#208 – Behind the Scenes at the CloudFest Hackathon

Transcript

[00:00:00] Nathan Wrigley: Welcome to the Jukebox Podcast from WP Tavern. My name is Nathan Wrigley.

Jukebox is a podcast which is dedicated to all things WordPress, the people, the events, the plugins, the blocks, the themes, and in this case a firsthand look at the CloudFest Hackathon.

If you’d like to subscribe to the podcast, you can do that by searching for WP Tavern in your podcast player of choice, or by going to wptavern.com/feed/podcast, and you can copy that URL into most podcast players.

If you have a topic that you’d like us to feature on the podcast, I’m keen to hear from you and hopefully get you or your idea featured on the show. Head to wptavern.com/contact/jukebox, and use the form there.

So on the podcast today, we have something different. Usually it’s me, Nathan Wrigley, chatting with a guest about something related to WordPress, whether that’s a plugin, Core updates, or perhaps an aspect of the WordPress community.

This time around, it’s me, and later on, a bunch of guests talking about an event. The event in question has already taken place, but the next iteration of it is just around the corner. And if you read the title of this episode, you’ll already know that I’m talking about CloudFest.

CloudFest is an unusual event. The most obvious indicator of this fact is that it takes place in Europa Park in Rust, Germany. It’s one of the world’s premier theme parks.

CloudFest is at its heart, a tech conference, but every year, just before the main CloudFest conference begins, a very different event takes place. It’s called the CloudFest Hackathon. So whilst the rollercoasters are testing the laws of physics outside, inside a group of developers, UX designers and system architects are testing the limits of the modern internet.

Dozens of the world’s most talented engineers strip away the corporate sales pitches and set themselves a variety of collaborative challenges to be completed in just three days. Now we see hackathons all the time. Usually they’re sponsored by a single company trying to get people to use their specific API, or their high pressure competitions, to build a disruptive startup in 48 hours. But the CloudFest Hackathon isn’t like this. It’s professional, it’s non-commercial, and its primary intention isn’t necessarily to build a product, it’s to maintain the ecosystem.

So let’s hear from somebody who knows all about the CloudFest Hackathon, and that person, is Carole Olinger.

[00:03:04] Carole Olinger: My name is Carole, and I am the head of CloudFest Hackathon. I’m very excited about my role here and to be able to connect so many awesome people and talent around the world.

So I think there are multiple definitions for a hackathon. In this case I would probably define it as a gathering of open source enthusiasts who are going to be working and coding and designing a lot of exciting projects together. They haven’t met before in many cases, and they are put in the same room for three days being fed, being caffeinated and trying to improve the open web.

[00:03:42] Nathan Wrigley: Who’s on the organizing team?

[00:03:44] Carole Olinger: So basically, I am leading the whole operation in my role for CloudFest and the World Hosting Days. And I have the most amazing supporting team around me that anyone could ever imagine. So it is Lucas Ratke from Automattic, Alain Schlesser from Yoast, and Thierry Muller from Google, who are on the project support team, and making sure that we have all these really valuable projects in our event. And that project leads are prepared in the best possible way.

And for the first time we also have a volunteer that is helping during the event. And is specifically helping me wrangling the 110 amazing sheep around me, and to make sure that there are accommodations are covered. That all the catering is being done. And that is a Simon Kraft from Group One.

[00:04:36] Nathan Wrigley: I show up to an event like this, all the jigsaw pieces are in place you think, oh, it just happens, but of course it doesn’t just happen. How long have you spent working on this event? How long have you been wrangling this whole thing into existence?

[00:04:49] Carole Olinger: Usually we start in September. And then it’s more okay, what are our objectives? What are our goals for this edition? We are really trying to take as much feedback as possible from previous year’s attendees, to make sure that we have improvements in place and new additions to the event for the following edition. So that happens in September. Creating the team, making sure that we have specific objectives and goals and those are manageable.

And then the actual work starts in October, and then becoming more and more intense over the upcoming months. And I would say January is probably the most crazy month. I barely slept.

[00:05:31] Nathan Wrigley: So the idea really is that you put. In this case, 10 projects in a room. You’ve got 10 project leads and, in some cases there’s multiple people leading a project. And then you add into that mix over a hundred people, many of whom appear to be developers, and you stir that pot up a bit, and hope that things come out the end that are useful, that have been enjoyable to work on.

How do you decide what the 10, in this case, projects were? And are you oversubscribed with people wishing to be a part of it? And so how do you decide what makes it? How do you decide which projects are interesting to CloudFest Hackathon each year?

[00:06:10] Carole Olinger: So this has been evolving over the years. So I remember additions, three or four years ago, or previous to the pandemic. Where our project team was pulling projects out of the different CMS communities, open source project communities. So we had ideas about what we wanted to tackle, and some projects came out of the communities. So we were like hunting ideas, and also planting ideas inside communities.

This year is the first year where we didn’t have to do any of that. We had 22 pitches from different CMS communities and other open source projects that were pitching their ideas to us. So it was like a kind of a hard choice to determine which ones are going to make it.

So usually we are trying to take into consideration what the theme of the main event CloudFest is, and obviously as everyone is excited about AI these days, that is something we wanted to cover. So we made sure we had some projects that had AI involvement. And then what is really important to my heart, and to the team’s heart, is that we are having cross CMS collaborations.

So we are trying to have WordPress people here, which is obviously the community that I am mostly connected with. But also TYPO3. TYPO3 is one of our, the W3 Association is one of our top level sponsors. We since years we have Joomla people, Drupal people joining us. We are trying to find a good mix to empower those cross CMS collaborations and also cross-project collaborations. So even within one platform, just to name WordPress as an example, we try to make sure that we have projects that could eventually benefit from each other.

[00:07:54] Nathan Wrigley: I’m guessing also, there’s a component of trying to work out projects that if you put a hundred people in a room, there’s not a hundred replicas of the same person. Each of them are different. That there are 10 different places where they can land. Because one thing that I didn’t realize and was really curious to me, is when the Hackathon started, apart from the project leads, nobody’s assigned a place to go. They listen to a little speech at the beginning. It’s like a promotional thing. Two minutes, this is what we want to do, the pitch. And then the people make a decision. And for 10 minutes or something, there’s this sort of chaotic moving of people around, and then it all settles down.

So presumably you have a wide array of project pitches, so that those a hundred plus people can decide, they’re not all surrounding the one table and there’s a table over there that’s empty, I guess that fits in the jigsaw as well.

[00:08:44] Carole Olinger: So we are taking very much care about the selection process of applicants. So when we know what our projects are going to look like, we are trying to match their needs in terms of skills that attendees are going to present to us during their application.

So usually we have between 300 and 400 applications for the Hackathon, and we have 110 slots. But, and this is important to understand, our partners are bringing team members within these 110 attendees. And our partners this year have been super actively involved, which I love. So they were not only giving us money to make this event possible, to be able to invite open source contributors to this place, including their hotel, accommodation and food. They were bringing, people resources. I hate the term, but you know what I mean. So they were sending their crew to lead, to participate in these open source ideas and projects. So in the end we had 60 available spots for open source contributors. And then we made sure that we are matching the skills that they were sharing with us in their applications with the needs that the projects will have on the table.

So we have a pretty good understanding already about who’s going to be at what table. And obviously we are monitoring that. So we give them some time to make the decisions. And if we see that there are skills missing at a certain table, or if there’s another table that is going to be too full and too complicated to manage by the project lead, we are kindly convincing, and reassigning people, to participate in different projects.

[00:10:19] Nathan Wrigley: So that’s interesting. So it’s not just a free for all? The idea is to maximize the output of the projects, the 10 different projects at the end, and you will, like you said, politely, ask people to move over if you believe that the thing that they have said that they’re good at, is, matching. And there’s a, I don’t know, a hole in one particular project.

That brings me to this question then. Is the intention that these projects have a life after this event has finished? Or is it just a case of, okay, we’ve all had a nice time, the event has closed, let’s all move on with our lives.

[00:10:48] Carole Olinger: This is becoming more and more important to us. As I said earlier, we are trying to improve something every year like, like focusing on something when we are fixing our goals that we can do better every year. And what we can definitely do better is spreading the word about what amazing achievements the teams have been building during the event, and make sure that this project’s become more sustainable. So that the world knows that there’s potential in the outcomes of CloudFest Hackathon project, and to potentially unlock support and resources for these projects to continue.

I would love to spread the word, making it possible to unlock these resources. And then also inside our team, building more and more resources to follow up with this project leads from our end.

[00:11:33] Nathan Wrigley: There’s an element of, how to describe it? There’s this time pressure in the whole event. So that the thing is basically three days long, from inception until final judging, three days. So the pressure is on, and I can feel at the moment the pressure is increasing slightly. You can sense that people are getting quite into the project they’re working on.

I noticed last night, long after the event had officially closed down, there was quite a lot of people still sitting there. They’re were obviously wedded to what they’re doing. There is this sort of like Shark Tank element where there’s going to be a judgment at the end and somebody’s going to win.

How does that work? Who gets to decide who’s the winner?

[00:12:06] Carole Olinger: So we do have a jury, and the jury is composed of representatives of our top tier partners. And they send one representative to the jury. Then we do have one representatives from the Groundbreaker Talents charity project. Because, on a side note, all these awards are being sponsored by companies, and everything that we are collecting in terms of sponsorships is going to the Groundbreak Talents initiative. And then we have the project support team, and myself being on the juries. And it’s an uneven number. So we have nine people, which is always good to have on a jury. And after the presentation of results on the last day of the hackathon, the jury is going to deliberate.

And then we are going to listen to the project support team, who has been working the room and connecting with the project tables during the three days in terms of technical achievements, challenges they have seen. So they’re going to give us some impression on that. And that is mostly important for the Tech Visionary Award. And then, all of us have had the chance to obviously see the presentations, which is important for the Pitch Perfect Award. Who has the most appealing presentation of results? We do have the Social Media Master Award, that is fully being tracked.

So Simon and I, we are going to give the jury some insights on who has created the most boss on social media. . And, then we do have the Breaking Barriers Award, which is a new one. So this is about using inclusive technologies, and getting some outputs that are going to be helpful for a diverse set of users, and connecting people on the user base, but also how the people have been working together in terms of having diverse skills and perspectives on the table.

So these are some of the awards, and there’s going to be an overall winner. We have five categories, and an overall winner. And the overall winner is, the one that has the most points.

Thanks to Carole for that comprehensive introduction to the CloudFest Hackathon. Now, let’s look at the why. Why do people travel from 30 plus different countries around the world to do all of this?

In our industry, we talk a lot about the cloud, but we often forget that the cloud is just a massive collection of interconnected open source projects. You have WordPress powering 40 plus percent of the web, you have the Linux kernel, you’ve got PHP and Python communities, and then you have the hosting providers and hardware manufacturers. Normally, these groups live in silos. They communicate via GitHub issues or formal API documentation. Well, the intention of the hackathon is to create what might be called the human API. It’s about taking a person who might maintain a security plugin, and sitting them at the same table as an engineer who manages millions of servers for a global host.

When you remove the barrier of the screen, the friction of the internet disappears. Problems that have been sitting in a backlog for six months get solved over a coffee, or a shared meal because the right people are finally in the same physical space.

Although, as Carole mentioned, there is a winner, this isn’t really about winning a prize. In fact, the prizes are almost secondary to the real goal, which is contributing back to open source projects, some of which already exist, some of which are new. The intention is all free and open source software or FOSS for short.

These contributors aren’t there to build something proprietary and closed. They’re there to ensure the plumbing of the internet stays robust, secure, and interoperable. Oh, and to have some fun collaborating at the same time.

Speaking of contributors, let’s hear from some of them now and get a little taste of what their project was all about.

The room as you will hear was a little noisy.

[00:15:52] Javier Casares: I am Javier Casares and I am one of the co-leads from for the CMS Cloud Manager Project.

[00:15:58] Nathan Wrigley: What does this project hope to achieve?

[00:16:00] Javier Casares: Usually when you have a cPanel or Plesk or some kind of panel, you can install a WordPress, for example, with one click, but the server is not configured.

So in this project, we want to configure not only the CMS, but also the server where the CMS is going to be

[00:16:20] Nathan Wrigley: And how’s it going so far?

[00:16:22] Javier Casares: It’s fine. We have the public part because we want to have a website so you can configure things and prepare everything. And then we have this software, the real software that creates everything. And more or less it’s, fine, at this moment. So I think we can achieve everything for the hackathon, for the finals.

[00:16:46] Mattias Pfefferle: I’m Mathias. I am working on Activity Pub and the Fediverse.

[00:16:51] Nathan Wrigley: And what’s the project that you’re working on at CloudFest, the Hackathon?

[00:16:55] Mattias Pfefferle: We are working on federated events. So it’s very much a special case of the Fediverse.

[00:17:02] Nathan Wrigley: And what is the intention of the project? What are you hoping to get out of it?

[00:17:05] Mattias Pfefferle: We try to build a decentralized, alternative to the big social networks around events, so that people does not have to rely on something they do not have control over. So we would hope to get an alternative to meetup.com, maybe, or any other big closed proprietary social network, around events.

[00:17:29] Nathan Wrigley: And how’s it going so far?

[00:17:30] Mattias Pfefferle: It’s mixed, because even if it’s a standard, there are different variations of using the standard. So we filed a lot of bug reports, and tried to work on a standard that better describes the standard , if that makes sense? And we’re trying to make federation happen using WordPress and some other platforms that are built by people that are part of the Hackathon team.

[00:17:58] Milana Cap: My name is Milana Cap, and I’m on a project WPCLI as MCP. MCP stands for Milana Cap pro. No, it doesn’t

[00:18:09] Nathan Wrigley: What is the intention of this project then?

[00:18:11] Milana Cap: We are introducing AI into WPCLI. So, you could use AI in different aspects of WordPress, like content creation and all of that. But it was missing in development process, especially in local instances. So now we have that, and it’s actually a lot of fun. Much more fun than I thought.

[00:18:32] Nathan Wrigley: And how’s it going so far?

[00:18:34] Milana Cap: So far we build a spam machine, and that’s official name. And we actually had a MVP on first day. It’s really fun. Yeah. And we are just learning how this AI is behaving by itself in our locals.

[00:18:52] Patricia BT: Hello, Nathan. I’m Patricia BT. I’m living in Geneva, Switzerland. I speak French, and I am my own boss, . And I came, with that pitch, as a project for the Hackathon because for me it’s very important to own your data and preserve what exists on the web, and not lose anything.

[00:19:11] Nathan Wrigley: And what’s the name of your project?

[00:19:13] Patricia BT: CMS Freedom.

[00:19:14] Nathan Wrigley: And how’s it going so far?

[00:19:16] Patricia BT: It’s going very well. We have, tech people here, engineer, who are doing amazing work with, especially LLMs. So we are using AI to grab any HTML content and discover the format, the elements, and then later be able to import that into WordPress block theme.

For now it’s WordPress block theme and content. And later the hope is that people from other CMSs, other system, can just modify that last bit and import what the tool extracts, and import to their own system. So we can move from any HTML, render any page on the web and create that for your CMS.

[00:19:59] Nemanja Cimbaljevic: Hi, I am Nemanja. I come from Serbia and currently I’m with GoDaddy as a software engineer.

[00:20:05] Nathan Wrigley: And what’s the name of your project?

[00:20:07] Nemanja Cimbaljevic: I will not break my tongue. We will call it AI Accessibility Content Updater.

[00:20:12] Nathan Wrigley: And what is the intention? What are you hoping to achieve over these three days?

[00:20:16] Nemanja Cimbaljevic: We will try to make a proof of concept that will allow us to move on in the future where the AI is capable to help with accessibility of the websites that can be improved? Not all of them.

[00:20:29] Nathan Wrigley: Are there any constraints around what it is that you are hoping to be able to do? Or is it literally all the accessibility?

[00:20:35] Nemanja Cimbaljevic: It’s all about accessibility, and yes, we will see where we will go. It was announced as a trial and error. So we will see if there is any trial or just error.

[00:20:48] Anne-Mieke Bovelett: Hi, I’m Anne Bovelett.

[00:20:50] Nathan Wrigley: And what’s the name of the project you’ve got at CloudFest Hackathon?

[00:20:54] Anne-Mieke Bovelett: It’s called Accessible Infographics.

[00:20:56] Nathan Wrigley: And what is the intention? What are you hoping to achieve in these three days?

[00:21:01] Anne-Mieke Bovelett: Right now we’re creating a plugin for WordPress. . And when you have infographics on your site because you produce medicine, or machines, or you have statistics on your site. You can use our WordPress block, and upload an image through that, and then it will help you to make it accessible by creating extra information under the hood.

And the best thing of this is it’s not just going to be for WordPress, because we’re conceptualizing that for others so they can easily recreate this in other open source CMSs. And it will save millions and millions of people from sitting in the dark with very important information on websites.

What it actually means is that, also when we manage to move on with this project in the next phase, we’re gonna try and do this in bulk. To do it backwards for companies that already have a lot of infographics on their website, and understand that they have to do it either by law, or because they’re smart and want higher converting web shops, for example.

And then the possibility will come that they can do that backwards in bulk, and it will save them thousands and thousands in money that they have to invest in making this happen.

[00:22:20] Wesley Stessens: Yes. My name is Wesley Stessens, and I’m from Belgium, and we work on the Peer-to-peer Federated RAG Framework with, our team.

[00:22:30] Nathan Wrigley: And can I ask, what is the intention of that project over the three days? What are you hoping to achieve?

[00:22:35] Wesley Stessens: We are hoping to achieve something that hasn’t been done before in the, in the RAG space.

So basically RAG, or Retrieval Augmented Generation is way how you can augment an LLM and AI with extra data. And we want to allow everyone to create their own databases. And anyone can just join our network with their own knowledge.

For example, someone who knows a lot about beers, they can join our network and have like a library full of all information about very specific niche beers, maybe beers they brew themselves or whatever. And then any other node in the network can ask a specific question. And then our purpose is to route that question to the best matching node in a decentralized way. So there’s no servers or big companies. In between everything is done in a peer to peer way. So they get back the best matching documents from other people’s libraries, so to speak.

And then we use that to ask an LLM, or an AI a question with the context that we got from other people’s databases that matched best. And now we show the results to the user, or we create like a chat interface maybe around that. That’s the end result that we hope to achieve.

[00:23:49] Tadas Pukas: I’m Tadas.

[00:23:50] Nathan Wrigley: And what is the name of the project?

[00:23:52] Tadas Pukas: It’s WordPress Staging Environment Manager. It’s a bit complex to understand, but it does very simple thing.

[00:23:59] Nathan Wrigley: And what’s the intention at the end of these three days? What would you ideally like to be shipping? Have finished?

[00:24:05] Tadas Pukas: Yeah, so we want to have open source plugin, and actually we have it almost, so it’s the final touches.

And this will be distributed. It’s already in the public GitHub repo. So people will be able to download zip file, install a plugin, and create staging environments. Not only create but sync changes from staging to live. Actually, our name of the plugin is Staging to Live, so it’s, yeah, almost done.

Almost ready.

[00:24:29] Nathan Wrigley: Okay. I hope that you enjoyed this different style of podcast. Hopefully you learned something about CloudFest and the CloudFest Hackathon.

You certainly got to hear from a wide variety of contributors, and got to peel back the curtain about what a hackathon is, and the different projects people work on. There’s a great energy at events like this, and maybe this will convince you to explore hackathons in the future.

You don’t need to be a coder. Each project needs a wide array of talents from coders to marketers, designers, to project wranglers.

Like I said, at the top of the show, CloudFest 2026 is just around the corner. There’s an annual event both in the US and the one discussed here in Germany.

If you’re interested in finding out more, you can find all of the links in the show notes by heading to wptavern.com/podcast.

And we’ll be back next week with more from CloudFest and the CloudFest Hackathon.

On the podcast today we have something different.

Usually it’s me, Nathan Wrigley, chatting with a guest about something related to WordPress, whether that’s a plugin, Core updates or perhaps an aspect of the WordPress community. 

This time around it’s me, and later on a bunch of guests talking about an event. The event in question has already taken place, but the next iteration of it is just around the corner, and if you read the title of the episode, you’ll already know that I’m talking about CloudFest.

CloudFest is an unusual event; the most obvious indicator of this is the fact that it takes place in Europa-Park in Rust, Germany. It’s one of the world’s premier theme parks.

CloudFest is at its heart a tech conference, but every year, just before the main CloudFest conference begins, a very different event takes place. It’s called the CloudFest Hackathon. So, whilst the roller coasters are testing the laws of physics outside, inside, a group of developers, UX designers, and system architects are testing the limits of the modern internet.

Dozens of the world’s most talented engineers, strip away the corporate sales pitches, and set themselves a variety of collaborative challenges, to be completed in just three days.

Now, we see “hackathons” all the time. Usually, they’re sponsored by a single company trying to get people to use their specific API, or they’re high-pressure competitions to build a “disruptive” startup in 48 hours. But the CloudFest Hackathon is not like this. It’s professional, it’s non-commercial, and its primary intention isn’t to build a product, it’s to maintain the ecosystem.

Today we’re going to be hearing from a variety of people who were involved in the 2025 event. The 2026 event is just around the corner.

You’ll hear from:

  • Carole Olinger (the Hackathon lead)
  • Javier Casares
  • Mattias Pfefferle
  • Milana Cap
  • Patricia BT
  • Nemanja Cimbaljevic
  • Anne-Mieke Bovelett
  • Wesley Stessens
  • Tadas Pukas

They’re a tiny sample of who was present at the event, but hopefully they will give you a flavour of what the CloudFest Hackathon is, why people attend, and what kinds of projects they’re involved in.

Useful links

CloudFest Hackathon 2025 Recap

CloudFest

CloudFest Hackathon

WordPress.org blog: WordPress 6.9.3 and 7.0 beta 4

WordPress 6.9.2 was released earlier today and addressed 10 security issues.

A few users have subsequently reported an issue where the front end of their site was appearing blank after updating to 6.9.2. The issue has been narrowed down to some themes using an unusual approach to loading template files via “stringable objects” instead of primitive strings for file paths.

Although this is is not an officially supported approach to loading template files in WordPress (the template_include filter only accepts a string), it nevertheless caused some sites to break. As a result, the Security Team has decided to address this in a fast follow 6.9.3 release.

As always, it is recommended that you update your sites to the latest version of WordPress immediately. This ensures your site is protected by all available security fixes in 6.9.2 and that you will not be affected by the bug fixed in 6.9.3.

Many thanks to those who reported the issue, assisted in narrowing down the problem, and helped with the fix.

You can download WordPress 6.9.3 from WordPress.org, or visit your WordPress Dashboard, click “Updates”, and then click “Update Now”. If you have sites that support automatic background updates, the update process will begin shortly. You don’t have to do a thing!

For more information on WordPress 6.9.3, please visit the version page on the HelpHub site.

WordPress 7.0 beta 4

The next major release of WordPress will be version 7.0, which is planned for April 9, 2026. The Security Team has decided to package a new beta release (7.0 beta 4) to keep everyone protected from the patched vulnerabilities, including the dedicated members of the community focusing their time and effort on testing the upcoming release.

This will be an additional beta release in the 7.0 release cycle. The schedule will remain the same going forward, but with five total beta releases instead of the previously planned four. The next 7.0 beta release is still scheduled for Thursday, March 12th.

This beta version of the WordPress software is still under development. Please do not install, run, or test WordPress 7.0 beta versions on production or mission-critical websites. Instead, you should evaluate Beta 4 on a test server and site.

Plugin Install and activate the WordPress Beta Tester plugin on a WordPress install. (Select the “Bleeding edge” channel and “Beta/RC Only” stream.)
Direct Download Download the Beta 4 version (zip) and install it on a WordPress website.
Command Line Use this WP-CLI command:
wp core update --version=7.0-beta4
WordPress Playground Use the WordPress Playground instance to test the software directly in your browser. No setup is required – just click and go!

Beta 4 updates and highlights

WordPress 7.0 Beta 4 contains the ten security patches shipped in WordPress 6.9.2, and more than 49 updates and fixes since the Beta 3 release, including 14 in the Editor and 35 in Core. 

Each beta cycle focuses on bug fixes. More are on the way, thanks to your help with testing. You can browse the technical details for all issues addressed since Beta 3 at these links:

As always, a successful release depends on your confirmation during testing. So please download and test!

Props @peterwilson, @desrosj, @marybaum, @amykamala for peer reviewing.

As Open Models Spark AI Boom, NVIDIA Jetson Brings It to Life at the Edge

The Cat 306 CR mini-excavator weighs just under eight tons and fits inside a standard shipping container. It’s the machine a contractor rents when the job site is tight: a utility trench near a foundation, a basement dig in a dense neighborhood.

The cab is roughly the size of a phone booth. The operator sits close to the controls, two joysticks, multiple functions per hand. It takes time to learn. It takes longer to speed up.

At CES earlier this year, that machine answered questions.

In the demo, the Cat AI Assistant ran on NVIDIA Jetson Thor, an edge AI platform built for real‑time inference in industrial and robotic systems, NVIDIA Nemotron speech models are used for fast and accurate natural voice interactions. Qwen3 4B, served locally via vLLM, interprets requests and generates responses with low latency, no cloud link required.

Beyond enterprise innovation, open models unlock new possibilities for developers to build and experiment freely. Running OpenClaw on NVIDIA Jetson enables developers to create private, always-on AI assistants at the edge — with zero application programming interface cost and full data privacy.

All Jetson developer kits support OpenClaw, offering the flexibility to switch across open models from 2 billion parameters to 30 billion. With a frontier-class AI assistant running locally, users can power morning briefings, automate daily tasks, perform code reviews and control smart home systems — all in real time.

From the Cloud to the Edge

For most of their recent history, open models lived where it was easiest to support them. 

They ran in data centers, backed by elastic compute and persistent networks. Cloud deployments carry costs in latency and ongoing compute spend that scale with every query.

Physical systems optimize for something else. Low latency because machines interact with people and environments. Limited power because devices have hard limits. And consistent behavior because variability introduces risk.

There’s also a supply question. Memory shortages have driven up costs across the industry. Jetson brings compute and memory together in a system-on-module, accelerating customer hardware design and making sourcing and validation easier than with discrete component approaches.

And as models have grown more efficient, developers have also started asking a different question. Not which model performs best in isolation, but where it makes sense to run. 

More often, the answer is on the device, starting from Jetson Orin Nano 8GB for entry-level generative AI models.  

Building Autonomous Physical AI Systems at Scale

For physical AI systems, generative AI models are expanding what’s possible. 

Caterpillar’s in-cab Cat AI Assistant, which is in development, runs speech and language models locally alongside trusted machine context, supporting operator guidance and safety features.

At CES, Franka Robotics showed what that looks like in robotics. The company’s FR3 Duo dual-arm system ran the NVIDIA GR00T N1.6 model end-to-end onboard, perception to motion, no task scripting. The policy executes locally.

In robotics research, the SONIC project from NVIDIA’s GEAR Lab trains a humanoid controller on over 100 million frames of motion-capture data, then deploys the resulting policy on a physical robot where the kinematic planner runs on Jetson Orin at around 12 milliseconds per pass. The policy loop runs at 50 Hz. Everything executes onboard.

The pattern reaches into the developer community. A team from UIUC’s SIGRobotics club built a dual-arm matcha-making robot on Jetson Thor running the GR00T N1.5 model. It took first place at an NVIDIA embodied AI hackathon.

This research momentum continues at the NYU Center for Robotics and Embodied Intelligence. The group recently ran its YOR robot on Jetson Thor, using NVIDIA Blackwell compute to handle the heavy processing required for AI-driven movement. Early results show YOR performing intricate pick-and-place tasks with better generalization to new objects and robustness to scene variation, accelerating readiness for a wide range of household tasks like cooking and laundry.

Independent researchers are finding the same. Andrés Marafioti, a multimodal research lead at Hugging Face, built an agentic AI system on Jetson AGX Orin that routes tasks across models and schedules its own work. Late one night, the agent sent him a message: Go to sleep. Everything will be ready by morning.

Developer Ajeet Singh Raina from the Collabnix community has shown how to run OpenClaw on NVIDIA Jetson Thor for a personal AI assistant that runs 24/7. This setup allows for private large language model inference for the user’s own data while the system manages emails and calendars through a local gateway.

Jetson Is the New Standard

NVIDIA Jetson has become a common platform for running open models at the edge.

It supports a wide range of open models and AI frameworks, giving developers flexibility for almost any generative AI workload at the edge. 

Model benchmarks are available at Jetson AI Lab, along with tutorials from the open model community. Jetson Thor delivers leadership inference performance across all major generative AI models.

Gemma: Built on Google’s Gemini research, Gemma 3 is a versatile workhorse for Jetson. It is multimodal out of the box, which means it can see and talk in over 140 languages. On Jetson Thor, it handles a massive 128K context window. This makes it perfect for robots that need to remember a long list of complex or multistep instructions.

gpt-oss-20B: This model from OpenAI lowers the barrier to deploying advanced AI by delivering near state-of-the-art reasoning performance in a model that can run locally on Jetson Thor and Orin for cost-efficient inference. 

Mistral AI: The new Mistral 3 open model family delivers industry-leading accuracy, efficiency and customization capabilities for developers and enterprises. This family includes small, dense models ranging from 3B to 14B, fast and remarkably smart for their size. Jetson developers can use the vLLM container on NVIDIA Jetson Thor to achieve 52 tokens per second for single concurrency, with scaling up to 273 tokens per second with concurrency of eight.

NVIDIA Cosmos: This leading, open, reasoning vision language model enables robots and AI agents to see, understand and act in the physical world like humans. Both the 8B and 2B models run on Jetson to deliver advanced spatial-temporal perception and reasoning capabilities. 

NVIDIA Isaac GR00T N1.6 is an open vision language action model (VLA) for generalist robot skills. Developers can use it to build robots that perceive their environment, reason about instructions and act across a wide range of tasks, environments and embodiments. On Jetson Thor, the full GR00T N1.6 pipeline executes onboard, delivering real-time perception, spatial awareness and responsive action.

NVIDIA Nemotron: A family of open models, datasets and technologies that empower users to build efficient, accurate and specialized agentic AI systems. It’s designed for advanced reasoning, coding, visual understanding, agentic tasks, safety, speech and information. The Nemotron 3 Nano 9B model effectively runs on Jetson Orin Nano Super with llama.cpp with 9 tokens per second performance. 

PI 0.5: A VLA model from Physical Intelligence that enables robots to understand instructions and autonomously execute complex real-world tasks with strong generalization and real-time adaptability, while NVIDIA Jetson Thor delivers 120 action tokens per second to power responsive, low-latency physical AI deployment.

Qwen 3.5: This family of models from Alibaba, including the latest Qwen 3.5 releases, offers a mix of dense and mixture‑of‑experts models that deliver strong reasoning, coding multimodal understanding and long‑context performance. Jetson Thor delivers optimized performance across Qwen models like the Qwen 3.5-35B-A3B model, which reasons at 35 tokens per second, making real-time interactivity possible. 

Any developer can fine-tune these models to create specialized physical AI agents and seamlessly deploy them into physical AI systems. The NVIDIA Jetson platform supports popular AI frameworks from NVIDIA TRT, Llama.cpp, Ollama, vLLM, SGLang and more.

Take On Open Models on Jetson

Developers can dive into Hugging Face tutorials — including Deploying Open Source Vision Language Models on Jetson — and catch the latest livestream. Learn from this tutorial and run OpenClaw on NVIDIA Jetson.

Join GTC 2026 next month to see it all in action. NVIDIA will show how open models are moving from data centers into machines operating in the physical world, including in a  panel on the Future of Industrial Autonomy.

Watch the GTC keynote from NVIDIA founder and CEO Jensen Huang and explore physical AI, robotics and vision AI sessions.

WordPress.org blog: WordPress 6.9.2 Release

WordPress 6.9.2 is now available!

This is a security release that features several fixes.

Because this is a security release, it is recommended that you update your sites immediately.

You can download WordPress 6.9.2 from WordPress.org, or visit your WordPress Dashboard, click “Updates”, and then click “Update Now”. If you have sites that support automatic background updates, the update process will begin automatically.

The next major release will be version 7.0, which is planned for April 9th, 2026.

For more information on WordPress 6.9.2, please visit the version page on the HelpHub site.

Security updates included in this release

The security team would like to thank the following people for responsibly reporting vulnerabilities, and allowing them to be fixed in this release:

  • A Blind SSRF issue reported by sibwtf, and subsequently by several other researchers while the fix was being worked on
  • A PoP-chain weakness in the HTML API and Block Registry reported by Phat RiO
  • A regex DoS weakness in numeric character references reported by Dennis Snell of the WordPress Security Team
  • A stored XSS in nav menus reported by Phill Savage
  • An AJAX query-attachments authorization bypass reported by Vitaly Simonovich
  • A stored XSS via the data-wp-bind directive reported by kaminuma
  • An XSS that allows overridding client-side templates in the admin area reported by Asaf Mozes
  • A PclZip path traversal issue reported independently by Francesco Carlucci and kaminuma
  • An authorization bypass on the Notes feature reported by kaminuma
  • An XXE in the external getID3 library reported by Youssef Achtatal

The WordPress security team have worked with the maintainer of the external getID3 library, James Heinrich, to coordinate a fix to getID3. A new version of getID3 is available here.

As a courtesy, these fixes are being backported, where necessary, to all branches eligible to receive security fixes (currently through 4.7). As a reminder, only the most recent version of WordPress is actively supported. The backports are in progress and will ship as they become ready.

Thank you to these WordPress contributors

This release was led by John Blackbourn. In addition to the security researchers mentioned above, WordPress 6.9.2 would not have been possible without the contributions of the following people: Dennis Snell, Alex Concha, Jon Surrell, Isabel Brison, Peter Wilson, Jonathan Desrosiers, Jb Audras, Luis Herranz, Aaron Jorbin, Weston Ruter, and Dominik Schilling.

NVIDIA Virtualizes Game Development With RTX PRO Server

Game development teams are working across larger worlds, more complex pipelines and more distributed teams than ever. At the same time, many studios still rely on fixed, desk-bound GPU hardware for critical production work.

At the Game Developers Conference (GDC) this week in San Francisco, NVIDIA is showcasing a new approach to bring together disparate workflows using virtualized game development on NVIDIA RTX PRO Servers, powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs and NVIDIA vGPU software.

With the RTX PRO Server, studios can centralize and virtualize core workflows across creative, engineering, AI research and quality assurance (QA) — all on shared GPU infrastructure in the data center. 

This enables teams to maintain the responsiveness and visual fidelity they expect from workstation-class systems while improving infrastructure utilization, scalability, data security and operational consistency across teams and locations.

Simplifying Complex Workflows

As game development studios scale, hardware can often sit underutilized in one location while other teams wait to access it for production work. QA capacity is hard to expand quickly. Over time, workstation hardware, drivers and tools diverge, making bugs harder to reproduce. AI workloads are often isolated on separate infrastructure, creating more operational overhead. 

The NVIDIA RTX PRO Server helps studios move from workstation-by-workstation scaling to centralized GPU infrastructure. Studios can pool resources, allocate performance by workload and support parallel development, testing and AI workflows without expanding physical workstation sprawl.

Centralized GPU infrastructure enables studios to run AI training, simulation and game automation workloads overnight, then dynamically reallocate the same resources to interactive development during the day, improving overall utilization and reducing idle capacity.

The NVIDIA RTX PRO Server supports virtualized workflows for 3D graphics and AI across the game development lifecycle for:

  • Artists: Providing virtual RTX workstations for traditional 3D and generative AI content-creation workflows.
  • Developers: Powering consistent, high-performance engineering environments for coding and 3D development.
  • AI researchers: Offering large-memory GPU profiles for fine-tuning, inference and AI agents.
  • QA teams: Enabling scalable game validation and performance testing using the same NVIDIA Blackwell architecture used by GeForce RTX 50 Series GPUs.

This allows studios to support multiple teams — including across sites and contractors — on one common GPU platform, improving collaboration and reducing debugging issues that can arise from disparate hardware.

Supporting AI and Engineering on Shared Infrastructure

AI is becoming a core part of everyday game development, spanning coding, content creation, testing and live operations. As these workflows expand, studios need infrastructure that can support AI alongside traditional graphics workloads without introducing separate, siloed systems.

With the RTX PRO Server, studios can support coding agents, internal model experimentation and AI-assisted production workflows without spinning up a separate AI stack for every team.

The NVIDIA RTX PRO 6000 Blackwell Server Edition GPU features a massive 96GB memory buffer, enabling developers to run multiple demanding applications simultaneously while supporting AI inference on larger models directly alongside real-time graphics workflows.

NVIDIA Multi-Instance GPU (MIG) technology partitions a single GPU into isolated instances with dedicated memory, compute and cache resources. Combined with NVIDIA vGPU software, MIG can help studios securely allocate GPU capacity across users and workloads. In combined MIG and vGPU configurations, a single RTX PRO 6000 Blackwell Server Edition GPU can support up to 48 concurrent users, maximizing utilization while maintaining performance isolation.

Enterprise-Ready Deployment for Game Studios

NVIDIA RTX PRO Servers are designed for enterprise-grade data-center operations. Studios can deploy virtual workstations on RTX PRO Servers via NVIDIA vGPU on supported hypervisor and remote workstation platforms.

That means RTX PRO Servers can fit into studios’ existing infrastructure and IT practices, rather than requiring one-off deployments.

Major game publishers already use NVIDIA vGPU technology to scale centralized development infrastructure and improve efficiency at studio scale.

Learn more about the NVIDIA RTX PRO Server.

See these workflows live by joining NVIDIA’s booth 1426 at GDC or attending NVIDIA GTC, running March 16-19 in San Jose, California. 

See notice regarding software product information.

Computers, networking, games, people management, call centers, reporting & whatever.