RobertLong

Oh, you're right! From the responses API reference:

> For function_call_output / custom_tool_call_output items. The output of the tool call, either a plain string or a

> list of input_text / input_image content parts. With the deepseek-v4-flash-vision-exp model, input_image parts in

> the output are processed as real images; with other models they are replaced with a placeholder text.

The benchmark results look promising when compared to Opus 4.8, but for agentic usecases it's lacking images as tool call result types. Giving the model a tool to take screenshots and verify its work is my main usecase for vision models, but this is more oriented towards "build a website that looks like this" type prompts. Hopefully we'll see this by release.

This ends up reading files into node.js and then running a command like grep but implemented in JS. I love the concept but isn’t this incredibly slow compared to native cli tools? Building everything in JS on top of just readFile and writeFile interfaces seems pretty limited in what you can do for performance.

I've been staying on iOS 17 and macOS 15 because I just don't see iOS/macOS 26 as an upgrade in any way. This is the first time I haven't been running a beta or getting a day one upgrade for Apple operating systems. I can't imagine moving away from iPhone or a MacBook, however if they want to avoid OS fragmentation and security issues caused by users like me refusing to upgrade, 27 better address these UX and stability issues.

All the AI labs are but Anthropic is the worst. Anyone serious about running Claude in prod is using Bedrock or Vertex. We've been pretty happy with Vertex.

Try updating your Synapse to 1.62 or higher. We added a header to files downloaded from the media repository so they can work on sites using SharedArrayBuffer. Unrelated, but you'll also want to make sure you have a TURN server set up and your media repo upload limit is set to something reasonably high like 100MB.

Ah, yeah what's happened with VRChat is a shame. I don't think we're building the same thing though. We'd like to go far beyond social 3D chat, that may be what it looks like right now, but our aspirations are towards building an open platform for publishing and hosting your own events, experiences, games, and applications. There's a much larger audience for that on desktop.

1. Thanks! I think we can play with the falloff a bit, but yeah I think so too.

2. Agreed, we erred on the side of caution for this release and made a bunch of public rooms since our per-room cap isn't super high while were still doing P2P WebRTC networking. We may dial this back to just one public room per environment.

3. It is! That first box where you select your homeserver can be typed in and you can put your own homeserver in there. Third Room is the first Matrix client using OpenID Connect which will avoid the whole issue entering your password on someone else's site, but we're still not quite at the point where just anyone can add it to their own homeserver. Hopefully very soon!

Yeah, curious why the next two months? We're definitely going to be implementing the VRM avatar format fairly soon, but not VR just yet. There are plenty of unsolved problems on desktop where there's a much larger user base. And we also still have mobile design and optimization to work on. I've worked on AltspaceVR and Mozilla Hubs and as much as I love VR, especially WebXR, it's still a relatively small market. We will support WebXR, but it's not our top priority right now.

Given some of the constraints of the immersive web currently, Matrix can provide some critical pieces of that user experience. Third Room worlds are Matrix rooms hosted across homeservers. You log in with your preferred client. You can now traverse worlds seamlessly without losing your virtual identity. This can even extend to WebXR where currently you can't traverse across origins while in the headset.

Seamless interoperability is what the "metaverse" is all about and we want to work with people who want that as well.

As one of the chairs of the Open Metaverse Interoperability Group and a member of the Metaverse Standards Forum, "metaverse" is a term with a ton of baggage and not one we really like to use internally at Matrix. However, it's also the one people know the most right now :( Multi-user interoperable virtual worlds doesn't quite flow off the tongue and immersive-web or wider-web haven't really caught on outside a developer audience. Maybe cyberspace will make a comeback?

Also, more importantly we don't believe in "a metaverse" in the same way we don't say "an internet". There's only one "metaverse". Open and interoperable should be baked into the definition. Matrix can play one part of it.