This article discusses how Anthropic and OpenAI are adjusting their LLM usage policies, hinting at future pricing models. Anthropic is implementing usage throttling for its Claude model during peak hours, even for paying subscribers, to manage growing demand. This move, announced by Thariq Shihipar, limits session duration during weekdays from 5 am–11 am PT / 1 pm–7 pm GMT, affecting approximately 7% of users, particularly those on Pro tiers.

In contrast, OpenAI, through its engineering lead Thibault Sottiaux, announced the removal of usage caps for its Codex coding assistant, potentially to attract users from Anthropic. This dynamic suggests a competitive landscape where companies might offer unlimited access initially to gain market share, only to introduce monetization later, creating a cycle of shifting user bases.

👉 Read the full article on Gizmodo

The March 2026 Gemini Drop introduces several updates to the Gemini app, focusing on making AI more seamless and intuitive. Key features include the ability to transfer past AI chat history from other providers, allowing Gemini to quickly adapt to user context. The update also introduces Personal Intelligence, now free for all U.S. users, which connects across Gmail, Photos, and YouTube to provide personalized assistance for tasks like vacation planning.

Furthermore, Google TV now offers a more interactive experience with Gemini-powered visual answers and narrated deep dives. For creative users, the new Lyria 3 Pro allows subscribers to compose longer music tracks up to 3 minutes, turning photos or thoughts into high-fidelity anthems with lyrics. Finally, Gemini Live receives a significant upgrade to version 3.1, enabling faster conversations that retain context for twice as long, facilitating more natural dialogue.

👉 Read the full article on Google Blog

The OpenAI Codex documentation outlines various use cases for its AI coding assistant, demonstrating its versatility in software development and related tasks. It can significantly accelerate workflows by enabling faster review of pull requests, catching regressions and potential issues before human intervention. Codex is also adept at building responsive front-end designs, converting screenshots and visual references into UI with visual checks, and even transforming Figma designs into polished code.

Beyond front-end, it supports building for iOS and macOS by scaffolding, building, and debugging SwiftUI apps. Other applications include analyzing datasets and generating reports, creating browser-based games, generating slide decks with image generation, and understanding large codebases by tracing request flows and mapping modules. Codex can also be integrated with platforms like GitHub, Slack, and Linear, and can be used to kick off coding tasks from Slack threads.

👉 Read the full article on OpenAI Developers