AI coding tools news
November 13, 2026 · 26 stories
The day's news for developers who use AI coding tools, picked and summarised from official changelogs, the media and Hacker News. Resets and limit changes are on the reset tracker.
Claude Code v2.1.325 fixes security issue and updates terminal tips
Claude Code version 2.1.325 fixes a security vulnerability in zstd-compressed response decoding. It adds a notebook cell spinner tip in VS Code and integrated terminals after reading or editing a Jupyter notebook, and introduces error recovery instructions for organization policy blocks.
Claude Code adds model configuration and task organization in Claude Cowork
Claude Code adds model configuration along with task organization via projects in Claude Cowork. The update explains model switches between Opus 5 and Opus 5.5, details training data freshness, and adjusts the Messages API to preserve thinking blocks against distillation.
Antigravity v2.32.0 adds Dynamic Workflows and new themes
Antigravity v2.32.0 adds Dynamic Workflows, allowing agents to orchestrate parallel sub‑agents and multi‑step tasks in Python using the bundled runtime. It also introduces Everforest Light and Dark themes, prompt navigation shortcuts, and optional Submit Prompt shortcut. The model selector now expands to fit screen height and shows a Refresh button when models fail to load, and agent‑controlled browser tabs gain download‑rejection notifications and hidden cross‑site frames.
ChatGPT Wallet launches for US Pro users
ChatGPT Wallet is rolling out today for ChatGPT Pro subscribers in the US. Eligible users must be 18 years or older, have a US billing address, and use a US‑issued American Express, Chase, Mastercard or Visa card.
Gemini can read Google Wallet for budgeting but cannot transfer money
Gemini can read your Google Wallet to help you budget using your actual spending. It cannot move any money.
GitHub Copilot CLI v1.0.112-0 adds macOS Ventura startup improvement and default context windows
Improved best‑effort startup on macOS Ventura 13.5 or later without expanding official OS support. Uses each model’s recommended context window by default while preserving explicit context choices. /chronicle cost‑tips now includes Copilot App sessions, recommends cheaper models and reasoning settings based on usage, and reports sooner when local usage is available. Fixed several error‑handling and plugin‑display issues.
LangChain raises valuation to $3 B
LangChain now claims 7,000 customers of its model and platform agnostic agent management tool LangSmith.
GitHub Copilot app adds GitHub Enterprise Server support
You can now add a GitHub Enterprise Server (GHES) account to the GitHub Copilot app. This lets you work with remote repositories, issues, and pull requests from GHES.
Export‑banned Nvidia chips used to train China’s most advanced AI models
New research shows the most advanced AI models from China were trained on banned Nvidia chips.
Claude Code update adds model configuration and settings info
Claude Code now includes a model configuration section. It explains why Claude switched models when using Opus 5 or Opus 5.5. It also notes how up‑to‑date Claude’s training data is and lets users change the model, effort, and thinking settings.
Perplexity adds Teammates feature to Mac app for Max subscribers
Perplexity's Teammates are now available in its Mac app for Max subscribers. The agents can own and divide ongoing work.
Introducing ChatGPT Wallet
ChatGPT Wallet is a secure way for agents in ChatGPT Work, Codex, and dots to complete purchases with user approval and existing card protections.
ChatGPT Wallet enables AI agents to shop for U.S. Pro users
ChatGPT Wallet lets AI agents shop and check out on users' behalf, starting with a limited rollout to U.S. Pro subscribers.
DeepSeek updates pricing and quota page
The DeepSeek pricing page now lists separate charges for Web Search & Fetch, File Parsing & OCR, and Speech services. Search (text, image) and web fetch are billed per 1,000 searches or fetches. File parsing and OCR cost $0.225 per 1,000 pages. Text‑to‑Speech is $0.45 per 10,000 characters (limited‑time $0.225) and Speech‑to‑Text is $0.116 per hour of audio (limited‑time $0.058).
Meta brings Muse AI agent to Chrome with new extension
Meta's Muse AI agent is now available as a Chrome extension. It provides sidebar conversations, task delegation, approvals, and connected services.
Chinese AI breakthrough in complex lung cancer diagnosis
Chinese AI achieved a breakthrough in complex lung cancer diagnosis. 250 top doctors from leading tertiary hospitals participated in the joint evaluation.
Codex 0.169.0 adds multi‑side‑chat tabs and custom model providers
The opt‑in terminal workspace now supports multiple side‑chat tabs that can be switched with keyboard or mouse and reopened after restart. Users can search older tasks by name, prompt preview, or working directory. A Collapse turn details option folds completed activity in the full‑screen transcript while keeping answers visible. Custom model providers can be added or edited from the terminal, with gateway onboarding via device code for SSH or browser‑less sign‑in. Ctrl+S in the terminal report editor saves a diagnostic ZIP locally even when uploads are disabled. On macOS and Linux, the opt‑in Cloud connection allows sending images, selecting a model and reasoning effort, and archiving.
Chinese Large Model Enters Top‑Tier Rankings
A Chinese large‑scale AI model has entered the top‑tier ranking of models.
openJiuwen releases RSI native AgentOS for self‑developing agents
openJiuwen released the RSI native AgentOS. The platform lets agents develop the features users request themselves.
Fx‑Bio‑1031 open‑source model released, supports local deployment
The Fx‑Bio‑1031 model is released as open source. It can be deployed locally without data leaving the domain.
Kiro IDE adds workflow step view, skill reload after compaction, and readable Mermaid diagrams
The release lets users follow a workflow run step by step from the Workflow tab. After Kiro compacts a conversation, the agent now reloads the requested skills and steering files. Large Mermaid diagrams in chat remain readable. Fixes include freezes after permission changes, stalls before large file writes, and missed reloads after editing configuration on a Windows network drive.
Codex v0.168.0 adds model provider selection, side‑conversation UI, audio recordings, animated GIF generation
The release adds a Model providers setting under Sessions → Model providers to choose the default for new chats. It introduces a side‑conversation pane that can be resized or zoomed with leader z. Audio‑capable models can now attach recordings, and short recordings are kept longer using clip duration. Animated GIF generation is supported with frame count and timing controls. Bug fixes include filtering of local TUI history and resume pickers.
GitHub Copilot CLI v1.0.111-0 adds regex support and host‑allow‑session‑creation
Added regular expression support to enterprise‑managed Shell, Bash, and PowerShell permission rules. Added --host-allow-session-creation to let remote clients create sessions on CLI hosts for one launch without changing saved settings. Improved organization support links in policy guidance and managed settings notices, and indexed search becomes available sooner after restarting with its enabled notice appearing only when the index can serve searches. Fixed refresh of /usage premium request usage, prevented duplicate MCP OAuth reconnects, corrected markdown table rendering of wide symbols, updated footer.showCodeChanges description, and adjusted the /usage activity graph to fit narrow terminals.
Google Gemini 4 Hydrogen pricing announced
Google launched Gemini 4 Hydrogen with a price of $1.50 per million input tokens, $0.075 per million cached input tokens, and $7.50 per million output tokens. Gemini 4 Carbon and Argon cost $2.00 per million input, $0.10 per million cached input, and $10.00 per million output through Nov 30 2026, then double to $4.00, $0.20, and $20.00 from Dec 1 2026. Argon is priced the same as Carbon but claims faster responses and lower token spend.
Amplitude launches Wave product agent built on OpenAI models
Amplitude launched a proactive product agent called Wave built on OpenAI models. Wave identifies what to build, helps teams ship changes with coding agents, and measures impact.
Kiro CLI 3.3 adds /copy commands and compaction threshold
This release adds /copy N and /copy last N to copy earlier responses or the last few messages, and lets you choose the context usage at which automatic compaction starts. It also shows each sub‑agent’s task in the agent monitor, moves the /model search, sort, and edit controls to Up, keeps a V2 agent’s allow rules from widening when upgraded to V3, reloads a stalled workflow session after an agent restart, and fixes headless, resume, /memories, /goal, voice, clipboard, session, and search display issues.