Claude Code

Goldman Sachs is removing Anthropic's Claude Code coding tool from its bankers and traders, restricting staff access to the AI assistant. The move has drawn widespread attention in tech and finance circles, with debate over why the bank is limiting an AI tool that had been embraced as a productivity booster for employees.

A software developer has published a write-up explaining a switch from Anthropic's Claude coding assistant to Pi, sparking discussion among developers about which AI assistant best fits day-to-day programming work. The piece, shared widely on Hacker News, is drawing hundreds of upvotes as readers weigh in on their own experiences moving between AI coding tools.

A claim is circulating that Anthropic's Claude Code regularly sends each background agent's full transcript to its model as part of normal operation, raising questions about how much context and user activity is transmitted during automated coding sessions. Developers are debating the privacy and cost implications of routine transcript uploads, and whether users are adequately informed about this behavior.

Anthropic has introduced measures preventing its Claude AI model from being used to reverse-engineer or clone video games. The move comes as AI tools increasingly demonstrate the ability to reproduce playable game mechanics from footage or code, raising copyright concerns across the games industry. Developers have welcomed the restrictions, while others argue it highlights unresolved questions about how far AI companies should police how their models are used.

GitHub is expanding its range of third-party coding tools to include Anthropic's Claude and OpenAI's Codex. In an analysis, Matt Asay writes that the company hopes the move will strengthen developers' loyalty to its own platform, as competition among AI coding assistants intensifies and GitHub seeks to remain developers' central workplace.

A GitHub project called free-claude-code by Alishahryar1 is trending. It claims to let developers use AI coding assistants including Claude Code, Codex, VSCode, Pi and OpenCode for free, advertising over 1.3 billion free tokens accessible from a terminal, IDE, app, phone or browser. The tool supports multiple harnesses and models and describes itself as terms-of-service friendly.

Anthropic has updated the system prompt of its Claude Code coding tool to stop users from employing it to recreate existing video games, as reported by Tom's Hardware. The move responds to a wave of AI-assisted decompilations, recompilations and ports flooding the retro gaming scene, raising fresh questions about copyright and how AI coding assistants should be restricted.

Security researchers at Noma Security report a vulnerability in Anthropic's Claude Code Desktop, warning that the app's preview pane can be used to exfiltrate user data. The disclosure is drawing attention among developers and security professionals, who are debating the risks of agentic coding tools having access to local files and how vendors should sandbox such features.

Linus Torvalds is now openly using AI in an experimental branch of the Linux kernel, under the name noref-filemap-read-with-llm. Reports say he uses Claude Code for tests and commit messages, framing it as speeding up monotonous work rather than replacing developers. Related AI news rounds also mention a four-person team that reverse-engineered Adobe technology.

A developer-run GitHub project called Claude-Code-Game-Studios is drawing attention for turning Anthropic's Claude Code into a simulated game development studio. It bundles 49 AI agents, 72 workflow skills and a coordination system that mirrors the hierarchy of a real studio, from producers to programmers. The project is sparking debate about how far AI agents can go in automating creative production work.

Andre Karpathy, the AI researcher and OpenAI co-founder, says he built a working video platform called HackerTube in just two hours using Claude Opus 5.5. The feat is drawing attention as a demonstration of how quickly AI coding assistants can now take a project from idea to functioning software, and as another showcase of Anthropic's latest model capabilities.

Weeks after the free-software internet embraced vibe-coded clones of Office, Photoshop and other commercial programs, that wave may be winding down. Tom's Hardware reports that users began noticing projects built with Claude are coming to an end, suggesting Anthropic has moved to restrict using its AI to develop lookalikes of competing commercial software.

Anthropic has introduced stricter rules for its Claude Code programming tool to prevent users from cloning games and other software. The company says the measures address how easily AI tools can reproduce existing products. Some developers, however, criticize the new restrictions as going too far, sparking debate about where to draw the line between inspiration and infringement.

Anthropic has published research describing how its Claude AI model can be used to reduce the attack surface of code, identifying and simplifying parts of software that could be exploited by attackers. The approach points to a security role for AI coding assistants beyond writing features, sparking discussion among developers about how far AI can be trusted to harden critical systems.

A 23-year-old quant developer in China says he ran 300 Claude Code agents in parallel as his own trading research desk, wiring them to a verification layer so outputs could not be taken at face value. The setup, described as trading research on autopilot with live monitoring, is drawing attention to how far individual developers can now push AI agent swarms.

Google has acknowledged that Gemini 4 Argon ranks well behind Anthropic's Claude and OpenAI's GPT-6 when it comes to coding for Android. The admission, reported by 9to5Google, is drawing attention for its unusual frankness about the model's weaknesses. It comes amid intense competition among AI labs over developer tools and coding performance, where rankings are closely watched.

Orkas, an open-source MIT-licensed desktop app for macOS and Windows, coordinates a team of AI agents locally. A Commander component converts user goals into to-dos and manages 16 built-in specialist agents alongside external tools including Claude Code, Codex, OpenCode, OpenClaw and Hermes. It supports projects, scheduled automations, connectors, built-in models or user-supplied API keys, and is free to use. Developer communities are discussing its local-first approach to multi-agent work.

A detailed write-up describes how a real company runs 17 Claude Code bots in production for roughly $100 per month, breaking down actual costs and operational experience. The account of how cheaply AI coding agents can be deployed at scale is drawing attention among developers weighing the economics of AI-assisted engineering.