qub1t

Founder of - The new way to collaborate - for top software teams. Previously @gnosis_ & @clickup

[qub1t @_qub1t]

@vibern0 @claudeai @ElevenLabs 👀

[qub1t @_qub1t]

@claudeai Opus 5.5 one shotted us a realtime virtual office with our up-to-date workspace data - our goals, releases and tasks tracked on TVs. Agents are also in, integrated with @ElevenLabs voice.

Now WeMachines apps are even capable of running peer to peer multiplayer.

...still on Slack? 😉

[qub1t @_qub1t]

@claudeai @ElevenLabs Opus 5.5 ran at max effort for the total of 3h

[qub1t @_qub1t]

WeMachines prompt:

Build a 3d virtual office for our team that follows our brand. Kind of Meta vibe with avatars like in instagram. Give everyone in the workspace their own avatar and desk, AIs too, so we can walk around, hang out and drink coffee together and stuff. Add a meeting room and couple big TVs with our latest github and live data from the workspace. Make it absolutely beautiful, tasteful, nice lightning, masterclass - you get the idea

[qub1t @_qub1t]

@eldsjal @Neko the taste and asthetics behind Neko 🤌

[shorn @sh______rn]

i did some more demos of our built-in app builder today and it blows my mind how much WeMachines unlocks for every role. there’s a separate million $ business inside our own app 🤯

[qub1t @_qub1t]

@claudeai I almost always think of effort after writing the prompt so it would be amazing if there was an "auto" mode for it - AI reads the prompt and then classifies whether I expect the model to go in detail or ship something fast and easy

[qub1t @_qub1t]

Built with Claude Opus 5.5 - WeMachines connected to my local harness.

[qub1t @_qub1t]

WeMachines AI built us an app that studies the components and design system in our repo over time and turns them into beautiful graphics.

Every app in WeMachines is usable by agents, so anyone on our team can now ask an agent to generate an image.

Best part is... this whole extension took one prompt.

[qub1t @_qub1t]

The prompt:

Build me a press mockups app. I want to type one sentence about a feature, AI draws a marketing image: one WeMachines screen caught mid-action, a few small bits floating on its edges (toast, stat, badge, teammate cursor) and a headline next to it, on a nice gradient bg. It has to work through chat with agents. give me like 20 bg variants from our design system so I can pick and tweak, plus uploading my own pics. Ratios 16:9, 1:1, 4:5, 9:16, logo in a corner, and a "change it" box so i can say "shorter headline, make it square" and get a new version. Keep every render in a library tab, export PNG in one click so i can post it on X. Pull real copy and UI from our repo, memorize it, don't make stuff up, go

[qub1t @_qub1t]

Very exciting improvements! Try it now with your team in WeMachines 🤖

[Claude @claudeai]

Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.

It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.

[qub1t @_qub1t]

We're at @AISummitBCN - chat with us and see how WeMachines could fit your team

[shorn @sh______rn]

@WeMachinesCom is at @AISummitBCN and it’s 🔥

if you’re here too drop me a DM! 🤝

[qub1t @_qub1t]

One thing I don't like about the tools companies use today is that you can't see what the AI consumes.

So we added context controls and settings that keep your WeMachines workspace data healthy. That is better AI performance, fewer hallucinations, lower costs.

On top of that organizations get an admin layer that manages spend, limits, performance and context health in one place for all your teams.

[qub1t @_qub1t]

Thanks for sending this, and sorry: this one's on us. When our AI built Audio Server, it wrote it in Python, but apps here only run on Node, so it could never start. And the error didn't tell you why. We'll fix both: the builder will stick to Node, and a failed start will say what went wrong.

To get it running, ask the AI in the app's thread to rebuild Audio Server in Node. One limit: apps can't use WebSockets yet, so it will use regular HTTP requests. That works for uploading and playing back audio, but not for live streaming.

Separately, a tip for your next app: the model matters. On your workspace credits, Claude Opus 5 and GPT-5.6 Sol handle apps well. If you have a Claude or ChatGPT
subscription, WeMachines for desktop can run builds on it, which unlocks Claude Fable 5.1 and GPT 6 Astra, our top picks for apps.

[qub1t @_qub1t]

You can now connect all repos from your organization to WeMachines. We've also built a deep study that agents perform to better understand your infrastructure and run efficiently at scale.

Agents now ship tasks across repos, answer anyone's questions and understand your business much better.

Extra fast shipping needs proper code review, so we're integrating with the industry leader @coderabbitai - more to come!

[shorn @sh______rn]

my current work stack (plus quite a few other dev harnesses)

who knows them all?

[qub1t @_qub1t]

You can now run Claude, Codex, Grok, Hermes and Ollama locally in WeMachines and spend your personal usage. One click ships a task or builds a fullstack app into your workspace - an integration, a report, a research, a fun game, anything your team needs.

[qub1t @_qub1t]

Funny that now with Astra, I'm using Fable 5.1 for small easy things as a less capable model🫡

[qub1t @_qub1t]

A couple of small features your team should enjoy