What’s new in Playcode

Clicking Logout could quietly do nothing: the menu closed, no error appeared, and you were still signed in. It happened when Playcode was open in more than one tab - the sign-out waited on a local data cleanup that a second tab was holding, and then waited forever.

Sign-out no longer waits for anything. Your session ends the moment you click Logout, and the local cleanup happens afterwards where it cannot get in the way.

Two related gaps closed at the same time:

• Every tab. Logging out in one tab now ends the session in your other Playcode tabs too, instead of leaving them signed in with a session that no longer exists. A tab holding unsaved work stays put rather than reloading over it.
• Only your own data. Signing out used to try to clear the cached data of every account that had used the browser. It now clears only yours, so signing out of one account no longer disturbs another on a shared computer.

If you work in a team workspace, the Home screen previously checked your personal credit balance - so it could warn "You're out of credits" even though your team workspace had plenty. Fixed:

• The right wallet. Home's composer now shows the credits of the

workspace your new project will be created in - the same number the workspace menu shows.

• The right button. "Top up" from Home adds credits to that workspace,

and team members without billing access see a note to ask a workspace admin instead of a checkout they can't complete.

• Warnings stay honest. Dismissing a low-credit warning in one

workspace no longer hides a real warning in another.

When we ship a new build, Playcode offers to reload so you get it. That reload used to happen immediately - and if you had a browser project you hadn't saved yet, it went with the page.

Now the reload saves first. If you're signed in, your project is written to your account before anything else happens, and edits still syncing get a moment to land. If it can't be saved - you're signed out, for instance - Playcode asks before reloading instead of deciding for you.

Two related gaps closed at the same time:

• Other tabs. Reloading in one tab used to reload the rest instantly. A tab holding unsaved work now keeps it and waits for you.
• Closing or refreshing the page. The browser's "leave site?" warning now appears for an unsaved project on any exit, not only when moving between pages inside Playcode.

Projects created with the AI builder and Cloud projects were never affected - those save to your account continuously.

If your first publish failed, the project could be left unable to publish at all: pressing Publish again would stop immediately, and the project card kept showing "Publishing..." forever. That is fixed. Every publish attempt now ends in a state you can act on, so you can always fix your app and publish again.

When a publish fails its health check, we also now show you your app's own startup log with the error. That is the part that actually explains the failure: your code deploys fine, but the app's server does not come up, so the published URL never responds. Previously that log was not kept, and the deploy log looked clean because the deploy itself had succeeded.

Two related improvements: the publish progress window no longer blocks the editor, so you can keep working while a publish runs or close it if something goes wrong. And if a publish is interrupted and left behind, we now clean it up automatically instead of leaving the project stuck.

If a health check keeps failing, check that your app listens on the path set as healthcheck.path in playcode.yml, or point it at a path your app actually serves.

Sharing an editable link from a Pro workspace now works the way it should: any number of collaborators can join at the same time, and they do not need a Playcode account - opening the link is enough.

Previously, collaborators who were not signed in could be turned away with a "collaboration limit reached" message even when the project owner was on Pro. That check now runs on our servers against the project's actual plan, so it can no longer misfire.

Free projects keep the existing limit of 2 people at once, and the project owner now sees a notice when someone was turned away - the AI agent working in your project no longer counts toward that limit.

When you switch your project's shared link from "can edit" to view-only, everyone who already has the project open now loses edit access within a second - their editor switches to a read-only state with a clear notice. Turning editing back on unlocks connected visitors just as instantly.

Previously the toggle only applied to people who opened the link *after* the change: visitors who were already connected could keep typing until they reloaded the page, which made the setting look broken.

Opening a shared project link without signing in now gives you a way to browse the whole project, not just the files the owner happened to leave open.

The project menu has a new Files entry under View. It opens the file tree in a panel on the left - click any file to open it, exactly like the project's members do. Close the panel with the x in its corner and reopen it from the menu any time.

On read-only links the tree is view-only: you can open files, split the editor, and copy paths, but options that would change the project - creating, renaming, moving, or deleting files - are not shown. Links that allow editing keep the full set.

Converting a browser project to a Cloud full-stack app no longer fails when the project carries a large media library.

Projects with many hundreds of images could get stuck at the import step with an unhelpful error, and retrying or restarting the project computer never helped. The importer now transfers big asset sets reliably, and if anything does go wrong during a conversion, the error now names the actual cause instead of a generic failure - so the AI can fix it or route it to us with real detail.

Cloud publishing got a stronger guarantee: Publish deploys the exact code that is in your workspace at the moment you press the button - every time, no matter what the AI has been doing with Git behind the scenes.

Previously, if the AI organized its work into a Git branch, a publish could quietly ship an older version of your app, or fail repeatedly with a cryptic Git error that retrying never fixed. That whole class of problems is gone:

• What you see is what ships. Publish captures your current workspace state as a single immutable version - unsaved edits included - and deploys exactly that. Branches, background commits, and Git bookkeeping can no longer change what goes live.
• Every publish shows its source. The publish dialog and history now identify each attempt by its branch and version - so "what exactly is live right now?" always has a clear answer, for you and for the AI.
• Honest errors instead of retry loops. In the rare case your workspace is genuinely out of sync with the published history, Publish stops before touching production and explains what happened in plain words - your live site stays untouched, and the AI can see the same explanation and fix the mismatch for you.

Your Cloud terminal now keeps running as long as its tab is open.

Before, switching to another tab and back would drop the session and start a fresh shell - losing whatever was running and your command history. Now you can move between the terminal, preview, and AI Chat freely: the terminal stays connected in the background, and switching back brings you right back to the same session, scrollback and history intact. Dragging the terminal to another panel keeps it too.

Closing the terminal tab still ends the session. And if you leave the whole app for a while, idle terminals disconnect on their own to keep things tidy - reconnecting the moment you come back.

Claude Sonnet 5 credit estimates now account for its launch pricing and newer tokenizer together.

The model picker badge and actual credit use are aligned, so Sonnet 5 should no longer overstate expected credits during the introductory pricing window.

Picking Premium mode in the AI chat sometimes did nothing in a freshly opened project - the switch quietly snapped back to Economy. That's fixed: your mode choice now sticks in every project, even before you've sent your first message.

Improvements

• Premium mode is selectable everywhere - the choice is saved the moment you make it, in any project
• Set your default mode - choose whether new chats start in Economy or Premium from Preferences > AI Model > Default Mode, and it applies across all your projects

Two billing fixes for existing subscribers. Opening the Billing Portal silently failed on mobile (it worked on desktop) - the link is now opened in a way phones don't block. And changing your plan (for example Pro → Business) no longer dead-ends on "redirecting to secure payment" - subscribers are now sent to the billing portal's plan switcher.

Fixes

• Billing Portal opens on mobile - previously nothing happened on phones; the portal now opens reliably
• Plan upgrades work for current subscribers - switching plans routes to the billing portal instead of failing

Cloud publishing now observes deploy logs in short, bounded checks instead of waiting on one long-running server step. This helps the publish dialog keep up with newer publish and rollback activity, especially when a build step is quiet for a while.

If a publish is cancelled or a deploy step takes longer than expected, Playcode can now notice follow-up publish activity faster instead of appearing stuck with no output.

We fixed a caching issue that could keep the Preview showing old CSS after you changed a cloud project. Cloudflare was telling browsers to keep some Preview stylesheets for too long, so a normal reload could bring back stale styles instead of the latest version from your project.

New Preview requests now respect the project server's cache headers, so CSS changes refresh normally.

If your browser already cached the old response, clear it once: open the affected project or folder, open the browser developer console, go to the Network tab, turn on Disable cache, and reload the page. After that one reload, the stale Preview cache is cleared and future updates should work normally.

When something goes wrong during a build, you used to see a generic "Processing failed" message with no clear way to recover. Now you get a clear explanation of what happened and a button to fix it - one click to continue, retry, or add credits.

Improvements

• Clear error messages - Every error now explains what happened in plain language, translated into all 10 supported languages
• One-click recovery - "Continue", "Try again", or "Add credits" buttons appear right where you need them. No more guessing what to do next
• AI timeout protection - If the AI gets stuck overthinking a step, it now automatically stops and lets you continue. Your progress is always saved
• Smoother long conversations - When your conversation gets summarized, you see a clear notice with a Continue button instead of a confusing pause

Some of you ran into a really nasty one: after a long building session, the AI would suddenly reply with jumbled text - bits and pieces of earlier messages, file names, error banners, all stitched together into something that made no sense. Once it started, it usually kept happening on every message in that chat, and the only thing that worked was starting a new chat.

This one was hiding deep in the plumbing and it took a while to track down. It's now fixed, and new replies in every chat will be clean from here on.

We also went back and cleared out the corrupted content from older messages across the database so it can't influence new replies. If you scroll back through an old chat you might see a message or two that looks empty where the garbled text used to be - that's the old corruption being cleaned up, not something new breaking. Your projects, files, and chat history are otherwise untouched.

We're really sorry to anyone who hit this. It was confusing, it wasted your time, and it stayed broken for longer than we'd have liked. Thank you to the people who took the time to report it in chat with screenshots - that's what made it possible to find. If anything still feels off after this update, please ping us and we'll jump on it.

We tuned Smart mode to make the AI feel more like a real designer working on your project. You should notice it right away.

• Cleaner, more polished results. Pages come out looking less like a template and more like something a designer actually thought about - better spacing, hover states, mobile layouts, the small details that make a site feel real.
• Stays focused on what you asked for. The AI is better at making the change you wanted without wandering off and rewriting things you didn't ask it to touch.
• Long building sessions go further. You can keep iterating in the same chat for longer before anything starts to slow down or get confused.
• Stops repeating itself. The AI was sometimes reading the same file two or three times in a single change. We fixed that, so changes are faster and use noticeably fewer credits.

Nothing in your workflow changes. Same chat, same buttons. Just open a project and try a redesign - you should feel the difference on the first turn.

Fixed: Voice Transcription Was Stuck on Russian

When you used the microphone in chat, your voice was being transcribed as Russian no matter what language you actually spoke - so English speakers got back garbled, transliterated text. The default was wrong and was getting saved permanently in your browser, which is why even people who never touched the setting kept seeing it.

Voice transcription now auto-detects your language from the audio. Speak English, Spanish, German, Japanese, anything - it just works. The fix runs across the whole stack, so it applies to every existing user automatically the next time you open the app.

New: Transcription Language Menu

Want to lock voice input to a specific language instead of auto-detecting? There's now a dedicated Preferences → Transcription → Language menu with all 10 supported languages plus Auto-detect. Picking a specific language can improve accuracy if you have a strong accent or if you switch between languages a lot and want voice input to stay consistent.

Heads up: as part of this fix, your chat preferences (selected AI model, custom instructions, open chat tabs) will reset to defaults the first time you open the app after this update. Your projects, files, and chat history are not affected. If you notice anything off, just refresh the page - and if something still feels wrong, ping us in chat and we'll sort it out fast.

Fixed: Multi-File Projects Now Build Correctly

When AI built websites with multiple files (HTML, CSS, JavaScript), it could sometimes produce files that didn't match each other - for example, CSS class names that didn't correspond to the HTML. This happened because the AI lost visibility of files it had just written. Now the AI always sees the full content of its recent work, so multi-file projects stay consistent.

Fixed: Conversations No Longer Get Stuck

In rare cases, if the AI took a full-page screenshot of a very tall website, the conversation could become permanently stuck - every message would fail with "Something unexpected happened." This has been fixed. Screenshots are now automatically resized to safe dimensions, and if an issue does occur, you'll get a clear message explaining what happened.

Improved: Web Research for Large Pages

The AI can now fetch content from larger websites (up to 5 MB). Pages that previously failed due to size limits - like sites with inline JavaScript - are now handled automatically. If a page is too large for direct fetching, the AI falls back to an intelligent content extraction service that pulls just the useful text.

Improved: AI Research Tasks

When the AI researches websites to gather content for your project, it now has more room to work - up to 20 steps instead of 8. This means it can thoroughly explore multi-page websites without running out of steps halfway through.

Fixed: AI Repetition Loops

In rare cases, AI models (especially Gemini Flash) could get stuck in a loop, endlessly repeating the same text. This wasted credits and could freeze your browser with a gigantic message. We've shipped a multi-layer fix:

• Automatic loop detection - The AI now detects when it starts repeating itself and stops immediately, saving your credits.
• Tuned model settings - Adjusted temperature and repetition penalty parameters to prevent loops from starting in the first place.
• Large message protection - If a response is unusually long, it's now truncated in the UI with a "Show full message" button, preventing browser freezes.

Fixed: Image Uploads in AI Chat

Over the past few days, some of you experienced issues when uploading images to AI chat - conversations would freeze or show an error message. We've been investigating this intensively and shipped multiple fixes:

• CDN compatibility fix - Google's AI servers couldn't fetch images through our CDN. We now route image URLs through a direct path that bypasses the issue entirely.
• Long conversation support - Conversations with many images (10+) would fail because of a provider limit on image URLs per request. We now intelligently manage the image budget across your conversation history, keeping the most recent images while preserving text references to older ones.

If you were affected, your existing chats with images should work again. Thank you for your patience while we tracked this down.

Improved: AI Clarifying Questions

The AI's question wizard now supports two types of questions: multiple choice (A/B/C/D) for decisions, and text input for open-ended questions like "What's your business name?" Previously, the AI was forced to turn everything into multiple choice, which sometimes caused errors or awkward options. Now it picks the right format automatically.

New

• "Undo from here" button - Every message you send to AI now shows an "Undo from here" button below it. Click it to undo all the AI's changes after that point and try again with different instructions. No more endlessly asking AI to fix its own mistakes.
• "Undo & resend" confirmation - When you undo, a clear confirmation dialog explains exactly what will happen before any changes are made. Edit your message if you want, then send it again.
• Checkpoints help page - New help article explaining how checkpoints work, when to use them, and how they relate to publishing.

Improved

• Versions renamed to Checkpoints - "Versions" was confusing for non-technical users. They're now called "Checkpoints" everywhere - menu, labels, help pages. Save a checkpoint before big changes, come back to it anytime.
• Checkpoints moved to top-level menu - Checkpoints were buried three levels deep (Menu > Project > Versions). Now they're right in the main menu, one click away. Includes a "How it works" link to the help page.
• My Projects redesigned - The "My Projects" menu item now has a distinct visual style with a project count badge, so you always know where to find your other projects.
• AI suggests undo and checkpoints - When things go wrong, the AI now suggests using "Undo from here" instead of trying to fix its own mistakes. It also recommends saving checkpoints before major changes.

Fixed

• CSS url() on published sites - Images and fonts referenced with url() in CSS (like background images) now work correctly on published sites. Previously, these paths broke after publishing.

Improvements

• Smarter file handling - The AI agent now sees file sizes before reading them and reads files in smaller chunks. No more wasting credits on loading giant files that don't fit in context.
• Faster AI responses - Improved prompt caching for Claude models reduces latency and cost on follow-up messages in long conversations.

Fixes

• Fixed AI agent loops - Resolved an issue where the AI could enter a repetitive loop after long conversations, re-reading the same files and burning credits. A new loop detection system automatically breaks the cycle.

New

• Image Attachments - Attach images directly to your chat messages. Paste screenshots with Cmd+V, pick files, or drag and drop. The AI can see and analyze your images - describe designs, debug UI issues, or use attached images in your project.
• PDF Attachments - Attach PDF documents to your chat for the AI to read and reference while building your project.
• Web Browsing - The AI can now fetch web pages and search the internet. Share a URL and it will read the content, or ask it to research topics, find documentation, or look up examples.

Improvements

• Economy Mode by Default - New chats now start in Economy mode (Gemini) for the lowest cost. Switch to Smart or Pro anytime from the mode selector.
• Smarter Chat Names - The AI automatically renames your chat to match the topic once it understands what you're building.

New

• Responsive Preview Mode - Switch between phone, tablet, and desktop views right in the preview toolbar. See exactly how your project looks at different screen sizes without leaving the editor.

Fixes

• Folder Drag & Drop Preserves Structure - Dragging a folder into your project now correctly places asset files (images, fonts, etc.) inside their original subfolders. Previously, all assets were flattened into the root directory.

Fixes

• Moving Files No Longer Deletes Images - Fixed a critical bug where moving or renaming image files (e.g., dragging to /public folder) would delete the actual image from storage. Your uploaded assets now stay safe when reorganizing your project structure.
• Upload Images to Shared Projects - You can now upload images to projects with web write access enabled. Previously, only project owners could upload assets. Storage quota is correctly tracked for the project owner.

Improvements

• Unified AI Experience - The AI now guides you from idea to published website like an agency would. It automatically asks clarifying questions to understand your vision, builds step by step, and helps you iterate. No mode switching required - just describe what you want and the AI figures out the rest.

Fixes

• AI Chat Stability - Fixed an issue where certain AI model responses could cause the chat to crash. Conversations are now more reliable across all models.

Fixes

• Usage History Now Shows Accurate Costs - Fixed an issue where the "Total spent" in Usage History showed lower amounts than what was actually deducted from your balance. The numbers now match exactly what you paid.

Fixes

• Assets Lost When Deleting Cloned Projects - Fixed a critical bug where images and other files would disappear from your original project after deleting a cloned copy. The issue occurred because cloned projects were sharing file references with the original instead of creating independent copies. Now when you clone a project or create a new version, all assets are properly duplicated to separate storage locations, ensuring each project's files are completely independent.
• Version Creation Reliability - Fixed an issue where creating a new project version could fail with a timeout error. The version creation process has been redesigned to work reliably without interrupting your current editing session.