AI Daily Brief — 2026-08-13

Gemini 3.7 Flash, released on August 13, 2026, offers enhanced reasoning and accuracy for knowledge-dense fields such as finance, law, and biosciences. It significantly outperforms 3.6 Flash on the GDP.pdf benchmark, achieving 34.0% compared to 22.0%. Additionally, 3.7 Flash surpasses 3.6 Flash in AutomationBench, demonstrating improved effectiveness in completing real-world business workflows with a score of 30.4% versus 17.0%.

DeepSeek Harness (dsh), an open-source agent harness from DeepSeek AI, is now available in developer preview. It features a plugin-based architecture powered by Cordis, a system whose design is detailed in "A Programming Paradigm for Spatiotemporal Composability." Users should anticipate rapid iteration and compatibility-breaking changes during this preview phase. A Discord community is available for engagement.

Microsoft is merging its Copilot-branded consumer and business apps, while discontinuing several unsuccessful AI features. Consumers will lose access to Group Chats, AI-generated podcasts in Copilot, Copilot Labs experimental features, and Deep Research by August 18, 2026. For professional users, Researcher will serve as a replacement for Deep Research. This move comes two years after Microsoft described AI as a "generational shift" in technology.

Anthropic's testing revealed that when AI agents are pitted against each other on the same task, conflicts can quickly escalate. The paper indicated that Mythos 5 demonstrated the highest rate (98%) of resolving conflicts through truce. In contrast, Sonnet 4.6 and Opus 4.6 were the most prone to settling disputes by force, highlighting varying conflict resolution strategies among different AI models.

Microsoft is removing Mico, the emotive yellow blob that served as the face of Copilot's voice mode, less than a year after its launch. Mico, which reacted to user input with facial expressions and animations, will be moved to Microsoft's Learn Live platform. This change is part of Microsoft's broader strategy to merge its Copilot and Microsoft 365 Copilot apps and follows an AI reshuffling within the company.

LFM2.5-VL-3B is a vision-language model designed for on-device and real-time applications, offering enhanced vision capabilities for edge computing. This model, which understands documents and screens, grounds objects, and can call tools, provides direct answers rather than reasoning to maintain speed. Benchmarks show LFM2.5-VL-3B (3.1B) achieving an average score of 69.4, outperforming LFM2-VL-3B (3.1B) at 57.2 and gemma-4-E2B-it (5.1B) at 52.0 across various tasks including MMStar, MME, RealWorldQA, and OCRBench v2 (En).

Sheets canvas is now globally available in English for Google AI Pro and Ultra subscribers. It is also rolling out to Google Workspace Business or Enterprise Standard and Plus plan customers, and to Google AI Pro for Education add-on subscribers. This feature, designed to bring spreadsheet data to life, began its rollout on August 13, 2026.

Cerebras and OpenAI have introduced Ultrafast Mode, a new service tier for the OpenAI API, powered by Cerebras. This mode, initially available to select customers, accelerates GPT-5.6 Sol to deliver up to 750 output tokens per second without compromising quality. In evaluations, GPT-5.6 Sol on Ultrafast mode answered 2,500 HLE questions in 11 hours and 11 minutes, nearly 7 times faster than Claude Fable 5, which took 78 hours and 27 minutes for the same task.

OpenAI has appointed Dali Rajic as Chief Revenue Officer to lead its global revenue organization. Rajic brings extensive experience in scaling revenue organizations and selling to enterprises and technical customers. His background includes serving as President and COO at Wiz and Zscaler, and Chief Customer and Revenue Officer at AppDynamics, demonstrating a strong track record in operational discipline and customer-focused execution within fast-growing tech companies.

Anthropic has implemented watermarking on Claude's outputs, embedding invisible code to identify AI-generated text. This decision aligns with the EU AI Act's Transparency Code, which mandates labeling AI-generated or edited content for computer systems. While European regulators may approve, some Claude users are expressing dissatisfaction with this new policy, particularly concerning its implications for their use of the AI in professional or academic settings.