Meta's Brain-to-Text Breakthrough, Consulting Pricing Shifts, and Cursor's Mobile Expansion
Meta advances non-invasive brain-to-text technology with Brain2Qwerty v2, while consulting firms begin shifting to outcome-based pricing models. Additionally, Cursor launches an iOS app for agent management, and Anthropic releases new data on AI usage patterns.
Podcast В· 4 min
Meta upgrades Brain2Qwerty to v2
Meta has introduced the second version of its Brain2Qwerty system, a non-invasive research pipeline designed to decode brain signals into text. Unlike the first version, which was limited to single characters, v2 is capable of decoding full sentences. The system was trained on 22,000 sentences of data collected from nine volunteers during 10 hours of brain scanning. In testing, the top-performing volunteer achieved 78% accuracy, with the system reaching an average word accuracy of 61%. This represents a significant improvement over previous non-invasive benchmarks, which typically struggled to exceed 8% accuracy. Meta has open-sourced the code and dataset for both versions. The development is notable because it narrows the performance gap between non-invasive systems and surgical brain-computer interfaces. By demonstrating that accuracy can be improved through data scaling, Meta suggests that non-invasive communication technology could eventually become a viable alternative for individuals who have lost the ability to speak.
Nvidia unveils CHORD project for robot manipulation
Nvidia has introduced the CHORD project, a new research initiative focused on transferring human dexterous manipulation skills to robot policies. The project utilizes contact-focused demonstrations to teach robots how to handle objects with human-like precision. This research addresses a critical bottleneck in robotics: the difficulty of programming robots to perform fine motor tasks. By leveraging AI to interpret human contact data, Nvidia aims to create more adaptable robot policies that can operate in unstructured environments, potentially accelerating the development of general-purpose robotic assistants.
Consulting firms pivot to outcome-based pricing
Major consulting firms are increasingly moving away from traditional hourly billing models in favor of outcome-based pricing. Reports indicate that firms like Deloitte and McKinsey are adapting to a market where AI significantly increases efficiency, making labor-based billing harder to justify to clients. Deloitte has reportedly projected that traditional labor-based consulting could shrink significantly by 2035. Firms are now testing fixed-fee structures and performance-based models where fees are tied to specific client outcomes. McKinsey noted that over 30% of its global fees are already linked to client results. This shift reflects a broader tension in the professional services sector. While outcome-based pricing aligns incentives with client success, it also introduces risks, such as the potential for firms to prioritize easily measurable metrics over long-term quality, or to use AI-driven efficiency as a justification for headcount reductions.
Arena reaches $100M ARR
Arena, a company specializing in real-world AI evaluation, announced that it has reached a $100 million annual revenue run rate. The milestone comes eight months after the company launched its core product. The rapid growth highlights the increasing demand for specialized evaluation tools as enterprises deploy more complex AI agents. Arena's platform focuses on assessing how AI models perform in real-world scenarios, a critical requirement for companies moving from experimental prototypes to production-grade agentic workflows.
Kling AI reports 100M users
Kling AI announced significant growth during its presentation at the Cannes Lions 2026 festival, reporting that it has surpassed 100 million global users and 50,000 enterprise clients. The company showcased new capabilities, including advanced motion control and 4K video production. The growth of Kling AI underscores the rapid adoption of generative video tools in commercial production. By integrating AI into hybrid workflows alongside traditional CGI and human filmmaking, the company is positioning itself as a key player in the evolving creative software market.
Anthropic publishes Claude usage report
Anthropic has released a new report in its Economic Index, providing granular, hourly data on how users interact with Claude. The study, which surveyed 9,700 users, reveals distinct patterns in how AI is integrated into daily routines. Data shows that news-related queries peak in the morning, while recipe and cooking-related prompts surge during dinner hours. Additionally, the report highlights a shift in personal usage, with Claude chats moving from roughly one-third of activity during the workweek to nearly half on weekends. Users who delegate more tasks to AI reported higher expectations for future AI capabilities and expressed more positive sentiments regarding their career stability.
Cursor launches iOS app for coding agents
Cursor has released a new iOS application designed to allow developers to manage coding agents from their mobile devices. The app enables users to launch agents, monitor progress, and review pull requests while away from their primary workstations. The app supports both cloud-based execution and local machine integration, allowing developers to switch between environments seamlessly. With features like push notifications for agent status updates, the tool aims to shift development workflows from desk-bound tasks to mobile supervision, reflecting the trend of developers acting more as supervisors of AI-generated code.
Artificial Analysis releases AA-Briefcase benchmark
Artificial Analysis has launched AA-Briefcase, a private benchmark designed to evaluate the performance of AI models in long-horizon agentic knowledge work. The benchmark tests capabilities across spreadsheets, presentations, and memos. As the industry shifts toward agentic workflows that require multi-step reasoning and document handling, standard benchmarks are becoming less effective. AA-Briefcase aims to provide a more realistic assessment of how models handle complex, multi-document tasks, offering developers a clearer view of model reliability in professional settings.
Tidal implements AI music policy
Music streaming service Tidal has announced a new policy regarding AI-generated content. The platform will now label music that is substantially generated by AI and will restrict monetization for such tracks. This move aligns Tidal with other platforms attempting to address the influx of AI-generated audio. By distinguishing between human-created and AI-generated content, the company aims to protect artist royalties and maintain the integrity of its catalog, reflecting the ongoing industry-wide debate over copyright and compensation in the age of generative AI.