Price moves, releases, and shutdowns across the tools in these stacks. Every entry links its source and the stacks it affects.
What changed in AI tools and the stacks that ship. One email a week.
Standalone Claude Design (claude.ai/design) will close by December 14, 2026. Notices have been added to /design-sync and /design-login to alert users. Developers should plan to migrate their work before the sunset date.
Fork subagents, which inherit full conversation context while isolating tool output from the context, are now enabled by default in the CLI, Agent SDK, desktop app, and VS Code. They can be disabled with CLAUDE_CODE_FORK_SUBAGENT=0.
Added environment variable to select which model the client-side auto mode classifier uses, independent of the server-side classifier. This provides developers with granular control over automatic tool invocation behavior.
Server-side auto mode classifier improved to automatically compact conversations exceeding safety limits rather than denying all subsequent tool calls. Maintains safety constraints while improving usability for longer conversations.
Reduced file system calls during plugin loading have accelerated initialization, with particularly notable improvements on Windows. Reduces startup delays for developers with many plugins enabled.
The Code tab now allows downloading large files that can't be previewed. The v0 iOS app adds a searchable skills picker, design system picker with web examples, and Quick Look video attachment preview support.
Cloud agents can create and share interactive canvases via link. The new Environments API allows programmatic creation and management of cloud environments. Cloud environments now support up to 300 repositories and load 5x faster.
Core and Pro builders can now select Claude Haiku 5.5 as their AI model in Power Mode. This extends Replit's available Claude models for AI-assisted development within the platform.
Enterprise customers can now run Replit in their own cloud environments with three options: full platform BYOC (private preview), BYOC for published apps (early access), and image exports for self-hosting. This gives enterprises control over deployment location and infrastructure.
Replit Core and Pro builders can now select Claude Haiku 5.5 from the model selector in Power Mode. This provides developers with access to Claude's latest smaller model for building within Replit's platform.
Replit has introduced a browsable directory of vetted agencies, consultants, and systems integrators that build with Replit. Developers can discover professional services partners, and qualified firms can apply to join the program.
Replit now offers multiple enterprise cloud deployment options: Full BYOC (private preview), BYOC for Published Apps (early access), and Image Exports (early access). Enterprises gain greater control over where the Replit platform and their applications run within their own cloud environments.
The Code tab now supports downloading large files that cannot be previewed, while the v0 iOS app gains a searchable skills picker, design system picker, and video attachment previews with Quick Look.
Fork subagents (created with subagent_type: "fork") now run by default across terminal, desktop, VS Code, and Agent SDK. This improves performance by isolating tool output from context. Set CLAUDE_CODE_FORK_SUBAGENT=0 to disable the behavior.
The standalone Claude Design at claude.ai/design will shut down on December 14, 2026. Users should migrate their design workflows before the deadline. A notice was added to /design-sync and /design-login warning of the closure.
The Code tab now allows downloading large files that cannot be previewed. The iOS app adds a searchable skills picker, design system picker, and video attachment preview capabilities.
Plugin command and agent file loading now requires far fewer file system calls, with the most improvement on Windows. Startup time is noticeably faster when many plugins are enabled.
Developers using Replit's Power Mode on Core and Pro plans can now select Claude Haiku 5.5 as their AI model. This gives builders access to Anthropic's latest lightweight model option alongside other available models.
Three deployment options now available: Full BYOC for running the entire Replit platform in your cloud tenant (private preview), BYOC for Published Apps to host Replit apps in your cloud (early access), and Image Exports to self-host Replit-built apps (early access). This gives enterprises greater control over where their platform and apps run.
Adds OpenTelemetry tracing guidance for API debugging and monitoring. Introduces LaTeX math rendering in VS Code responses, supporting $$ blocks, inline $...$ notation, and \[...\] blocks. Improves workflow rendering performance with incremental progress updates.
Lovable now supports GPT-6.1 Sol, OpenAI's latest model for complex coding and professional work. The model accepts text and image input. You can request it in your app or let Lovable automatically choose the best model.
Apps can now use OpenAI's GPT-6.1 Sol model for complex coding and professional work. The model accepts text and image input, complementing existing model options for AI-powered app features.
Expo SDK 58 is now available with new features and improvements for React Native development. Developers building with Expo should update to access the latest functionality.
OpenAI's GPT-6.1 Sol model is now available for AI features in Lovable apps. Supports text and image input for complex coding and professional work tasks.
New collaboration features allow Pro and Premier users to share song libraries and custom models with invited team members. Shareable custom models include 2x inputs for collaborative experimentation. This enables teams and studios to work together on music creation and model customization.
Pro and Premier users can now share workspace libraries with team members for collaborative song creation. Premier users can also share custom models with 2x the inputs to iterate on model development together.
Teams can now share song libraries and custom models with invited collaborators, with custom models supporting 2x the inputs for richer experimentation. These features let development teams collaborate on music generation and custom model creation.
Lovable apps can now use OpenAI's GPT-6.1 Sol model for complex coding and professional work tasks. The model accepts text and image input and can be specified in your app or automatically selected by Lovable.
Business and Enterprise plan users can enable SSO authentication for the Lovable MCP server in Claude, allowing team members to sign in with Okta or their company's existing identity provider. Users keep their existing workspace and project permissions.
Fast mode is no longer in research preview for Claude Opus 5.5 and no longer requires the beta header. This enables production use with default rate limits available across Start, Build, and Scale usage tiers.
Fast mode for Claude Opus 5.5 is now out of research preview and available in production without requiring beta header access. Claude Opus 5.5 now has standard fast mode rate limits in Start, Build, and Scale usage tiers. Earlier Opus versions continue to offer fast mode in research preview.
Fast mode is now generally available for Claude Opus 5.5, no longer requiring the beta header. It's included in the Start, Build, and Scale usage tiers and available on Claude in Microsoft Foundry for Azure deployments.
The Teams plan drops from $40 to $30 per month. Read from cursor.com on 2026-10-14, and on 1 earlier read(s) since the old price. Affects 6 stacks.
Adds OpenTelemetry GenAI semantic-convention attributes (gen_ai.provider.name, gen_ai.tool.name, etc.) for improved observability. Supports rewriting workflow agent prompts and models, introduces /advisor command on Claude Platform on AWS, and adds Slack Connect channel support for Claude Tag.
Two features were renamed for clarity: Fix comments (formerly Autofix) fixes selected review comments as a one-off repair, while Autofix (formerly Autopilot) brings pull requests to completion by repeatedly fixing review feedback and CI failures. Old command aliases remain for backward compatibility during a limited transition period.
Autofix was renamed to Fix comments for one-off review comment repairs, while Autopilot was renamed to Autofix for iterative PR completion. Old command aliases remain for backward compatibility during transition.
Lovable now provides integrated growth planning tools including audience analysis, ad campaign setup on Google and Meta, and creative generation. The tools recommend next steps based on your app's metrics and automate ad setup with tracking integration.
Workspace admins can register OAuth clients to let third-party applications request user access to Lovable profiles, projects, and workspaces. Sensitive operations require Business or Enterprise plans and client verification.
The .NET SDK now emits traces and metrics through System.Diagnostics for Auth, Postgrest, Storage, and Functions clients. Developers get SDK calls in application traces with no additional instrumentation overhead or dependencies.
Lovable added Growth tools that recommend growth strategies, set up Google and Meta ad campaigns, generate creative assets, and track results—all integrated into the project chat with no extra fees beyond standard costs and ad platform charges.
Claude SDK and API now support exporting spans to observability backends via OpenTelemetry. Available in Python, TypeScript, C#, Go, Java, PHP, and Ruby SDKs, as well as via the API Console. This enables developers to trace Messages API requests and Claude Managed Agents sessions in their existing observability infrastructure.
Lovable introduces Growth tools to help developers find and acquire customers. Includes AI-powered growth advice, Google/Facebook/Instagram ad campaign setup, and creative generation.
CodeRabbit renamed two features to match their actual functionality. The one-off repair feature (formerly Autofix) is now Fix comments, while the repeated auto-fixing feature (formerly Autopilot) is now Autofix. Old commands remain as backward-compatible aliases for a limited time.
The .NET SDK now emits traces and metrics through System.Diagnostics for all major clients (Postgrest, Auth, Storage, Functions). The SDK has no OpenTelemetry dependency and zero overhead when not in use, making observability seamless for .NET developers.
The .NET SDK now emits traces and metrics through System.Diagnostics for Auth, Postgrest, Storage, and Functions clients. Developers can subscribe to these diagnostics without adding OpenTelemetry dependencies to the SDK itself, integrating with existing observability infrastructure.
Lovable now includes growth tools to help find customers and grow your app, including traffic analysis, ad creation and management for Google and Meta platforms, and automated growth recommendations. Creative assets like videos and ad copy are generated from your app's content.
The .NET SDK now emits traces and metrics through System.Diagnostics for Auth, Postgrest, Storage, and Functions clients, with zero dependency overhead and no credential data in tags. Developers integrate it by adding OpenTelemetry hosting and exporter packages.
Two features were renamed for clarity: Autofix (formerly Autopilot) repeatedly fixes pull requests until ready, while Fix comments (formerly Autofix) performs one-off repairs on selected review comments. Old command aliases continue working for backward compatibility.
Origin now integrates with deployment, CI, code review, security, monitoring, and project management tools including Cloudflare, Vercel, Linear, and Sentry. Any developer tool team can now publish apps to the marketplace.
Google's Gemini 4 Argon is now available for Lovable apps as a frontier model for complex reasoning and long-horizon coding tasks. It accepts text, image, audio, and video input at premium pricing.
Showing 50 of 1470
What changed in AI tools and the stacks that ship. One email a week.
