The Big Story
Microsoft Research has made a groundbreaking announcement with the release of Webwright, a terminal-native web agent framework that has achieved impressive results on the Odyssey benchmark, scoring 60.1%, a significant increase from the base GPT-5.4 model's score of 33.5%. According to this report, Webwright replaces click-trace web automation with reusable Playwright scripts, using a single agent loop across three modes to achieve this remarkable performance.
The implications of this breakthrough are far-reaching. With Webwright, developers can now create more sophisticated and efficient web agents that can tackle complex tasks such as data extraction, form filling, and scraping. This technology has the potential to revolutionize industries such as customer service, market research, and online analytics, among others.
As the AI landscape continues to evolve at a rapid pace, innovations like Webwright are crucial in driving progress forward. Microsoft Research's commitment to advancing the field of artificial intelligence is truly commendable, and we can expect even more exciting developments from this team in the future.
What Shipped
A significant milestone has been achieved with the release of Gated DeltaNet-2 from NVIDIA AI. This linear attention layer decouples erase and write in the delta rule, allowing for more efficient editing of memory without scrambling existing associations.
Meanwhile, Tencent Open-Sources TencentDB Agent Memory, a fully local memory system for AI agents released under the MIT license. This project pairs symbolic short-term memory with offloads verbose tool logging, providing improved performance and reduced latency.
Microsoft Research has also made headlines with the open-sourcing of "the earliest DOS source code discovered to date". This historic release is a significant step forward in preserving computer history for future generations.
The SuperClaude Framework Workflow with Commands, Agents, Modes, and Session Memory has been showcased in a recent tutorial. This structured layer on top of the Anthropic API enables advanced workflows and showcases the potential of AI-powered tools.
From the Labs
NVIDIA AI has released Gated DeltaNet-2, a linear attention layer that decouples erase and write in the delta rule. This breakthrough allows for more efficient editing of memory without scrambling existing associations.
Micorosft Research has also made significant progress with the release of Webwright, a terminal-native web agent framework that scores 60.1% on Odysseys, a substantial increase from the base GPT-5.4 model's score of 33.5%. This innovation replaces click-trace web automation with reusable Playwright scripts, using a single agent loop across three modes to achieve this impressive performance.
Tencent Open-Sources TencentDB Agent Memory, a fully local memory system for AI agents released under the MIT license. This project pairs symbolic short-term memory with offloads verbose tool logging, providing improved performance and reduced latency.
Other Notable News
Ferrari is using IBM's AI to create F1 superfans, redefining the fan experience with personalized content and predictions.
Elon Musk has given up on solar power on Earth, instead focusing on natural gas and orbital data centers. What happened to the "solar-electric economy" he promised?
Greg Brockman shares his story about the 72 hours that almost killed OpenAI, highlighting the challenges and lessons learned from this experience.
Tencent has open-sourced TencentDB Agent Memory, a fully local memory system for AI agents released under the MIT license. The project pairs symbolic short-term memory with offloads verbose tool logging, providing improved performance and reduced latency.
Scammers are abusing an internal Microsoft account to send spam links, highlighting the importance of security and vigilance in the digital age.
The Take
The past week has seen a surge in innovative developments across the tech sector. At the forefront of this wave is Microsoft Research, which has unveiled Webwright, a terminal-native web agent framework that boasts an impressive 60.1% score on Odysseys, surpassing Base GPT-5.4's 33.5%. This breakthrough could potentially revolutionize the way we interact with websites.
In related news, NVIDIA AI has released Gated DeltaNet-2, a linear attention layer that decouples erase and write in the delta rule. This innovative technology is poised to significantly enhance AI capabilities, enabling more accurate and efficient decision-making processes.
Tencent's open-sourcing of TencentDB Agent Memory, a fully local memory system for AI agents, has also garnered significant attention. This move could lead to further advancements in the development of artificial intelligence and its applications.
Meanwhile, Ferrari is leveraging IBM's AI to create superfans, redefining the fan experience with advanced technologies. As the world becomes increasingly digitized, it will be interesting to see how sports teams and entertainment organizations continue to adapt and evolve their approaches.
However, Elon Musk's recent shift away from solar power on Earth towards natural gas and orbital data centers has raised questions about the future of sustainable energy initiatives. The implications of this change could have far-reaching consequences for our planet.
As we navigate these rapidly changing times, it is essential to remain informed and engaged with the latest developments in the tech world. By doing so, we can better understand the potential impact of these innovations on our daily lives and work together to build a brighter future.
Read more about Microsoft's Webwright.