What AI Update Did You Miss Today?

Mistral Vibe: In-App Session Picker with Live Preview

Mistral Vibe shipped an in-app session picker with live preview in version 2.24.2, letting developers browse and jump between past sessions without leaving the terminal. The release also sped up startup by rendering the TUI before session initialization completes, and cleaned up several worktree-related rough edges. Most of the remaining changes in this release were internal fixes and admin-config plumbing that don't warrant separate coverage.

Mistral Vibe

Codex CLI: Export Conversations to Markdown

Codex CLI 0.148.0 introduces a built-in `/export` command that saves a complete TUI conversation as a clean Markdown transcript, either to the system clipboard or to a new file. Previously, developers who wanted a readable record of a Codex session had to parse raw `.jsonl` session logs manually or rely on third-party community scripts. The addition also ships alongside startup improvements that let developers draft prompts while the TUI initializes, showing resume and fork progress in the background.

Codex

Claude Code Fixes Permission Dialog Over-Grant Bugs

Claude Code 2.1.235 patched two permission-dialog bugs that could leave a user granting more access than intended. Pressing Shift+Tab while typing a comment inside a permission prompt was silently approving the pending edit and granting session-wide edit permission, instead of simply closing the comment field. Separately, Anthropic corrected cases where a permission dialog's displayed text and its "don't ask again" option did not accurately reflect the scope of the grant it produced, and the option is now withheld whenever a dialog cannot fully display what is being approved.

Claude Code

Claude Code Adds Optional Spellcheck for the Prompt Input

Anthropic added an optional spellcheck setting to Claude Code that underlines misspelled words in the prompt input as a developer types. The feature works by shelling out to whichever spell-checking engine, aspell, hunspell, or ispell, is already installed on the machine, rather than bundling its own dictionary. It ships off by default and can be turned on in configuration, giving users a lightweight way to catch typos before they land in a prompt sent to Claude.

Claude Code

Codex CLI: Thread Credits and Cost Visibility

Codex CLI 0.148.0 adds visibility into estimated thread credits or cost directly in `/status`, the status line, and the terminal title, for workspaces on credit-based plans. Instead of switching context to a billing dashboard, developers can now see a running estimate of what a session is costing right where they're working. This follows a pattern from earlier releases like the token usage dashboard, giving users tighter feedback loops around usage on metered Codex plans.

Codex

Cursor Origin: Native Code Hosting Rivals GitHub

Cursor launched **Origin**, a native code hosting platform now rolling out in early beta to all paid-plan users, moving the company beyond being just an editor that sits on top of GitHub. Origin supports full **repo hosting**, **two-way GitHub sync**, and built-in **pull requests** with real-time comment syncing between the two platforms. It ships alongside integrations with **Vercel**, **Depot**, and **Buildkite**, with deeper **agent-native** capabilities — like agents opening and iterating on PRs autonomously — promised in later updates.

Cursor

Claude Code Adds a GitLab Merge Request Badge to the Footer

Claude Code now surfaces a live GitLab merge request badge in the footer and statusline, mirroring the GitHub pull-request badge that already existed. Developers on a GitLab remote with an authenticated `glab` CLI see the MR number (`!N`) along with its current state — draft, pending, or green — without leaving the terminal. The change closes a long-standing gap for teams whose primary git host is GitLab rather than GitHub.

Claude Code

Claude Code Lets You Open /permissions and /add-dir While Claude Is Working

Claude Code no longer forces users to wait for a turn to finish before adjusting settings. `/permissions` can now be opened mid-turn, with rule changes taking effect for the rest of the current turn, and `/add-dir` can add a working directory while Claude is actively running. Several other dialogs — including `/autocompact`, `/theme`, `/help`, `/config`, and `/advisor` — now open mid-turn in the fullscreen TUI as well.

Claude Code

Claude Code Fixes Permission Preview Leaks That Could Hide Commands From Approvers

Claude Code patched three related weaknesses in how permission previews — the summaries relayed to a remote approver before a tool runs — are secured and displayed. Credential masking could previously over-redact and hide the actual command, path, or destination from the person approving it, meaning an approver could unknowingly greenlight an action they couldn't fully see. A separate gap let previews relay to channel servers that hadn't been vetted by the inbound trust gate, and provider API tokens weren't always masked when directly followed by shell delimiters.

Claude Code

Grok 4.6 Now Available in GitHub Copilot

GitHub Copilot has added xAI's Grok 4.6 as a selectable model across eight surfaces, including VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the Copilot app, JetBrains, Xcode, and Eclipse. This is a third-party model integration (Grok is built by xAI, not GitHub) tuned for agentic coding and long-horizon, multi-step reasoning tasks. It is available on Copilot Pro, Pro+, Max, Business, and Enterprise plans, billed at usage-based provider list pricing, though Business and Enterprise admins must explicitly enable it since the policy is off by default.

GitHub Copilot

Replit Launches Native MCP Server for ChatGPT, Claude, and Slack

Replit launched a native MCP (Model Context Protocol) server, allowing ChatGPT, Claude, Slack, and any compatible MCP client to connect directly to Replit. Users can now create, find, inspect, update, and publish Replit apps — or duplicate an existing app to edit — entirely from within their AI assistant's chat interface. The move turns Replit into a callable backend for other AI tools rather than a standalone destination, extending Replit's reach to wherever developers already work.

Replit

Figma: Text Wrap with Balance and Pretty

Figma introduced native text wrap controls that make text layers responsive across different screen sizes, adding two new settings: Balance, which distributes text evenly across lines, and Pretty, which prevents a single word from being stranded alone on the last line. Either option can be applied to a text layer, a text style, or an individual paragraph directly from text settings. The update mirrors the CSS text-wrap: balance and text-wrap: pretty properties, closing a long-requested gap between Figma's typography controls and how modern browsers actually render text.

Figma

V0: Ask V0 to Manage Your Other Chats

V0 can now list, inspect, create, and continue other chats directly when asked within a conversation — turning V0 into a lightweight orchestrator across a user's own workspace instead of confining each chat to itself. This meta-level capability lets users manage multiple in-progress projects conversationally rather than manually switching between chats in the sidebar.

V0

Claude Code Turns Off Todo Tracking by Default on Newest Models

Claude Code 2.1.233 turns off its built-in todo and task-tracking tools by default when running on Opus 4.8, Sonnet 5, Fable 5, Mythos 5, and newer models. The TaskCreate, TaskGet, TaskUpdate, TaskList, and TodoWrite tools, which power the visible checklist Claude Code shows while working through multi-step work, no longer load automatically on those models. Setting the CLAUDE_CODE_ENABLE_TODO_TOOLS=1 environment variable restores the old behavior. Anyone accustomed to watching Claude Code check off a running task list will notice its absence unless they opt back in.

Claude Code

Replit Workspace Skills: Sharing and Admin Controls

Replit extended its Agent Skills system so workspace members can create skills and share them with the whole team, not just use them individually. Shared skills start turned off for everyone else, and each teammate chooses which ones to enable. Administrators gained the ability to manage every skill across the workspace and mandate specific ones for all members, backed by a redesigned Skills Library that makes skills easier to discover and administer.

Replit