Follow the change ·
What happened in AI and robotics — and the change behind it. A reading path through 19 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 companies plan a Seoul system to test mixed-chip AI inference.
The plan combines AMD CPUs and GPUs with FuriosaAI neural processors and MangoBoost data processors and software.
A Seoul reference cluster is planned for spring 2027 to test performance and interoperability.
Lower inference costs and power use are goals, not demonstrated results.
Uber reported about 500 agent-sourced food orders and four rides during one week.
The figures cover one week, not the quarter.
The agent services, markets and failed attempts are unspecified.
An assistant search followed by an app checkout differs from an agent placing an order.
OpenAI outlines local commitments and separate policy proposals for compute campuses.
OpenAI promises to cover power costs and provide financial assurances.
Covered sites would have water plans, public reports and local input.
The separate blueprint asks governments to set broader rules for developers; it is a proposal, not enacted law.
Dharma reports extraction results and costs.
Schemas specify fields; open requests do not.
Dharma reports higher benchmark scores.
A trial is linked; weights are unconfirmed.
Researchers compare robot touch cues.
One fingertip yields multiple touch cues.
Matched demonstrations support comparison.
Findings concern tabletop trials.
NVIDIA describes agents investigating GPU cluster issues.
A runtime routes chat requests to specialists.
Engineers review evidence before acting.
NVIDIA reports internal use but no measured accuracy or time savings.
Researchers simulate adaptive ocean sampling.
Forecasts estimate where uncertainty is highest.
Currents constrain the robots’ options.
Observations reshape the next forecast.
Researchers test changing simulated manipulation tasks.
Six tasks include variations in pose, layout and clutter.
Generated and human-authored programs share robot skills.
Simulated results do not establish real-world reliability.
