What's New: perf, speed, stability
Faster document streaming, AI model selection, custom MCP connections, longer conversations, and a more reliable ChatPRD experience.
Claire Vo
Originally sent on May 1, 2026 in the ChatPRD newsletter. Product details, events, and offers reflect that date; past deadlines and discount codes may have expired.
Hi Friends!
Welcome back to the ChatPRD newsletter. April was a big month. We shipped live document streaming, custom MCP connections, model selection in chat, and a ton of quality-of-life improvements that make ChatPRD faster and more reliable.
Let's jump in 🥳
✨ Faster Document Streaming
This is a big one. We've made ChatPRD much faster when writing or rewriting a document. You'll see your PRD come together in real time, which makes it much easier to course-correct early or just enjoy watching the magic happen.
🧠 Choose Your AI Model
Pro and Team users can now pick their preferred AI model right from the chat interface. Options include GPT-4.1, GPT-5.4, GPT-5.5, Claude Sonnet 4.6, and Claude Opus 4.6. Different models have different strengths; now you can choose the one that works best for the task at hand.
🔌 Custom MCP Integrations
You can now connect your own third-party MCP (Model Context Protocol) tools to ChatPRD. Just provide the server URL, select OAuth, and ChatPRD will auto-discover your OAuth endpoints and register automatically -- no more hunting for client IDs and secrets.
Your connected MCPs now sort to the top of the connectors list too, so they're always easy to find.
💬 Longer Conversations, No Lost Context
We've added automatic context compaction for chats. When conversations get really long, ChatPRD intelligently summarizes older context, preserving key decisions, requirements, and in-progress work, so you can keep going without hitting a wall.
🖥️ Responsive Editor Toolbar
The document editor toolbar now adapts to your screen. When your editor panel is narrow (sidebar open, comments panel open, etc.), secondary formatting buttons collapse into a clean overflow menu instead of wrapping to a second line.
📎 Smarter File Uploads
Uploading images and files just got smoother:
- Oversized images are automatically resized before upload — no more manual resizing
- Long filenames are auto-truncated so they don't break the UI
- Friendly, human-readable error messages when something goes wrong
⚡ Faster, More Reliable Chat
We made some serious performance investments this month:
- Thread loading is noticeably faster thanks to optimized queries and smarter document counting
- Database caching added to frequently-used queries for snappier navigation across the app
🐛 Bug Fixes
- Fixed a clipboard error when using "Copy as .md" on documents
- Language and locale preferences now apply across chats and docs
- You'll see a clear error message when inviting someone who's already on your team
- Fixed pagination when loading Confluence spaces via the integration
- Chat messages are now preserved when closing the document editor — no more lost drafts
- Added a helpful empty state and tooltip for the comments panel
- Tool approval output and loading indicators now display correctly
🎙 How I AI
This week, Jason Levin shares how he started Memelord as a $6.90-per-month newsletter sending subscribers to a Google Slides deck, grew it to $100K ARR on Bubble without hiring engineers, then raised $3M to build it into an API-first product:
In this mini episode, I break down OpenAI’s new GPT 5.5 and GPT 5.5 Pro after weeks of early testing:
Be sure to watch these other recent eps, too:
- Brian Scanlan, Sr. Principal Engineer at Intercom
- JJ Englert, Community Enablement leader at Tenex
- Yash Tekriwal, Head of Education at Clay
- Al Chen, Field Engineer at Galileo
And don't forget to check out the individual workflows.
Reminder: If you are liking the show, please give it a rating or review ⭐⭐⭐⭐⭐! It helps people find our episodes and helps us build early momentum with the show.
You can find us wherever you like to find your podcasts:
And if you have an idea for a guest, please submit a use case!
🌟 Other Updates from the ChatPRD Universe
I teamed up with Zach Davis of LaunchDarkly to deliver the first part of our cohort, the Executive Playbook for AI in Engineering, Product, and Design.
Our next class is the last weekend in June!
I built a reading practice app for my son using gpt-realtime-1.5:
I hacked a bluetooth speaker for the fun of it:
I used my platform to promote critical AI workflows:
