This month has been packed with 🔥 technical insights. Evan Ovadia’s talk on the hidden superpowers of linear types revealed how they help with inconsistent data, forgotten function calls, and more. Meanwhile, Ash Vardanian’s blog post, featuring Chris Lattner, Daniel Lemire, and Evan Ovadia, deep dived into SIMD parallelism, showing how modern CPUs can unleash GPU-like performance.

The Modular community has been buzzing with activity over the past month. 🐝 MAX 24.5 featured improvements to Llama 3.1 CPU performance, our biggest Mojo update ever, and more. The community has been busy building creative MAX and Mojo projects, including a terminal snake game by DobyDabaDu and a shader transpiler by Hammad Ali.

We’re excited to announce that Magic 🪄 is now the official package and virtual environment manager for MAX and Mojo projects. Going forward, we recommend you manage your MAX and Mojo installations using the magic CLI. It brings a number of improvements specific to Modular pipelines and virtual environments, and we will be adding new features for managing, developing, and deploying pipelines to future magic releases.

The Mojo Community Meeting has been renamed to the MAX + Mojo Community Meeting. We explained how we combined Mojo and MAX, and updated the license for both to ensure you can be confident building free (and open source) technologies, as well as using Mojo and MAX for your own commercial projects.

Last week we began alpha testing our new magic CLI for installing Mojo and MAX through the Conda ecosystem. It allows you to pin Mojo/MAX to different versions, and add Python dependencies so you can reliably share your Mojo and MAX projects. If you want to try it out before release, you can find the doc for installing it here.

Check out the new section for nightly updates. If you're not keeping track on the #nightly Discord channel, you can catch up on everything new for the last week in this newsletter.

The 24.4 release introduced GGUF support and quantisation for models such as Llama3 and many new features for Mojo. With that behind us Modular is now working hard towards the next release. To try out the new features press the Nightly tab on the MAX installation guide. There's an active community in the #nightly Discord channel to discuss the upcoming features with, and Modular staff are there to answer questions and respond to your feedback!

We hosted the third Mojo Community Meeting this week, with great demonstrations from the community on constraints and the Mojo HTTP serving library lightbug. We also previewed our upcoming AI pipeline serving feature with llama3, using a GUI to interact with it. And ended by answering community questions on async, and when Python will be able to call Mojo.

This week we released Mojo🔥 and MAX⚡️ 24.4, there is too much in the release to cover here! Check out the launch blog, the Mojo changelog, and the MAX changelog.

This week we held the first ever Mojo community meeting! Many topics were discussed, including priorities, diving into specific features like async, and topics set on the agenda by the wider Mojo community. Next week we'll be hearing from community members about the awesome projects they've been working on. Make sure to add the next community meeting to your calendar.