Daily Brief

ByteDance's Seed team steers the company into autonomous driving

The Arena

ByteDance's Seed team steers the company into autonomous driving

Physical AI is pulling the big platform players toward the road: ByteDance is lining up a driving effort built on its world models, and the companies already putting robotaxis on city streets are now fighting about the rules that govern them. ByteDance is moving into autonomous driving, with its world-model research team spearheading an early-stage push that sources say is aimed at unmanned logistics. The effort is being led by the world-model group inside Seed, ByteDance's flagship large-model

Samsung puts HBM base dies on a logic node to reclaim bandwidth

The Stack

Samsung puts HBM base dies on a logic node to reclaim bandwidth

The memory stack feeding AI accelerators is getting a quiet overhaul. At Hot Chips 2026, Samsung laid out its plan to build High Bandwidth Memory base dies on a logic process node — a step change that frees up die area it wants to fill with memory controllers, sensors, and eventually compute. Samsung revealed it now fabricates its HBM4 and HBM4E base dies on a 4nm logic node, breaking with the long-running practice of building them on a DRAM process. The move is a direct answer to the memory wa

Apple cuts video-AI latency 5x by internalizing visual thinking

The Frontier

Apple cuts video-AI latency 5x by internalizing visual thinking

A new Apple machine-learning paper takes aim at one of the more expensive habits of multimodal models: rendering future frames to "see" where a video is heading. Its move is to push that thinking into training — and drop frame generation at inference entirely. Apple researchers show video models can learn to reason visually during training and then skip frame generation at inference, matching or beating the standard approach while cutting end-to-end latency more than fivefold. The method, descr

Nvidia's Groq 3 LPX inference chip enters full production

The Stack

Nvidia's Groq 3 LPX inference chip enters full production

Nvidia commercialized its largest acquisition at Hot Chips 2026 on Monday, while a New York startup betting that video games can teach robots how to move pulled in fresh capital at a $6 billion valuation. Here's what moved in AI hardware and physical AI. Nvidia said its dedicated inference accelerator Groq 3 LPX has entered full production, turning the company's record $20 billion purchase of Groq into a shipping product. The chip is built as an extension of the Vera Rubin data center platform,

Taiwan indicts nine over AI servers allegedly smuggled to China

The Guardrails

Taiwan indicts nine over AI servers allegedly smuggled to China

Taiwan's export-control enforcement just got sharply personal. Prosecutors in Keelung indicted nine people — including employees of Nvidia and Super Micro — for allegedly helping high-end AI servers reach China despite US and Taiwan bans, in a case that confirms the black market around frontier chips is alive and well. Taiwanese prosecutors in the northern city of Keelung indicted nine people, including Nvidia and Super Micro employees, for allegedly colluding to illegally export AI servers to

AWS patches SDK flaw that turned a region into credential theft

The Guardrails

AWS patches SDK flaw that turned a region into credential theft

A cloud and silicon day: Amazon quietly closed a class of bug that let a region field leak live credentials, and China's homegrown AI-chip push got a fresh public-market test with Enflame's STAR Market launch. AWS quietly patched a flaw across seven of its software development kits after security firm Pi Inc. traced one bug report to roughly 2,500 copies of the same defect. The weakness sits in how the SDKs assemble an endpoint: the region value drops into the middle of a hostname like https://

## Thomson Reuters builds its own legal AI instead of renting from OpenAI

The Arena

## Thomson Reuters builds its own legal AI instead of renting from OpenAI

Thomson Reuters builds its own legal AI instead of renting from OpenAI Thomson Reuters launched "Thomson," its first in-house language model for legal work — built on Alibaba's open Qwen rather than fine-tuning a frontier model from OpenAI or Anthropic, a bet on owning AI instead of renting it. The company says it spent about $40 million over two years on the project (the final training run alone came to roughly $450,000). Thomson starts from Alibaba's Qwen — most recently Qwen3.5-397B — which

ShengShu lays out a five-level roadmap to general world models

The Frontier

ShengShu lays out a five-level roadmap to general world models

ShengShu founder Zhu Jun used a World Robot Conference keynote to put a name and structure on where embodied AI is heading: a five-level roadmap for a general world model he argues must unify generating the world, interacting with it, and acting inside it — not a stack of separate models bolted together. Zhu, the ACM/IEEE/AAAI fellow who built video model Vidu, described the general world model as a closed loop rather than a single generator or robot policy: a system that understands the world,

Ireland reconsiders its nuclear ban as data centres eat a quarter of the grid

The Guardrails

Ireland reconsiders its nuclear ban as data centres eat a quarter of the grid

Ireland is Europe's data-centre capital, and it's running out of cheap ways to power the AI boom. Lawmakers are now reopening a debate the country closed in 1999, when it banned nuclear power — a sign of how hard the AI buildout is bending national energy policy in the world's smaller grids. Ireland is reconsidering its 1999 ban on nuclear power, driven by data centres that now account for about a quarter of the country's electricity use. The Financial Times reports that politicians are reopeni

**IBM's dual-architecture chip brings Arm-native AI to mainframes**

The Stack

**IBM's dual-architecture chip brings Arm-native AI to mainframes**

IBM's dual-architecture chip brings Arm-native AI to mainframes IBM used the Hot Chips 2026 conference to unveil the first concrete product to come out of the strategic partnership it signed with Arm back in April — a new dual-architecture processor designed to run both IBM's z/OS and Arm-native Linux workloads on the same cores, a first for the mainframe world. The unnamed chip will power the next-generation IBM Z mainframes and LinuxONE systems, and in doing so it aims to pull Arm's sprawling

Autonomous Russian drone killed three in a strike Ukraine calls a first

The Guardrails

Autonomous Russian drone killed three in a strike Ukraine calls a first

A fully autonomous, AI-guided Russian drone killed three Ukrainians at a gas station in Zaporizhzhia — an attack Ukrainian officials describe as a milestone in weaponry untethered from human control. Two developments anchor the morning: the deadly autonomous strike itself, and the quiet fact that the computer guiding it is the kind of civilian chip Nvidia sells for robot vacuums and machine-vision toys. Ukraine says a fully autonomous Russian drone — guided by an off-the-shelf Nvidia chip — kil

Cerebras CS-4 doubles accelerator performance on the same chip

The Stack

Cerebras CS-4 doubles accelerator performance on the same chip

A new rack-scale AI accelerator, a nine-figure robotics raise, and fresh evidence that chatbots quietly steer sensitive-health decisions — all in the last few hours. Cerebras has introduced the CS-4, a rack-scale AI accelerator its CEO Andrew Feldman calls the fastest system in the industry. The machine still runs on the same 5nm WSE-3 wafer as its predecessor, but Cerebras says it doubles the CS-3's performance by pushing clock speed higher through more power and better cooling. A single rack

See all