October 20, 2025 News
Former OpenAI and Google Brain Researchers Launch AI-Powered Materials Science Startup with $300M
Periodic Labs, founded by OpenAI's Liam Fedus and Google Brain's Ekin Dogus Cubuk, emerged from stealth with a $300 million seed round to automate materials science discovery using AI. The startup combines robotic synthesis, ML simulations, and LLM reasoning to discover new compounds, particularly superconductors, in a fully automated lab environment. The team has recruited over two dozen top AI and scientific researchers and is already conducting experiments, though robotic systems are still being trained.
Skynet Chance (+0.01%): The closed-loop system of AI hypothesis generation, robotic execution, and automated analysis represents increased AI autonomy in physical experimentation, though focused on beneficial scientific discovery. The risk remains low as the system operates in controlled lab environments with clear objectives.
Skynet Date (+0 days): The integration of AI reasoning with physical robotic systems and real-world experimentation modestly accelerates the timeline toward more autonomous AI systems capable of independent action. However, the narrow domain focus and controlled environment limit broader implications for AI autonomy.
AGI Progress (+0.02%): This represents meaningful progress in AI's ability to conduct autonomous scientific reasoning, hypothesis testing, and physical interaction with the real world through robotic systems. The closed-loop learning from experimental failures and successes demonstrates enhanced real-world grounding that addresses a key AGI capability gap.
AGI Date (+0 days): The substantial funding, talent acquisition including key OpenAI researchers, and focus on generating novel real-world training data accelerates AGI development by addressing the critical bottleneck of grounded, experimental data. The system's ability to learn from physical experiments provides a new pathway for AI advancement beyond purely digital training.
Anthropic Expands Claude Code AI Coding Assistant to Web Platform
Anthropic launched a web-based version of Claude Code, its AI coding assistant that allows developers to create and manage AI coding agents from their browser. The tool, available to Pro and Max subscribers, has grown 10x in users since May and now generates over $500 million in annualized revenue. Anthropic claims 90% of Claude Code itself is written by AI, reflecting the shift toward agentic AI coding tools that work autonomously rather than as simple autocomplete.
Skynet Chance (+0.04%): The widespread deployment of autonomous AI agents that can write complex code with minimal human oversight increases the surface area for potential misalignment and reduces human understanding of software systems. The fact that 90% of the product itself is AI-written demonstrates recursive self-improvement capabilities and reduced human control in critical software development.
Skynet Date (-1 days): The rapid commercial success and 10x user growth accelerates the deployment of autonomous AI agents in critical software development roles, potentially hastening timeline concerns. However, these remain narrowly-scoped coding assistants rather than general agents, moderating the acceleration effect.
AGI Progress (+0.03%): The shift from autocomplete to autonomous agentic coding represents meaningful progress toward AI systems that can independently complete complex, multi-step tasks in specialized domains. The ability to write 90% of its own codebase demonstrates approaching human-level performance in software engineering tasks, a key capability for AGI.
AGI Date (-1 days): The commercial viability ($500M+ revenue) and rapid adoption of agentic AI coding tools accelerates investment and development in autonomous AI systems. The demonstrated capability of AI writing most of its own code could create positive feedback loops that speed AGI development timelines.