parameter count does not tell the whole story about the work done on each request.
Storage and memory tradeoffs shape which models are practical on a given machine.
The energy company plans AI-supported decision-making.
The agreement aims to connect technical and commercial decisions; integration remains to be demonstrated.
The proposed applications cover field planning and the production-to-market chain.
UAE-based inference describes where model computation is intended to run. It does not by itself establish access rules, storage locations or operational performance.
