Magpai

oh we back

If you’re in town for @sxswsydney hit me up for a free coffee and @TryMagpai demo 😉

Magpai - Your Unified AI Workspace

I spent the last two weeks imagining what an AI workspace could look like. Magpai.app has always focussed on empowering humans. Combining tech breakthroughs with traditional, trusted paradigms. But I was wrong about no-code interfaces. Here's why:

August 8th will mark 1 year of Magpai.app being in public beta. On the 8th August 2024, Magpai will leave beta and it will look very different... x.com/TryMagpai/stat…

Just added extraction prompts for files: These smart prompts persist through file updates, retaining variable name context over time. This context updates instantly across reports, presentations, workflows, and agents. 🎉

Magpai now can fetch web pages! Maybe we should make workflows like this easier to run... magpai.app/workflow/J5Ayd…

Just added a context input to the "Chat to an LLM" node. I think inputs should probably be able to convert from PDF/CSV/Markdown to Strings automatically...

Did you know Magpai can generate property signage? Magpai allows AI to actually understand your existing creative workflows. All Mapgai workflows can be used with a chatbot on Magpai, or in ChatGPT! Get started today! magpai.app/solutions/real…

Went to an absolutely cracked houdini talk on Wednesday at the Sydney Houdini User Group (HUG) @TryMagpai looks sick as hell

@JungleSilicon 👀👀

I’d like to start an AI Interface Innovation Lab to explore and discover how to create a new kind of virtual operating system. My goal is to build an OS that runs smoothly in a web browser or as a local app, allowing you to create and consume media in whatever way feels most natural. This isn’t just about creating a workspace; it’s about rethinking the internet and how we interact with it. AI is flipping the creative process on its head. Instead of starting from individual pixels or words, you can now think about high-level concepts and only refine the details if necessary. Using a computer should feel like magic. Picture editing audio as if it were a document, like in Descript, or using drawing software like tldraw to create real software. Imagine being able to re-light a scene by converting it into a 3D environment or changing its composition by moving around high-level concepts. This OS will let you transform any media into another form—extracting its essence and remixing it effortlessly. You could take data from a spreadsheet and blend it with unstructured audio or visual data. The system will help you combine these seamlessly, making it easier to create rich, complex projects. My mission is to rethink and research the fundamentals of AI interfaces, making them more intuitive, adaptive, and collaborative. By focusing on seamless multi-modal interactions and flexible systems, I aim to create an environment where creating and consuming media feels natural and integrated. I believe we can reinvent what a computer is. We moved from hyper-specific hardware and interfaces to adaptive ones, and now it’s time for software to evolve in the same way. An operating system designed from the ground up for seamless collaboration, not just with other humans but with AI as well, will make working with different types of media feel intuitive and natural. I want to redefine what a file means. A file should be a multi-modal representation of a concept, not just a static piece of data. File types should be seen as representations at a point in time, capturing the essence of an idea in various forms—text, audio, video, or 3D. I’m exploring how to make interactions feel natural. Think voice commands, gestures, and direct manipulation that make tech feel like an extension of your mind. The goal is to make the technology fade into the background so you can focus on your work. I want to rethink version control to make it lightweight and intuitive. Imagine a system where files are represented as blocks or widgets, blending the idea of file and app. Each file’s history of changes is captured, and you can see previews before merging. This system will lean into familiar folk patterns like duplicating files to make changes and then merging them back. Semantic merging will ensure that changes are meaningful and coherent, whether you’re working with text, audio, visuals, or 3D content. I’m building for messy, creative workflows. The system should adapt to fit your needs. You can mold your workspace to fit whatever project you’re working on. A couple of promising technologies are data detectors and reverse embeddings. These will enable higher-level thinking when it comes to working with media and allow for remixing features across different modalities. If you share this vision and want to get involved, let’s connect. Whether you’re an individual or part of a company, there are many ways to contribute. Support by helping with research, sharing this vision, or spreading the word. Together, we can create an OS that enhances creativity and bridges the gap between consuming and creating.

A setup in Houdini, that has been exposed to ChatGPT as a custom GPT. This runs directly on your machine, so source code/files are safe. We have versions of this for @sidefx's Houdini, and @Adobe After Effects. Reach out if you're interested in using this!

We've been working hard on making @TryMagpai useful for teams and agencies... Part of the frustration with ChatGPT is not being able to run meaningful workflows. That's always been at the heart of Magpai, but here's how that actually might look: youtu.be/hU4mkmEMLUI

@andymac3d More perfect every day! 🚀

More props to Magpai here - node graph? (Check), 2d/3d integration? (Check), Rich set of utility nodes? (Check), Integration with a variety of AI technologies? (Check). Not perfect - but alludes to the future of where generative AI tooling is heading 😉👍

GPT Vision is now on magpai.app 🎉

One of these features is using ChatGPT to run workflows! This opens up a very natural way to run one or more jobs, and have the LLM fill in the gaps. Stay tuned for more features like this 👀

This makes it super easy to work with custom workflows in production. Set up your prompts and AI Checkpoints, and then go wild!