Archive

Daily edition November 20, 2026

Daily AI-generated issue

AI engineering, November 20, 2026

CoreWeave measures 10× throughput per megawatt, Google opens Genie 4 to developers via Gemini API and Google AI Studio, TSMC and Samsung suspend some China chip shipments

This issue is researched and written by AI models, and every fact is checked against its cited source. No human edits it before it is sent.

  • Agents
  • APIs
  • Audio
  • Benchmarks
  • Business
Edition
Daily
Published
November 20, 2026
Read time
3 min
Stories
3

You can change the weather inside an AI-generated world while exploring it. Genie 4 takes live prompts to change characters and action, with audio that adapts along the way.

Meanwhile, CoreWeave measured 10× the DeepSeek R1 throughput per megawatt on NVIDIA Vera Rubin NVL72 versus NVIDIA GB200 NVL72, at matched per-user throughput. TSMC and Samsung suspended some chip shipments to customers in China after new US export controls.

If you read one thing today: Genie 4. Test world generation with the interactive canvas and starter apps in Google AI Studio, then use the Interactions API to integrate it into your app.

Key takeaways

  1. 01The rack connects 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs through a 260 TB/s all-to-all NVLink 6 fabric.
  2. 02Native NVFP4 support comes through the Transformer Engine, with NVFP4 precision used in the comparison.
  3. 03The inference setup combined large-scale expert parallelism, multi-token prediction, and disaggregated prefill and decode, enabled by NVIDIA TensorRT-LLM and NVIDIA Dynamo.
  4. 04CoreWeave completed bring-up and validation in early June, verifying power, cooling, networking, and compute at production scale.

Top News · Official · coreweave.com

In CoreWeave's DeepSeek R1 evaluation, NVIDIA Vera Rubin NVL72 delivered 10× the tokens per second per megawatt of NVIDIA GB200 NVL72. Both systems were measured at a matched per-user throughput target.

Here's what went into the test:

  • The rack connects 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs through a 260 TB/s all-to-all NVLink 6 fabric.
  • Native NVFP4 support comes through the Transformer Engine, with NVFP4 precision used in the comparison.
  • The inference setup combined large-scale expert parallelism, multi-token prediction, and disaggregated prefill and decode, enabled by NVIDIA TensorRT-LLM and NVIDIA Dynamo.
  • CoreWeave completed bring-up and validation in early June, verifying power, cooling, networking, and compute at production scale.

One catch: CoreWeave calls this a starting measurement and expects further gains as it tunes scheduling, networking, and software.

Sources: wf.coreweave.com

Top News · 288 HN points

Genie 4 is Google DeepMind's interactive world model, now available to developers through the Gemini API via the Interactions API and through Google AI Studio. It generates navigable worlds from text, a reference image, or both.

Here's what you can build with it:

  • Continuous world generation runs at 720p and 24 frames per second, with synchronized native audio.
  • Real-time keyboard controls and live prompts steer the world, while Gemini reasoning helps keep its state coherent.
  • Live edits can change weather, characters, and action during navigation.
  • Project Genie's browser experience is rolling out to eligible Google AI Ultra subscribers. Google AI Pro members can join the waitlist.

One catch: Google identifies ongoing limitations in camera latency, intricate physical interactions, real-time audio precision, and long-term scene permanence.

Try it: Open the interactive canvas or starter apps in Google AI Studio, or use the Gemini API via the Interactions API. Also available via Google DeepMind

Sources: blog.google · blog.google

Top News · Reported · ft.com

If you are waiting for chip deliveries in China, some shipments from TSMC and Samsung are now on hold. Both chipmakers suspended some shipments to customers in China after the US Commerce Department introduced new export controls.

Here's what changed:

  • TSMC suspended some chip shipments to customers in China.
  • Samsung also suspended some chip shipments to customers in China.
  • The halt followed the introduction of new US Commerce Department export controls.

One catch: The suspension covers some shipments to customers in China, not a blanket halt to all shipments from TSMC and Samsung.

Why care? If your deployment depends on chip deliveries to China, confirm whether the specific shipments you are waiting for are affected before planning around their arrival.

Sources: ft.com

Also in this issue

Signals

Keep reading