Back to MergeSort

AI News Digest - February 15, 2026

15 stories · February 15, 2026

Top Stories

Product

OpenAI Tests Ads in ChatGPT, Launches GPT-5.3-Codex 'Fast Mode'

OpenAI has begun testing advertisements in ChatGPT for free and Go subscription tiers in the U.S., marking a step towards monetizing the chatbot. They also launched GPT-5.3-Codex-Spark, a 'fast mode' offering 15x speed increase using a smaller model on Cerebras chips, and signaled a shift towards 'builder tooling' as the primary consumer interface.

Sources: Hacker News, Latent Space, Michael Parekh AI, The Neuron

More Stories

Funding

Anthropic Raises $30 Billion at $380 Billion Valuation, Plans Massive AI Data Centers

Anthropic secured $30 billion in Series G funding, valuing the company at $380 billion, potentially leading to an IPO. The company is planning to invest hundreds of billions to build multi-gigawatt AI data centers to scale its compute capabilities and aims for 10 gigawatts of capacity.

Sources: via Michael Parekh AI

Research

AI Agents Aletheia and Moshi Demonstrate Advanced Capabilities in Mathematics and Conversation

Researchers, including those from DeepMind, introduced Aletheia, an AI agent that autonomously generates and verifies mathematics research, authoring a complete paper. Kyutai unveiled Moshi, an open-source, full-duplex conversational AI model capable of real-time interaction with ultra-low latency.

Sources: via Hacker News

News

AI Industry Faces Massive Compute and Power Spending

The AI industry is projected to spend over $700 billion on compute and power infrastructure in 2026, driven by major players like Anthropic. Major tech companies are exploring diverse financing options, including bond sales, to fund these substantial AI capital expenditures.

Sources: via Michael Parekh AI

Product

Zhipu AI Launches GLM-5, an Opus-Class Open-Weight Model

Zhipu AI has released GLM-5, an "Opus-class" open-weight model scaling to 744B parameters with 40B active. The model increases pre-training data to 28.5T tokens and integrates DeepSeek Sparse Attention for efficiency.

Sources: via Latent Space

Product

Hugging Face Enables CUDA Kernel Generation and Evaluation with OpenEnv

Hugging Face has developed an agent skill enabling AI agents like Claude and Codex to automatically write and optimize production-grade CUDA kernels. Meta and Hugging Face's OpenEnv framework provides a standardized method for evaluating AI agents against real-world systems and APIs.

Sources: via Hugging Face Blog

Funding

Gradium Raises $70M to Commercialize Audio AI from Kyutai

Gradium, spun out of the open audio lab Kyutai, has raised $70 million to bridge the gap between cutting-edge audio AI research and production-grade products. This funding aims to bring Kyutai's foundational audio AI technologies to market.

Sources: via Hacker News

News

Anthropic's Enterprise Subscriptions Surge, Competing with OpenAI

Anthropic's share of U.S. companies paying for its AI tools increased to 20% in January, while OpenAI's share slightly declined, indicating growing competition. A report indicated that Anthropic commands 40% of the enterprise spending on Large Language Models.

Sources: via Michael Parekh AI

Product

Google DeepMind Launches Gemini 3 Deep Think V2

Google DeepMind has rolled out Gemini 3 Deep Think V2, an upgraded reasoning mode for Google AI Ultra subscribers and select enterprise users via Vertex AI/Gemini API. The model achieves new SOTA benchmarks.

Sources: via Latent Space

Product

klaw Launches as Open-Source Platform for AI Agent Orchestration

klaw is an open-source platform designed to deploy, orchestrate, and scale AI agents, drawing parallels to Kubernetes for container management. It aims to simplify the management of intelligent agents.

Sources: via Hacker News

Product

GitHub Launches Agentic Workflows for Automated Repository Management

GitHub has introduced Agentic Workflows, a new feature designed to automate various repository tasks such as issue triaging, CI failure investigation, and test improvements using AI coding agents. This aims to streamline software development processes.

Sources: via TLDR AI

Executive

OpenClaw Creator Joins OpenAI to Advance AI Agent Development

The creator of the widely recognized OpenClaw project has announced their move to OpenAI, where they will focus on building accessible AI agents. This signals OpenAI's strategic interest in the area of AI agent development.

Sources: via Hacker News

Research

Anthropic Publishes Report on AI's Potential to Facilitate Chemical Weapon Creation

Anthropic released a report demonstrating that AI, even without direct human intervention, could be used to facilitate heinous crimes, including the creation of chemical weapons. This research highlights the potential risks associated with advanced AI technologies.

Sources: via Michael Parekh AI