Five things shipped last week, and the through-line is how Zero meets you where you already work. Your existing Claude or ChatGPT subscription now plugs straight in, your phone gets a friendlier handshake, video generation got a new backend, and the model lineup just grew.
Use your own Claude or ChatGPT subscription, no API key needed
If you're already paying for Claude or ChatGPT, that subscription can now power your runs in Zero. Open Settings → Model → Personal Model and you'll see two new options: Claude Code OAuth and ChatGPT (Codex). Click connect, sign in once, and you're done. No API key to copy, no billing to set up twice.
It's personal: the connection lives on your account, not the workspace. Your runs route through your subscription, but the rest of the team doesn't see it (and isn't billed for it). If you're a heavy individual user with a Claude or ChatGPT plan, this is the cheapest way to get top-tier models running inside Zero.
Here's the two-step flow, with Codex as the example. First, find the ChatGPT (Codex) card under Settings → Model → Personal Model.
Click it and Zero hands you a device code. Paste the code into OpenAI's approval page, approve access, and you're done.
Phone ships to everyone
Phone is now open to everyone, including SMS and iMessage access. To connect your mobile account, send one message from your phone to the number Zero gives you. That first text verifies the account and links it to your Zero user.
Once connected, you can talk to Zero from your phone the same way you do in Web Chat. SMS, iMessage, and Web Chat all share the same connector context and memory, so Zero keeps the same understanding of your tools, files, and work no matter where the conversation starts.
New connectors from the week
This week's additions span a few clusters:
- AI and agent platforms: Novita AI, OpenRouter, Atlas Cloud, Base44, and Parallel.
- People, sales, and business data: Clearbit, Coresignal, People Data Labs, Nyne, Hunter, BuiltWith, Clado, and Diffbot.
- Operations and workflow systems: Clerk, Brex, Checkr, Faire, Segment, SproutGigs, and WeRead.
- Location, documents, and data utilities: Google Maps Platform, Mapbox, OpenWeather, AviationStack, Mathpix, and Reducto.
None of these require a new workflow. Connect the account or API key, then ask Zero to pull context, update records, enrich leads, read docs, check weather or location data, or route model calls as part of the task you're already doing.
Built-in video generation now runs on Seedance
When you ask Zero to make a video, the work now flows to Seedance, ByteDance's video model. The interface is unchanged: you still ask Zero in natural language and get a finished clip. But the new backend gives us better motion quality and faster turnaround for the same credit cost. The switch is automatic; existing flows pick it up.
VM0 now supports GPT-5 and GPT Image 2
If you've been waiting on OpenAI's latest models inside Zero, they're here. The GPT-5 series is live. Pick GPT-5.5, 5.4, or the mini variant as your reasoning model from any chat. GPT Image 2 joins the image-generation roster alongside SeedDream and Flux Pro, and the video lineup picked up Seedance 2.0, Kling V3 4K, and Veo 3.1 Fast.
Off-product: we ran our first community labs
Outside of shipping, two firsts last week: a hands-on lab in San Jose on May 16, then a more conversational one in SF on May 20. About twenty builders showed up to each, picked one real task from their week, and we built AI teammates for it together. No coding required.
What surprised us most was the room. People running traditional businesses sitting next to people who build with AI every day, both leaving with the same thing: a working agent for a job they used to do by hand.
We'll do more of these. Keep an eye on @vm0_ai for the next dates.
From the blog
Two reads from the past week, if you want to go deeper than the release notes.
Beyond Claude Code & Codex: 10 best AI agents for non-developers in 2026
Claude Code and Codex are great if you're a developer. We rounded up the ten best AI agents for everyone else, ranked by who they actually help and what they do well.
How we built a Snowflake connector for AI agents in 6 hours
How our connector skill turned every teammate, not just engineers, into an integration author. From an empty repo to a working Snowflake connector in an afternoon.
That's the week. As always, jumping into a chat and asking "what can you do now?" is the fastest way to see this stuff in motion. More next week.
