|
|
A journey across time, space, and AI
Welcome to the October 2026 edition of Canonical’s newsletter, The Source! The year is flying by, but no year is complete without an interim Ubuntu release. Ubuntu 26.10 is out now, and you can find out all about it in our press release. October also heralds another important event: Ubuntu Summit 26.10. The Ubuntu Summit is a showcase of the best in open source, how talented teams are testing the limits of what software can do. Who knew that you could build an entirely internet-based OS? Or create a digital twin of Earth’s climate to better predict weather patterns? The latest Summit is taking place online from 12-13 November, and promises an exciting line-up of talks that will take us on a journey across time, space, and AI. So strap in and get ready to find out how open source relates to exoplanets, how we need a new way to keep time, and how you can learn to trust AI to build for you and keep your data private.
Keeping us on time, across the internet
Talk title: Time for a new way to sync time
Time synchronization is something we take for granted, but it’s easy to forget that before modern technology, people relied on bell towers and sundials. We’re far beyond that now, but it’s still surprising to learn that time synchronization over the internet has remained largely the same for the last 35 years. The legacy approach of Network Time Protocol (NTP) was not built with security in mind, and in a modern connected world, everything from Transport Layer Security (TLS) certificates to distributed databases depends on accurate, trustworthy time. Getting it wrong can break encrypted connections or corrupt data. In a world where we have connected devices needing to execute operations to the millisecond, it’s time for a new way to sync time. Ruben Nijveld (Trifecta Tech) has been working on a new kind of time synchronization daemon that can use multiple protocols, like Network Time Security (NTS) or Client Server Precision Time Protocol (CSPTP), and multiple clock devices. This new system allows for cross-checking, ensuring that drift, tampering, and inaccuracies are spotted and corrected. The result is time sync that is more secure and stable. We hope you’ll find the time to join Ruben for a deeper dive into this fascinating topic.
Hunting exoplanets with open source
Talk title: Astra: An open source observatory control software for exoplanet hunting
Trying to detect heavenly bodies that are millions of lightyears away requires careful coordination. It’s less about the stars aligning, and more about aligning hardware, software, schedule observations, and safety. However, it’s often the case that existing observatory control software is proprietary. Peter Pihlmann Pedersen (ETH Zurich) wants to change that. He’s created Astra (Automate Survey observaTory Robotised with Alpaca), an open source Python system for the sustained, fully autonomous operation of astronomical observatories. This software enables observatories to conduct pre-scheduled operations, with autofocus, pointing correction, and a catalogue of over 300 million stars. If the planet belongs to everyone, Peter’s making sure the heavens do too.
From prompt to product, with agentic CAD
Talk title: Text to CAD: How I designed a robot with LLMs and open source tools
For our next talk, we’re moving from exploring outer space to making use of the space right here on our planet. Agentic design is a new approach, where AI agents receive text-based input, turn into a visual design, and iterate on it by checking geometry, manufacturability, and refining the result in a feedback loop. Jake Fitzgerald (co-founder of Thompson Labs) built an open source library of agentic skills and tools to generate 3D models as STEP and mesh files from text and image prompts, as well as to perform manufacturability checks for fabrication services like 3D printing, sheet metal cutting, and Computer Numerical Control (CNC) machining. He then put them to the test, by designing and building a fully functional 6 degrees of freedom (6DoF) robotic arm. He’ll take us step-by-step through the library, the design process, and what he learned when agentic CAD met the real world.
Trusting AI to code reliably
Talk title: Agentic coding without the slop: 700K lines in 7 months
The term “AI slop” is pretty ubiquitous. We know that AI tools often can do great work, the difficulty is they often don’t. Building Woxi required finding ways to avoid exactly that. Woxi is an open source interpreter for the Wolfram Language, the programming language behind Mathematica and Wolfram Alpha, widely used for scientific computing, data analysis, and mathematical modelling. It was written almost entirely by coding agents. In this talk, Adrian Sieber (Software Engineer, Founder of Symiro, and CEO of Feram GmbH) talks about how these agents produced 700,000 lines of Rust to implement more than 3,000 Wolfram Language functions in Woxi, with output matching the original implementation byte-for-byte. He’ll walk through the full toolchain: prompts, coding harnesses, autonomous retry loops, as well as how he kept the AI agents on track, and prevented them from producing AI slop.
Keeping speech-to-text between you and your computer
Talk title: How local AI is transforming Ubuntu accessibility
Sticking with ideas about trust, AI offers huge potential for assistive technology, but cloud-based models come with real drawbacks: they require an internet connection, and they send personal data to remote servers. For users who depend on these tools daily, that’s not always acceptable. With the advent of local AI processing, there are privacy-conscious options on the table for users of assistive technology: but what does this look like in practice? Ubuntu 26.10 ships with Myna, a speech-to-text tool that runs entirely on local models. In this talk, Paul Geohegan (Accessibility Engineer, Canonical) will cover the behind-the-scenes work on Myna and give a broader overview of how local AI can improve accessibility across Ubuntu.
Democratizing development while preserving stability
Talk title: EPUBs, ESP32s, and everyone’s a developer now
Thanks to AI, more people can contribute code, often without requiring a background as a developer. It’s exciting, but it raises a practical question: when your contributors aren’t experienced developers, how do you keep a growing codebase stable and reliable? In his talk, Justin Mitchell (CrossPoint) talks about it from his perspective: eBook readers. CrossPoint is open source firmware for tiny ESP32 e-readers, that grew rapidly from nine downloads to over 400,000 installs and 2,000 forks in a matter of months. The majority of the community aren’t developers: they’re end users who discovered a cheap e-reader online, and wanted to make it their own. Justin will share what he’s learned about welcoming that wave of contributors through strong project leadership and safer spaces to experiment, without letting every idea land in the core firmware.
See the full schedule
Ubuntu Summit 26.10 has a packed schedule of talks, demos, and discussions over two days. You can tune in from anywhere, and dip in and out. Register on the Ubuntu Summit page, and tune into the livestream. Be sure to participate in the discussions – the chatbox is your place to get involved, share your thoughts, and ask questions. Access the schedule here. |
|
|
|
|
Canonical Group Limited | 5th floor, 3 More London Riverside, London, SE1 2AQ United Kingdom For further information on data collection, please refer to Canonical's privacy policy and privacy notice. Manage your preferences | Unsubscribe |