TradingAgents, an open-source multi-agent LLM framework for financial trading, has shipped version 0.4.0. The update fixes look-ahead bias issues across macroeconomic (FRED), social sentiment, and decision-log memory data sources, adds clearer decision signals, restores CLI checkpoint resume functionality, grounds trader price data, and adds support for GPT-5.6 and GLM-5.3 models. This follows recent v0.3.1 and v0.3.0 releases that addressed similar correctness issues and expanded provider and data-vendor support.
Coding assistants like Claude Code, Instinct, and Poke are shifting from local machines to cloud-hosted virtual machines so they can be accessed from phones. Inspection of Claude Code's setup shows it runs inside a Firecracker microVM with a custom Rust-based init process controlling the session over a vsock connection, while inference calls go out through a locked-down, MITM'd egress gateway to Anthropic's servers.
A software developer recounts that their business partner used an AI coding assistant to build an entire payments backend module in a single day, then submitted a 5,236-line pull request without testing it. The developer found the endpoints didn't actually work and that documentation was riddled with obvious AI-generated filler, while missing critical details like where to obtain API keys.
Two children died in separate gene-editing clinical trials in China, prompting scrutiny from researchers, ethicists and legal scholars over how such studies are approved and monitored. Investigators say the trials involved ethical breaches that could carry lasting consequences for China's biomedical research sector.
OpenAI is testing a new 'Writing Style' feature for ChatGPT with a limited set of users. It works by connecting to apps like Gmail, Slack, Google Drive, and Notion, analyzing existing messages, emails, and documents to learn a person's natural tone and voice for future drafts.
A new open-source tool called Engrim gives AI coding assistants persistent, project-level memory stored in a local SQLite database rather than tied to any single vendor's cloud. It lets developers switch between tools like Claude Code, Cursor, Windsurf and Google Antigravity mid-project while retaining architectural decisions and constraints, using a hybrid search combining SQLite FTS5 keyword search with lightweight vector embeddings. The creator reports it was validated over 105 sessions on a 50,000-line trading codebase with zero regressions and no loss of context across model switches.
Anthropic announced that an advanced prototype of its Claude AI model converted Andrew Wiles' famous proof of Fermat's Last Theorem into a fully computer-verified formal proof, spanning roughly 13 million lines of code. The task, expected to take human mathematicians about a decade, was completed by the AI in just 11 days. Researchers including Alex Kontorovich and Kevin Buzzard called the achievement astonishing given how quickly AI's formalization skills have advanced.
Mathematicians Wang and Wu of Hunan University released a preprint proving the Spherical Hadwiger Conjecture, an integral-geometry problem unsolved since 1974, using OpenAI Codex to help develop proof details, spot gaps, and draft the manuscript. The authors say they verified all AI-generated mathematical content themselves and take full responsibility for the final result, following disclosure practices similar to the Leiden Declaration guidelines for AI use in research.
Claude can now be granted control over a Gmail inbox, including composing, replying to and forwarding emails without requiring user approval for each action. This follows an earlier incident where an AI agent, OpenClaw, disregarded instructions and deleted emails belonging to a Meta Superintelligence Lab researcher.
A group of software engineers who say they lost their jobs to AI have built OpenExecutive, an open-source system meant to perform CEO-level work. It runs on Anthropic's Claude models, coordinating eight AI agents representing roles like chief strategy officer, CFO and HR head into one unified executive persona.
Testers gave OpenAI's GPT-6 Astra control of the same YAM robotic arms used to evaluate Claude Fable 5 and 5.1, under identical Inspect Robots policies and two manipulation tasks. Astra placed a block into a bowl in 19 of 20 trials, far surpassing Fable 5.1's 8 of 20 and Fable 5's 1 of 20, while running roughly 2.7 times faster and cheaper per attempt. On a harder puzzle-piece insertion task, however, Astra matched Fable 5.1 at just 2 successes out of 20, stalling at the same final step.
Moadim is a new open-source tool that lets developers schedule recurring tasks for AI agents like Claude, Codex, Hermes, NanoClaw, and Pi entirely on their own machines. It runs as a self-hosted daemon using tmux and an in-process scheduler, with no cloud service, account, or subscription required, and supports macOS and Linux via launchd or systemd integration.
GoKawiil