February 25, 2026 News
Anthropic Acquires Computer-Use AI Startup Vercept in Strategic Talent Play
Anthropic has acquired Vercept, an AI startup that developed tools for complex agentic tasks including a cloud-based computer-use agent capable of operating remote Macbooks. The acquisition brings several co-founders and researchers to Anthropic, though one co-founder had already been poached by Meta for $250 million, and Vercept's product will be shut down on March 25th. The deal follows Anthropic's December acquisition of coding agent engine Bun as part of its strategy to scale Claude Code capabilities.
Skynet Chance (+0.01%): The consolidation of computer-use agent capabilities into Anthropic's Claude system slightly increases autonomous AI capabilities that could operate computer systems, though Anthropic has demonstrated safety-conscious approaches. The competitive talent acquisition dynamics suggest rapid capability advancement across multiple labs.
Skynet Date (+0 days): Anthropic's aggressive acquisition strategy for agentic capabilities and the high-stakes talent competition (evidenced by Meta's $250M offer) indicates accelerated development of autonomous AI systems. The consolidation of Vercept's computer-use technology into Claude could speed deployment of agents with broader system access.
AGI Progress (+0.02%): Computer-use agents that can autonomously operate full computing environments represent meaningful progress toward AGI-relevant capabilities, demonstrating improved perception, planning, and action in complex digital environments. The acquisition strengthens Anthropic's position in building more generally capable AI systems.
AGI Date (+0 days): The rapid consolidation of specialized agentic capabilities into major AI labs, combined with intense talent competition at astronomical salaries ($250M), signals aggressive acceleration in the race toward more capable autonomous systems. Anthropic's strategic acquisitions (Bun in December, Vercept now) demonstrate a focused push to rapidly scale agent capabilities.
Nvidia Reports Record $68B Quarterly Revenue Driven by Exponential AI Compute Demand
Nvidia reported record quarterly revenue of $68 billion, up 73% year-over-year, with $62 billion coming from its data center business driven by exponential demand for AI compute. CEO Jensen Huang emphasized that demand for tokens has gone "completely exponential" and positioned compute investment as directly tied to revenue generation, while announcing the company is close to finalizing a reported $30 billion investment partnership with OpenAI. The company noted competitive pressure from Chinese AI chip makers following recent IPOs.
Skynet Chance (+0.04%): Exponential scaling of AI compute infrastructure and massive capital deployment accelerates the development of increasingly powerful AI systems without corresponding mention of safety measures or alignment progress. The focus on token generation economics and profit motive over control mechanisms modestly increases uncontrolled AI risk.
Skynet Date (-1 days): The exponential growth in compute availability and aggressive capex spending by tech companies significantly accelerates the pace at which powerful AI systems can be trained and deployed. Nvidia's characterization of demand as "completely exponential" and compute-as-revenue model suggests accelerating timeline for advanced AI capabilities.
AGI Progress (+0.03%): Record compute infrastructure growth and exponential scaling of GPU deployment directly enables training of larger, more capable models approaching AGI-level performance. The $215 billion annual revenue and massive data center expansion represents substantial progress in the hardware foundation required for AGI development.
AGI Date (-1 days): The exponential increase in available compute, sustained massive investments (including pending $30B OpenAI partnership), and Nvidia's assertion that profitable token generation is already happening all indicate significant acceleration toward AGI timelines. The characterization of reaching an "inflection point" suggests AGI development is proceeding faster than previously expected.
Google Integrates Intrinsic Robotics Platform to Advance Physical AI Capabilities
Alphabet is moving its robotics software subsidiary Intrinsic under Google's umbrella to accelerate physical AI development. Intrinsic, which builds AI models and software for industrial robots, will work closely with Google DeepMind and leverage Gemini AI models while remaining a distinct entity. The move aims to make robotics more accessible to manufacturers and advance factory automation, particularly through Intrinsic's partnership with Foxconn.
Skynet Chance (+0.04%): Integrating advanced AI models (Gemini) with physical robotics systems and factory automation increases the deployment of AI in physical domains with real-world consequences, creating more potential pathways for unintended autonomous behavior. However, the focus on industrial applications with human oversight provides some containment.
Skynet Date (-1 days): Consolidating robotics capabilities under Google with direct access to frontier AI models (Gemini) and DeepMind resources accelerates the development and deployment of increasingly capable physical AI systems. The Foxconn partnership for full factory automation suggests rapid real-world scaling.
AGI Progress (+0.03%): This represents significant progress in embodied AI, a critical component of AGI, by combining advanced language/reasoning models (Gemini) with physical manipulation capabilities and real-world learning environments. The integration of perception, planning, and action in industrial settings advances toward more general-purpose intelligent systems.
AGI Date (-1 days): Bringing together Google's substantial AI infrastructure, DeepMind's research capabilities, and Intrinsic's robotics platform creates powerful synergies that should accelerate progress on embodied intelligence. The focus on making robotics accessible to non-experts also broadens the developer base working on these problems.
States Across US Propose Data Center Moratoriums Amid Growing Public Opposition to AI Infrastructure
Public opposition to AI data center construction is intensifying across the United States, with several states and municipalities proposing or passing temporary moratoriums on new facilities. New York has introduced a three-year statewide construction ban while communities study environmental and economic impacts, joining local bans in New Orleans, Madison, and other cities. The backlash is driven by concerns over rising energy costs, environmental pollution, and strain on local resources, even as tech companies plan to spend $650 billion on data center infrastructure.
Skynet Chance (-0.03%): Public and regulatory resistance to AI infrastructure buildout may slow the concentration of compute power and impose environmental accountability measures, slightly reducing risks from unchecked AI capability scaling. However, the impact on control mechanisms or alignment research is minimal.
Skynet Date (+1 days): Moratoriums and regulatory resistance could delay the rapid infrastructure expansion needed for training increasingly powerful AI systems, potentially slowing the timeline toward scenarios involving uncontrollable AI. The magnitude is moderate as companies are finding workarounds and the policies remain localized.
AGI Progress (-0.03%): Regulatory barriers and public opposition to data center construction directly constrain the compute infrastructure necessary for scaling AI models toward AGI-level capabilities. This represents a modest but tangible impediment to the compute scaling pathway that many organizations are pursuing.
AGI Date (+1 days): Construction moratoriums and potential elimination of tax incentives could materially slow the pace of compute infrastructure deployment, delaying the timeline for achieving AGI by restricting the rapid scaling of training capacity. The $650 billion planned expenditure faces meaningful regulatory headwinds that could extend development timelines by months or years.
Google Expands Gemini AI with Multi-Step Task Automation on Android Devices
Google announced updates to its Gemini AI features on Android, including beta multi-step task automation for ordering food and rideshares on select devices like Pixel 10 and Galaxy S26. The update also expands scam detection for calls and texts, and enhances Circle to Search to identify multiple items on screen simultaneously. The automation feature includes safety protections like explicit user commands, real-time monitoring, and limited app access within a secure virtual window.
Skynet Chance (+0.01%): The automation operates in a controlled sandbox with explicit user commands and real-time oversight, demonstrating responsible deployment practices that slightly mitigate loss-of-control risks. However, expanding AI agent capabilities into real-world task execution does incrementally increase the surface area for potential misuse or unintended consequences.
Skynet Date (+0 days): The release of practical AI agents that can execute multi-step real-world tasks represents incremental progress toward more autonomous AI systems. However, the limited scope (food delivery, rideshares) and extensive safety guardrails suggest a cautious, measured deployment that only slightly accelerates the timeline.
AGI Progress (+0.02%): Multi-step task automation with real-world application integration demonstrates meaningful progress in agentic AI capabilities, including planning, tool use, and sequential reasoning. This represents a concrete step toward more general-purpose AI systems that can handle diverse tasks autonomously.
AGI Date (+0 days): The commercial deployment of AI agents capable of multi-step task execution across multiple applications indicates major tech companies are successfully translating research into practical agentic systems. This accelerates the pace toward more capable and general AI systems, though the current limitations keep the acceleration modest.
MatX Secures $500M Series B to Challenge Nvidia with Next-Generation AI Training Chips
MatX, a chip startup founded by former Google TPU engineers, raised $500 million in Series B funding led by Jane Street and Leopold Aschenbrenner's Situational Awareness fund. The company aims to develop processors that are 10 times more efficient than Nvidia's GPUs for training large language models, with chip production planned through TSMC and shipments expected in 2027.
Skynet Chance (+0.01%): Increased competition in AI chip development could lead to more distributed access to powerful AI training infrastructure, slightly reducing concentration of control. However, the focus on 10x efficiency gains for LLM training also enables more actors to develop potentially uncontrollable advanced systems.
Skynet Date (-1 days): The planned 10x improvement in training efficiency and increased competition in specialized AI chips would accelerate the development of more powerful AI systems. However, chips won't ship until 2027, somewhat limiting near-term acceleration effects.
AGI Progress (+0.02%): A 10x improvement in training efficiency for large language models represents significant progress in overcoming compute bottlenecks, a key constraint in AGI development. The involvement of former Google TPU engineers and substantial funding suggests credible technical advancement toward more capable AI systems.
AGI Date (-1 days): If MatX delivers on its 10x efficiency promise by 2027, it would substantially accelerate AGI timelines by making advanced model training more accessible and cost-effective. The significant funding and experienced team increase the likelihood of successful execution, compressing development cycles.