What AI Update Did You Miss Today?

Cursor Design Mode: Multi-Select and Voice Input

Cursor 3.7 expands Design Mode in the in-app browser with two significant improvements. Developers can now select multiple UI elements simultaneously, giving the agent full visibility into their code, layout, and visual relationships to make coordinated cross-component edits. A persistent voice input overlay is also available throughout agent runs, allowing users to narrate changes or queue the next instruction by voice without waiting for the current operation to complete.

Cursor

Replit Custom Domains Now Configure DNS Automatically

Replit has eliminated one of the most friction-filled steps in app deployment: manual DNS configuration. When connecting a custom domain, Replit now automatically sets up the required DNS records through Entri, removing the need to copy records between Replit and a domain registrar. The feature works for most major domain providers and falls back to manual setup for unsupported registrars.

Replit

Replit Stripe Payments Now Available in Collaborative Workspaces

Replit has expanded its Stripe payment integration to work within collaborative workspaces, enabling entire teams to build, test, and monetize apps together using a shared Stripe account. Previously, Stripe was limited to individual Starter, Core, and Pro accounts; it now works for all collaborators in a shared workspace without requiring each person to manage their own connection.

Replit

GitHub Copilot: GPT-5.2 and GPT-5.2-Codex Models Deprecated

GitHub retired GPT-5.2 and GPT-5.2-Codex from most GitHub Copilot experiences on June 5, 2026, affecting Copilot Chat, inline edits, ask mode, agent mode, and code completions. Developers using these models explicitly are directed to switch to GPT-5.5 (replacing GPT-5.2) and GPT-5.3-Codex (replacing GPT-5.2-Codex). An exception exists for code review, where GPT-5.2 continues to be available.

GitHub Copilot

Claude Code fallbackModel: Automatic Model Failover When Your Primary Is Overloaded

Claude Code 2.1.166 introduces a fallbackModel setting that lets developers configure a chain of up to three backup models, automatically tried in sequence when the primary model is overloaded or unavailable. The feature applies to both interactive sessions and backgrounded workers, so long-running agents no longer hard-fail during peak load. The update also adds full support for disabling thinking on models that reason by default, giving teams finer control over cost and latency.

Claude Code

GitHub Copilot: One-Million-Token Context Windows and Configurable Reasoning Levels

GitHub Copilot introduced support for one-million-token context windows across VS Code, Copilot CLI, and the Copilot desktop app, enabling developers to work across larger codebases and multi-file projects without losing context. Alongside this, configurable reasoning levels allow users to dial the balance between speed and analytical depth, selecting higher effort for hard architectural decisions and lower effort for straightforward generation tasks. Both features increase AI credit consumption and are available on supported models in the currently listed surfaces, with expansion to additional platforms planned.

GitHub Copilot

Cursor: Canvas Design Mode and Context Usage Report

Cursor 3.7 introduces Design Mode for canvases, allowing developers to select and annotate UI elements directly on a canvas to guide edits visually rather than through text descriptions. Cursor also added an interactive Context Usage Report that shows agents' token allocation broken down across system prompts, tools, rules, and skills, surfaced as a live canvas artifact that users can query and debug further. Additional canvas improvements include agent-embeddable action buttons, better type error resolution, and expanded chart customization.

Cursor

Cursor SDK: Custom Tools, JSONL Stores, and Nested Subagents

Cursor's June 4 SDK update significantly expands what developers can build with programmatic agents. The TypeScript and Python SDKs now support custom tool definitions that agents invoke through the built-in MCP server, flexible storage backends including JSONL files and custom implementations, and deeply nested subagents that can spawn additional subagents to unlimited depth. Auto-review controls are also extended to SDK-driven runs, giving developers the same approval-gating workflow available in the Cursor desktop app.

Cursor

GitHub Copilot in Visual Studio: Plan Agent for Pre-Code Implementation Planning

GitHub Copilot's May 2026 update for Visual Studio introduced the Plan Agent, a dedicated planning mode that lets developers collaborate with Copilot before writing any code. The agent uses read-only codebase exploration tools, asks clarifying questions, and generates a detailed markdown plan file at .copilot/plans/plan-{title}.md. When the developer is ready, a single "Implement plan" button hands the plan off to agent mode for execution. The May update also added a unified Multi-File Change Summary view, giving developers one place to accept or undo all Copilot-proposed changes across files.

GitHub Copilot

Figma: Check Designs, Built-In Design System QA

Figma introduced Check designs, a native design quality assurance tool that automatically audits files against a team's design system and surfaces inconsistencies in one panel. The feature flags hard-coded values (colors, typography, spacing, radius) and suggests the correct design tokens as replacements, detects color contrast violations with WCAG 2.0 AA/AAA-compliant fixes, and identifies components detached from their source library or referencing unsubscribed libraries. Check designs is available on Organization and Enterprise plans.

Figma

Replit Agent Now Builds and Launches Custom Shopify Storefronts

Replit Agent can now design, build, and provision a complete custom Shopify storefront from a single conversational prompt, replacing hours of manual theme configuration with a ten-minute workflow. The Agent generates a unique frontend, creates a new Shopify store, and populates it with products, all from inside Replit. Users only need to visit Shopify once to claim their store and activate payments, then return to Replit to manage products, update listings, refine the design, and deploy to a custom domain. Unlike template-based approaches, every storefront is generated from scratch to match the user's specifications.

Replit

Codex CLI: TUI Keybinding Upgrades and Expanded Web Tool Access

Codex CLI 0.137.0 introduces expanded TUI keyboard support, including F13-F24 keybindings, paste functionality in searchable menus, and a compact reasoning-only status display for cleaner workflows. The release also brings hosted web and image tools to more code-mode flows, with standalone web searches now able to run in parallel, improving performance for research-heavy tasks.

Codex

GitHub Copilot: Fix Failing Actions Jobs in One Click, Now on Pro, Pro+, and Max

GitHub expanded the "Fix with Copilot" button for failing GitHub Actions workflow runs to Copilot Pro, Pro+, and Max subscribers, having previously restricted the feature to Business and Enterprise plans since May 18. When triggered, the Copilot cloud agent investigates the workflow failure, pushes a corrective change to the developer's branch, and tags the developer for review, all from its own isolated cloud environment. The feature is designed to delegate repetitive CI debugging tasks like broken tests and linter errors so developers can stay focused on higher-value work.

GitHub Copilot

GitHub Copilot Chat for Pull Requests: Side-by-Side Diff View Now Generally Available

GitHub Copilot Chat for pull requests on github.com moved from public preview to general availability on June 4, 2026, and is now accessible to anyone with a Copilot license. The GA release introduces a side-by-side interface that keeps the code diff and the chat conversation visible simultaneously, eliminating the need to toggle between views when asking questions or requesting inline edits. Underlying the experience is an improved pull request understanding model that automatically injects relevant diff and repository context into every conversation about a PR, delivering faster, more accurate answers without manual context-setting.

GitHub Copilot

Mistral Vibe: Image Attachments for Vision-Capable Models

Mistral Vibe v2.14.0 introduced image attachments via @-mentions in the TUI, enabling users to reference images directly in the chat interface when working with vision-capable models. Users can now @-mention an image file in the prompt input and the model will receive and reason over it inline. This removes the need for workarounds such as describing images in text or pre-processing them separately, making multimodal workflows a first-class experience in the terminal.

Mistral Vibe