Search “Google”

Google Research proposes a sleep framework for continual learning of LLMs

Google Research proposes a sleep framework for continual learning in LLMs, introducing periodic sleep phases into the model to consolidate short-term context into long-term parameters, thereby mitigating catastrophic forgetting and the static knowledge cutoff. The framework consists of two stages: memory consolidation and dreaming. The former uses MoE parameter expansion and knowledge seeding based on GKD and reinforcement learning imitation, while the latter achieves self-improvement through random expert activation, screening of synthetic samples, and RL refinement. On BABILong, this method scales nearly perfectly within 10M tokens, and achieves accuracies of 79.2%, 69.0%, and 46.1% on AIME 24/25 and HMMT 25, respectively. There are currently no follow-up developments.

View story:Google Research proposes a sleep framework to help LLMs retain knowledge in continual learning

Google Developers Blog publishes HappyDelivery multi-agent prototype

On October 27, 2026, Prateek Khushalani published the HappyDelivery multi-agent conversational commerce prototype on the Google Developers Blog. The prototype is built on LangGraph, Gemini, FastAPI, and Google Cloud, implementing the previously proposed three-layer decoupled architecture of Ingress, Memory, and Cognition. The prototype abstracts channels such as WhatsApp, RCS, Apple Messages for Business, and SMS through platform-specific signature verification and a unified message schema; the Ingress layer was measured returning HTTP 200 in approximately 295–302 ms and forwarding to Pub/Sub. The article also provides code examples for the state machine, deterministic financial tools, and an automated validation suite. There are currently no subsequent developments.

View story:Prateek Khushalani publishes the HappyDelivery multi-agent prototype

Google launches replaceable reasoning format A2UI Express for A2UI

On October 27, 2026, Google launched a replaceable reasoning format for A2UI, with A2UI Express becoming the new default format. According to the Google Developers Blog, Express can reduce latency by up to 3x and reduce the tokens needed to generate comparable UI by up to 3x; when used with decision models such as Jev for layout decisions, the latency reduction can reach 80x. The format has landed in the Python agent SDK (a2ui-agent-sdk 0.9.0), the TypeScript SDK (@a2ui/agent), and the Dart SDK (a2ui_agent 0.0.1-wip004), and developers can switch formats with a single parameter when creating a generator. There are currently no reports of further developments.

View story:Google launches A2UI Express, a replaceable reasoning format for A2UI

Google demonstrates Gemini 4 Argon's ability to handle complex tasks

On October 27, 2026, Google released its latest model, Gemini 4 Argon, and demonstrated its ability to handle complex real-world tasks. Google says the model can reason, code, and coordinate tools, and the company gave 8 examples: first, writing 5,000 lines of mathematical code from scratch with zero source material, rendering a 3D sunlit library; second, processing more than 80,000 raw files from the NASA Mars mission in Antigravity and generating executive briefing slides. There is currently no information on further developments.

View story:Google releases Gemini 4 Argon and demonstrates complex task capabilities

Google Workspace supports native Markdown editing in Docs and Drive

Google Workspace announced that Google Docs and Drive now natively support viewing, editing, and collaborating on Markdown (.md) files, with no need for users to manually convert formats. According to information published on the Google Developers Blog on October 26, 2026, rendered tables and hyperlinks can be previewed directly in Drive, and a .md label will be displayed in the document header. The feature also supports real-time collaboration, line-level comments, and version history, making it convenient for team members and AI agents to co-edit the same file. No further progress has been disclosed at this time.

Google Research and Stanford launch AI Quest "Polar Peaks"

Google Research and the Stanford Learning Accelerator launch "Polar Peaks," the third lesson in the AI Quests series, aimed at students aged 11 to 14. In the lesson, students play the role of researchers, scanning and mapping brain neurons, practicing data cleaning and model testing, and understanding that AI requires human oversight. The lesson was inspired by more than a decade of Google's connectomics research. AI Quests has completed over 400,000 missions in total, has been localized into 10 languages, and a companion teacher training module will launch next month.

Sierra launches on Google Cloud Marketplace

Sierra launched on Google Cloud Marketplace on October 27, 2026. According to a Sierra blog post, enterprises can purchase using their existing Google Cloud commitment credits, and eligible Marketplace purchases can count toward committed spend; after signing through an existing Google Cloud relationship, billing is consolidated in the Google Cloud Console and supports Private Marketplace governance. There is currently no further development information.

Sierra launches on Google Cloud Marketplace

Sierra launched on Google Cloud Marketplace on October 27, 2026. According to a Sierra blog post, enterprises can purchase using their existing Google Cloud commitment credits, and eligible Marketplace purchases can count toward committed spend; after signing through an existing Google Cloud relationship, billing is consolidated in the Google Cloud Console and supports Private Marketplace governance. There is currently no further development information.

View story:Sierra launches on Google Cloud Marketplace

Google launches A2UI Express, a replaceable reasoning format for A2UI

On October 27, 2026, Google launched a replaceable reasoning format for A2UI, with A2UI Express becoming the new default format. According to the Google Developers Blog, Express can reduce latency by up to 3x and reduce the tokens needed to generate comparable UI by up to 3x; when used with decision models such as Jev for layout decisions, the latency reduction can reach 80x. The format has landed in the Python agent SDK (a2ui-agent-sdk 0.9.0), the TypeScript SDK (@a2ui/agent), and the Dart SDK (a2ui_agent 0.0.1-wip004), and developers can switch formats with a single parameter when creating a generator. There are currently no reports of further developments.

Google releases Gemini 4 Argon and demonstrates complex task capabilities

On October 27, 2026, Google released its latest model, Gemini 4 Argon, and demonstrated its ability to handle complex real-world tasks. Google says the model can reason, code, and coordinate tools, and the company gave 8 examples: first, writing 5,000 lines of mathematical code from scratch with zero source material, rendering a 3D sunlit library; second, processing more than 80,000 raw files from the NASA Mars mission in Antigravity and generating executive briefing slides. There is currently no information on further developments.

Google podcast discusses how AI Overviews and AI Mode are changing search habits

On October 28, 2026, Google Ads podcast Ads Decoded released an episode hosted by Google Ads product liaison Ginny Marvin, in conversation with Shivani Mohan, Vice President of Google Search Data Science and UXR, discussing how AI Overviews and AI Mode are changing the search habits of billions of users. The episode focuses on Shivani Mohan's team's research into the modern search experience and what the long-form query trend means for businesses. The episode has currently been published on Google's official blog, with no further developments at this time.

View story:Google podcast discusses how AI Overviews and AI Mode are changing search habits

LiteRT in Google Play services gets GPU inference upgrade

On October 28, 2026, the Google Developers Blog announced that LiteRT in Google Play services has received its biggest upgrade yet: GPU inference based on ML Drift is about 1.4 times faster than the old GPU delegate and is provided through the new CompiledModel API. After switching to this runtime, app size shrinks by about 8 MB, because the runtime is distributed uniformly with Play services and is no longer packaged into the APK. Currently, more than 100,000 Android apps already bundle LiteRT or TensorFlow Lite. The solution supports Android 7 and above, Wear OS 4+, and Android Automotive OS (Android 11+), while NPU support is still under development.

View story:Google upgrades LiteRT GPU inference in Play services

Google upgrades LiteRT GPU inference in Play services

On October 28, 2026, the Google Developers Blog announced that LiteRT in Google Play services has received its biggest upgrade yet: GPU inference based on ML Drift is about 1.4 times faster than the old GPU delegate and is provided through the new CompiledModel API. After switching to this runtime, app size shrinks by about 8 MB, because the runtime is distributed uniformly with Play services and is no longer packaged into the APK. Currently, more than 100,000 Android apps already bundle LiteRT or TensorFlow Lite. The solution supports Android 7 and above, Wear OS 4+, and Android Automotive OS (Android 11+), while NPU support is still under development.

Google podcast discusses how AI Overviews and AI Mode are changing search habits

On October 28, 2026, Google Ads podcast Ads Decoded released an episode hosted by Google Ads product liaison Ginny Marvin, in conversation with Shivani Mohan, Vice President of Google Search Data Science and UXR, discussing how AI Overviews and AI Mode are changing the search habits of billions of users. The episode focuses on Shivani Mohan's team's research into the modern search experience and what the long-form query trend means for businesses. The episode has currently been published on Google's official blog, with no further developments at this time.

Google Docs and Drive now natively support Markdown file preview and editing

Google Workspace announced that Google Docs and Drive now natively support viewing, editing, and collaborating on Markdown (.md) files, with no need for users to manually convert formats. According to information published on the Google Developers Blog on October 26, 2026, rendered tables and hyperlinks can be previewed directly in Drive, and a .md label will be displayed in the document header. The feature also supports real-time collaboration, line-level comments, and version history, making it convenient for team members and AI agents to co-edit the same file. No further progress has been disclosed at this time.

View story:Google Workspace supports native Markdown editing in Docs and Drive

Google Research and Stanford launch AI Quest "Polar Peaks"

Google Research and the Stanford Learning Accelerator launch "Polar Peaks," the third lesson in the AI Quests series, aimed at students aged 11 to 14. In the lesson, students play the role of researchers, scanning and mapping brain neurons, practicing data cleaning and model testing, and understanding that AI requires human oversight. The lesson was inspired by more than a decade of Google's connectomics research. AI Quests has completed over 400,000 missions in total, has been localized into 10 languages, and a companion teacher training module will launch next month.

View story:Google Research and Stanford launch AI Quest "Polar Peaks"

Google says partnership with Ignite Energy Access cut 160,000 tons of carbon emissions in the first half of the year

On October 26, 2026, Google stated that its partnership with Ignite Energy Access achieved emissions reductions of more than 160 kT CO2e in the first half of 2026. According to information published on the Google Blog the same day, the two parties have delivered 3.7 million solar devices in sub-Saharan Africa since 2023. The two parties use satellite imagery, the Open Buildings dataset, and intelligent routing algorithms to locate remote households without addresses, and use a computer vision model fine-tuned on Gemini for Science to identify individual rooftop solar panels from space. The partnership has now expanded to modeling microgrids for rural households to support clean electric cooking solutions. There are currently no follow-up independent reports or third-party verification.

Google says partnership with Ignite Energy Access cut 160,000 tons of carbon in the first half of the year

On October 26, 2026, Google stated that its partnership with Ignite Energy Access achieved emissions reductions of more than 160 kT CO2e in the first half of 2026. According to information published on the Google Blog the same day, the two parties have delivered 3.7 million solar devices in sub-Saharan Africa since 2023. The two parties use satellite imagery, the Open Buildings dataset, and intelligent routing algorithms to locate remote households without addresses, and use a computer vision model fine-tuned on Gemini for Science to identify individual rooftop solar panels from space. The partnership has now expanded to modeling microgrids for rural households to support clean electric cooking solutions. There are currently no follow-up independent reports or third-party verification.

View story:Google says partnership with Ignite Energy Access cut 160,000 tons of carbon emissions in the first half of the year

Google.org is investing US$2.5 million to support the World Bank Group in building the Open Network for Agriculture (ONA)

Google.org is providing US$2.5 million in support to the World Bank Group to scale digital public infrastructure for agriculture. Following a successful pilot in India, the World Bank Group plans to build the AI-driven Open Network for Agriculture (ONA) in multiple countries across Latin America, Southeast Asia, and Sub-Saharan Africa, enabling farmers to access digital farmer IDs, credit, subsidy payments, sales markets, and voice-based agronomic advice in local languages through a single platform. The goal of this support is to help 1 million farmers access key services that improve their livelihoods by the end of 2027.

Google.org supports World Bank Group in building the Open Network for Agriculture (ONA)

Google.org is providing US$2.5 million in support to the World Bank Group to scale digital public infrastructure for agriculture. Following a successful pilot in India, the World Bank Group plans to build the AI-driven Open Network for Agriculture (ONA) in multiple countries across Latin America, Southeast Asia, and Sub-Saharan Africa, enabling farmers to access digital farmer IDs, credit, subsidy payments, sales markets, and voice-based agronomic advice in local languages through a single platform. The goal of this support is to help 1 million farmers access key services that improve their livelihoods by the end of 2027.

View story:Google.org is investing US$2.5 million to support the World Bank Group in building the Open Network for Agriculture (ONA)