This observation mainly covers 8:47 am on 27 August to 8:47 am on 28 August 2026 in Australia/Sydney and includes five publicly available, verifiable technical changes.
Anthropic opens a Model Hardware Standard research preview
On 27 August, Anthropic opened a research preview of the Model Hardware Standard (MHS) to an initial group of scientific laboratories and advanced manufacturers. MHS uses standardised drivers, read-and-write primitives and device description files to connect instruments with programmable interfaces; agents can control multiple devices through MCP, a command-line interface or APIs. Compared with the previous need to write integrations device by device, the new fact is that a model-agnostic hardware interface has now been trialled with microscopes, liquid handlers, robotic arms and a quantum laser system. QuEra confirmed that Claude used MHS to develop and validate a laser auto-relock controller on about US$700,000 of test equipment. The deployed controller is a deterministic, inspectable program with the AI outside the runtime auto-relock loop, while the tuning workflow still uses AI. The standard remains a limited research preview, is not yet open source, and still requires work on physical-reasoning limits, safety evaluations and additional device drivers.
Google releases Gemini Omni 1.1 Flash
Google released Gemini Omni 1.1 Flash on 27 August and began making it available through the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform. The update increases the existing footage considered during video extension from the final one second to as much as ten seconds, permits ten-second extensions up to a cumulative 40 seconds, and adds first-and-last-frame interpolation, 360p drafts and upscaling to 4K. Claims that 360p generation is up to 60 per cent faster and costs one-third as much as 720p come from Google. No independent quality evaluation has been published, so consistency and complex camera control remain developer-reported capabilities.
Gemini 3.5 Transcribe enters public API preview
Google released Gemini 3.5 Transcribe and the real-time gemini-3.5-transcribe-live on 26 August. They are provided through the Interactions API and Live API respectively, with support for more than 85 languages, custom vocabulary and automatic text clean-up; file transcription also provides word-level timestamps and speaker identification for up to three people. Developer and enterprise access is in public preview, the Gemini app on macOS is limited to English, Android Rambler is limited to selected countries and languages, and Chrome support has not launched. Artificial Analysis independently recorded 2.6 per cent AA-WER for the file model and 4.0 per cent for the streaming model. Google's claim of a 70 per cent reduction in final-transcription time relative to Chirp 3 remains a developer comparison.
NVIDIA Vera CPU moves from early hand-offs to scaled shipments
On 27 August, NVIDIA updated material first published on 18 May to state that its first custom CPU, Vera, had begun shipping at scale and that AWS had received its first Vera CPU server and a Vera Rubin GPU. Earlier recipients included OCI, Anthropic, OpenAI and SpaceXAI. Vera has 88 Olympus cores and 1.2 TB/s of memory bandwidth and targets CPU work such as agent orchestration, tool use, sandboxing and reinforcement learning. The claim of up to 1.8 times greater per-core performance is based on NVIDIA testing and has not been independently reproduced publicly. AWS confirmed that Vera infrastructure is coming to its cloud but did not provide a date for general customer availability; delivered evaluation systems are not equivalent to a generally available AWS service.
Pollen Robotics opens Microduck pre-orders and its open development stack
Pollen Robotics, owned by Hugging Face, opened pre-orders for Microduck on 27 August at US$399, targeting first deliveries before Christmas 2026 in North America and Europe. The 25-centimetre biped has 15 motors, a camera, LiDAR, two IMUs and an articulated beak for grasping objects. Its official site also makes available an Apache-2.0 software stack, MuJoCo simulation, reinforcement-learning workflow and seven pretrained movement policies. What is confirmed today is the opening of pre-orders, code and demonstrations, not volume delivery. Camera, LiDAR, radio, SDK-language and age-guidance specifications remain provisional, and no independent long-duration test of real-world performance has been published.
Sources
Anthropic, Previewing the Model Hardware Standard, 27 August 2026
https://www.anthropic.com/news/model-hardware-standard-research-preview
Reuters, Anthropic unveils new framework allowing AI agents to operate physical devices, 27 August 2026
https://www.reuters.com/technology/anthropic-unveils-new-framework-allowing-ai-agents-operate-physical-devices-2026-08-27/
QuEra Computing, Holding the Light: Teaching an AI to Lock and Tune our Quantum Computer's Lasers, 27 August 2026
https://www.quera.com/blog-posts/holding-the-light-teaching-an-ai-to-lock-and-tune-our-quantum-computers-lasers
Google, Gemini Omni 1.1 Flash lets you build with more control, 27 August 2026
https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-omni-1-1-flash/
The Decoder, Google's Gemini Omni 1.1 Flash makes AI video generation cheaper and more flexible, 27 August 2026
https://the-decoder.com/googles-gemini-omni-1-1-flash-makes-ai-video-generation-cheaper-and-more-flexible/
Google, Intelligent transcription with Gemini 3.5 Transcribe, 26 August 2026
https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5-transcribe/
Artificial Analysis, Speech to Text Providers Leaderboard, accessed 27 August 2026
https://artificialanalysis.ai/speech-to-text/non-streaming
NVIDIA, Delivering Vera: NVIDIA's First CPU Built for Agents Is Shipping Now, updated 27 August 2026
https://blogs.nvidia.com/blog/vera-cpu-delivery/
Amazon, AWS and NVIDIA expand partnership for next-gen AI infrastructure, 26 August 2026
https://www.aboutamazon.com/news/aws/aws-nvidia-2-million-gpus-ai
Reuters, Nvidia begins Vera CPU sales pitch to Chinese clients, sources say, 12 June 2026
https://www.reuters.com/world/china/nvidia-begins-vera-cpu-sales-pitch-chinese-clients-sources-say-2026-06-12/
Pollen Robotics, Microduck, 27 August 2026
https://pollen-robotics.com/microduck/
The Verge, Hugging Face's new robot is an adorable rollerskating duck, 27 August 2026
https://www.theverge.com/gadgets/985549/hugging-face-microduck-robot
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.