Code Velocity

Anthropic

AI assistance impacting coding skills research from Anthropic
AI Research

Coding Skills: AI Assistance's Dual Impact on Developer Growth

Anthropic research reveals AI assistance can boost coding productivity but may hinder skill development, particularly debugging. Learn how interaction patterns influence mastery.

·10 min read
The Anthropic Institute logo, symbolizing responsible AI research and societal impact
AI Research

Anthropic Institute: Confronting AI's Societal Impact

Anthropic Institute launches to confront the profound societal challenges posed by powerful AI, leveraging interdisciplinary research for responsible AI development and global governance discussions.

·7 min read
Illustration of AI agent tool evaluation and optimization using Claude Code for enhanced performance.
Developer Tools

Agent Tools: Enhancing AI Performance with Claude Optimization

Discover how to write and optimize high-quality AI agent tools with Anthropic Claude. Learn building prototypes, comprehensive evaluations, and agent collaboration for enhanced AI performance.

·7 min read
GitHub Copilot interface showing Claude and Codex AI coding agents integrated, enhancing developer workflow.
Developer Tools

Claude & Codex on Copilot Business/Pro: Enhanced AI Coding

GitHub Copilot Business and Pro users now gain access to powerful AI coding agents, Claude by Anthropic and OpenAI Codex, directly within GitHub workflows, enhancing developer productivity.

·5 min read
Graphic illustrating the concept of AI fluency and human-AI collaboration, with data points.
AI Research

AI Fluency Index: Measuring Human-AI Collaboration Skills

Anthropic's new AI Fluency Index measures critical human-AI collaboration skills, highlighting the importance of iteration and critical evaluation for effective AI use.

·7 min read
Anthropic's official statement regarding the Department of War's potential supply chain risk designation over AI ethics.
AI Security

Anthropic Defies War Sec on AI, Cites Rights and Safety

Anthropic defies Department of War's supply chain risk designation, standing firm on ethical AI use, banning mass domestic surveillance and unreliable autonomous weapons.

·4 min read
Diagram showing distillation attack flow from frontier AI model to illicit copies through fraudulent account networks
AI Security

Anthropic Exposes Distillation Attacks by DeepSeek and MiniMax

Anthropic reveals DeepSeek, Moonshot, and MiniMax ran 16M illicit exchanges to distill Claude's capabilities. How the attacks worked and why they matter.

·4 min read
Claude Code Security scanning pipeline showing code analysis, data flow tracing, verification, and patch suggestion stages
Developer Tools

Claude Code Security: AI Vulnerability Scanner

Claude Code Security found 500+ zero-day vulnerabilities in open-source code. It scans, reasons about, and patches security flaws that traditional tools miss.

·5 min read
Claude Sonnet 4.6 OSWorld benchmark progression showing 65% improvement from Sonnet 3.5 to 4.6
AI Models

Claude Sonnet 4.6: Frontier Coding at Sonnet Pricing

Claude Sonnet 4.6 delivers Opus-level coding at $3/$15 per million tokens. 65% better computer use, 1M context window, and stronger prompt injection resistance.

·6 min read
GitHub Agentic Workflows diagram showing an AI agent triaging issues and creating pull requests inside a GitHub Actions pipeline
Developer Tools

GitHub Agentic Workflows: AI Agents in CI/CD

GitHub Agentic Workflows let AI agents automate repo tasks inside GitHub Actions. Supports Copilot, Claude Code, and Codex in isolated containers.

·6 min read
Claude Opus 4.6 benchmark comparison chart showing #1 rankings on Terminal-Bench 2.0, Humanity's Last Exam, and GDPval-AA
AI Models

Claude Opus 4.6: #1 in Coding and Reasoning Benchmarks

Claude Opus 4.6 ranks #1 on Terminal-Bench 2.0 and Humanity's Last Exam, beating GPT-5.2 by 144 Elo. New features: agent teams, compaction, and 1M context.

·7 min read
Xcode 26.3 interface showing an AI coding agent autonomously building a SwiftUI view with Xcode Previews verification
Developer Tools

Xcode 26.3: Apple Adds Agentic AI Coding

Xcode 26.3 adds agentic AI coding with Claude and Codex agents via Model Context Protocol. Agents search docs, explore files, and verify work through Previews.

·5 min read
NASA Perseverance rover on the Martian surface following the AI-planned 400-meter route through a rock field
AI in Practice

Claude AI Powers NASA's First AI-Planned Mars Rover Drive

Claude AI planned NASA Perseverance's 400-meter Mars drive, the first AI-planned route on another planet. JPL estimates 50% faster route planning.

·8 min read