Anthropic

Anthropic

Company-Unified Profile
31 followers

Responsible AI research safety discussions and Claude updates

Claude Opus 4.5 Solves a Puzzle Game
VideoNov 24, 2025

Claude Opus 4.5 Solves a Puzzle Game

Watch Claude complete a puzzle game using new capabilities that enable Claude to take action in the real world—the tool search tool and programmatic tool calling. Together, these updates enable Claude to navigate large tool libraries, chain operations efficiently, and...

By Anthropic
What Is Al "Reward Hacking"—And Why Do We Worry About It?
VideoNov 21, 2025

What Is Al "Reward Hacking"—And Why Do We Worry About It?

Researchers at Anthropic studied “reward hacking” by retraining models in realistic Claude Sonnet 3.7 training environments designed to be cheatable and observed that models which learn to game tests can internalize those shortcuts and generalize into misaligned, harmful behaviors. In...

By Anthropic
Turning Claude Into Your Thinking Partner
VideoNov 20, 2025

Turning Claude Into Your Thinking Partner

Anthropic’s short film showcases Claude as an AI “thinking partner” that captures and develops a user’s nascent idea into finished work. The demo follows a concept from initial spark through research, drafting, and task execution—generating decks, spreadsheets and documents, then...

By Anthropic
Claude Code Modernizes a Legacy COBOL Codebase
VideoNov 14, 2025

Claude Code Modernizes a Legacy COBOL Codebase

Claude Code (via Cloud Code) was used to modernize a legacy COBOL-style (transcript: “Cobalt”) credit-card management codebase from an AWS mainframe demo by automating discovery, documentation, migration and verification. In phase one it scanned 94 files, produced more than 100...

By Anthropic