October 12th

Venture capital funding for AI startups is projected to account for more than half of all VC money invested in 2025, confirming AI's dramatic dominance across the startup economy and contributing to ongoing concerns about market sustainability.

SoftBank committed $5.4 billion toward AI robotics, signaling confidence that the next major commercial phase of AI will involve physical, real-world automated systems and marking robotics as a key area for high-dollar funding.

Supabase secured a $5 billion valuation as of October 3, marking an astounding surge from its $2 billion valuation just four months prior and highlighting the speed at which developer-focused infrastructure startups are scaling.

JPMorgan Chase announced its long-term blueprint to become the first fully AI-powered megabank, indicating that traditional financial giants are betting their future infrastructure on comprehensive AI integration.

OpenAI launched AgentKit on October 6, delivering a cohesive development platform—including a visual Agent Builder and embeddable ChatKit UI—to accelerate the creation and deployment of specialized interactive AI agents, aiming to establish ChatGPT as the central marketplace for developers.

Zendesk introduced a new AI agent designed to autonomously solve 80% of customer support issues, setting an aggressive benchmark for task resolution rates and promising a massive shift in customer service automation.

Google Gemini launched the 2.5 Computer Use model API on October 7, demonstrating an AI capable of navigating a web browser to execute complex tasks on both mobile and web platforms, moving AI interaction beyond chat inputs.

Notion overhauled its underlying software architecture for its 3.0 version released in September, rebuilding its tech stack to support self-selecting, multi-step agentic AI that overcomes the limitations of older, rigid prompt-based workflows.

AMD finalized a multi-billion dollar AI chip-supply deal with OpenAI on October 6, agreeing to provide 6 gigawatts worth of its Instinct MI450 GPUs starting in the second half of 2026, causing AMD shares to surge over 34% and intensifying competition with NVIDIA.

AI21 Labs released the Jamba Reasoning 3B LLM, a small model with only 3 billion parameters that can efficiently handle a 250K context window on a standard laptop, suggesting a major breakthrough in making complex, high-context AI accessible on conventional consumer hardware.

Samsung AI debuted the Tiny Recursion Model (TRM), an open reasoning model that, for specific problem sets, outperforms models 10,000 times larger, demonstrating that specialized, smaller architectures can achieve superior performance without requiring vast scale.

Anthropic released Petri on October 8, an MIT-licensed, open-source framework that uses AI agents for automated auditing against 36 safety dimensions and 111 seed instructions, aiming to create a new industry standard for security testing.

Google initiated an AI vulnerability reward program on Monday, offering bug hunters up to $30,000 for identifying rogue actions, such as prompt injections that lead to unauthorized physical actions like unlocking a door via Google Home.

Deloitte publicly committed to refunding the Australian government for a policy report marred by AI hallucinations, demonstrating a clear financial and reputational cost for unchecked generative AI usage in major consulting projects.

The State of New Jersey established a State Artificial Intelligence Task Force on October 10, formalizing state-level efforts to create local AI policy and governance rules, following similar moves in states like Pennsylvania and North Dakota.

IBM announced productivity gains of 45% with its multi-model integrated development environment (IDE), "Project Bob," which provides LLMs with complete code repository context for software development, reinforcing the immediate value of AI assistance for coding professionals.

Job applicants are actively employing "prompt injection" techniques, sneaking hidden text into resumes to circumvent AI-powered job screeners, confirming that the automated hiring process has created a competitive, adversarial environment.

Creative professionals, exemplified by YouTuber MrBeast on October 6, voiced concerns that generative AI could severely threaten the livelihoods of creators, describing the current moment for the creative industry as "scary times."

Google introduced a new feature for Gemini to explain formula errors in Google Sheets, augmenting the daily tasks of white-collar workers by offering immediate, contextual troubleshooting for common productivity challenges.

OpenAI debuted the consumer Sora 2 AI video generator app, which now includes audio and allows for "self-insertion cameos," though the product's release was immediately complicated by unexpected copyright concerns.

Google deployed a new AI agent on October 6 focused on improving cybersecurity that automatically rewrites code to fix software vulnerabilities, marking a major step toward fully autonomous cyber defense mechanisms.

Google's virtual shopping features expanded, introducing an AI try-on feature that generates images showing what new shoes look like on a user's feet, leveraging computer vision to bridge the gap between digital retail and physical reality.

MIT CSAIL developed a generative AI tool on October 8 that produces realistic virtual environments, such as kitchens and living rooms, specifically to scale up training data needed for robot foundation models and accelerate robot skill acquisition.