OpenAI's GitHub repository has shown signs of GPT-6, with a developer submitting a pull request that adds two environments (gpt6-js-image and gpt6-py-image) for the Computer Use project. Sam Altman said Astra marks a milestone where AI models can operate computers at a human-like level.
A developer connected Banini’s Threads posts to Claude and Playwright to build an open-source tracker that sends Telegram alerts on her stop-loss updates, sparking mock reverse trades online.
Anthropic engineer Thariq explained how the team uses Skills inside Claude Code, outlining nine common use cases and four practical design techniques.
Anthropic accidentally exposed Claude Code's full source code due to a misconfigured npm file, revealing 512k lines of TypeScript, a three-layer memory system, KAIROS daemon mode, autoDream, undercover contribution mode, and internal model names. Concurrent npm supply chain attack raises AI security and crypto market worries.
Anthropic exposed the full source of Claude Code via a misconfigured npm source map, revealing a 3-tier memory system, KAIROS daemon, Undercover mode, and internal model benchmarks. The leak coincided with an npm supply chain attack, raising security concerns across the AI industry.
Anthropic accidentally leaked Claude Code's full source code on npm (60MB, 512K lines TypeScript), revealing three-layer memory, KAIROS daemon mode, autoDream consolidation, Undercover Mode for anonymous open-source contributions, internal model names (Capybara, Fennec, Numbat) and performance data (Capybara false claims 29-30%). Concurrent npm supply chain attack via malicious axios package urges users to rotate credentials.
A misconfigured npm source map file exposed Anthropic's Claude Code full source code, revealing a three-layer memory system, KAIROS daemon mode, autoDream memory consolidation, internal model codenames (Capybara, Fennec, Numbat), and an Undercover Mode for anonymous AI contributions to open source. A simultaneous npm supply-chain attack also raises security concerns.
