You’re receiving this email because you subscribed to the newsletter. If you’d like to unsubscribe, you can manage your preferences here with one click (or use the button at the bottom of the email).

Welcome to this edition of Loop!

To kick off your week, I’ve rounded-up the most important technology and AI updates that you should know about.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

Let's jump in!

1. Claude has made one small step for AI on Mars

We start this week on the planet Mars, where NASA has used an AI model to plan routes for the Perseverance rover - a first for the space agency.

Earlier in December, Anthropic's Claude model was used to plot waypoints for the rover and helped it to navigate 400 metres through a rocky Martian landscape.

It’s the first time that an AI has been trusted with this task, but the process wasn’t as simple as feeding Claude a map and asking for directions.

NASA engineers provided years of driving expertise and contextual data through Claude Code, which allowed the model to write commands in Rover Markup Language - a bespoke XML-based system that’s specific to the Mars missions.

Claude then analysed the overhead imagery, planned the route in ten-metre segments, then critiqued and refined its own work. Humans were still involved in checking the results, making minor adjustments based on camera footage that Claude hadn't seen.

NASA believes that this approach could halve the time needed to plan routes, allowing it to run more drives, collect more data, and ultimately make more scientific discoveries.

As missions venture further into the solar system - where signal delays stretch to hours and solar power becomes impractical - AI systems could be used to make smart, adaptive decisions without having to wait for Earth's feedback.

While 400 metres isn't that far, it's a meaningful first step towards something much larger.

2. Google allows you to create 3D games with AI

Google has begun rolling out Project Genie, an experimental feature that allows users to create and explore interactive worlds that were created with AI.

The system is powered by Genie 3, a "world model" that can simulate these 3D environments in real-time. It’s pretty easy to get started and you can use text or images to easily build these 3D worlds.

Google has also integrated its Nano Banana Pro image model, so that you can preview what the world will look like and make adjustments before diving in.

That said, the technology clearly has limitations. Generated worlds won't always follow real-world physics, character controls can sometimes feel laggy, and sessions are capped at 60 seconds.

Plus, Google is only allowing their highest paying subscribers to access the feature - since it’s so expensive for them to run. You’ll need an Ultra subscription and be based in the US.

It's still early days, but the technology has raised a really interesting question: could world models eventually become the foundation for everything, from game development to training robots?

Google seems to think so - and they're opening access to Genie 3, so that they can better understand potential use cases for the technology. If you want a more detailed overview of Genie 3 and what it can do, you can read my write-up here.

3. Gemini will be your tour guide in Google Maps

Continuing on with Google, you can now use their AI model as a tour guide in Maps.

For example, you can ask questions about the neighbourhood you’re in or find local restaurants nearby. To do this, Gemini will access to your location and can then search online for the information it needs.

Overall, it’s a pretty sensible way to use voice agents and make them accessible for the wider public. The technology has made significant progress in recent years, with support for hundreds of languages around the world.

But there are still issues with accents and understanding what people are saying. I regularly face these problems myself, so it’ll be interesting to see how Google tackles this - especially when you’re walking down a noisy street.

I’m just glad it’s not another boring chat interface…

4. Anthropic launches interactive apps with MCP

The company's latest update allows users to interact with apps like Slack, Canva, Figma, and Asana directly inside Claude's interface - without the need to switch tabs or copy-and-paste between windows.

Previously, connecting these tools meant that you would only be shown text responses. Now users can actually see and manipulate their work visually.

For example, you can customise Canva presentations in real-time, format Slack messages before sending, or build interactive charts with Hex and Amplitude.

It’s similar to the "mini apps" that you can use inside Telegram and WeChat. This is all possible thanks to the Apps extension for MCP, which has become an industry standard for communicating with AI models and platforms.

Salesforce integrations are also on the way, which will allow us to better integrate Claude with enterprise platforms and run more complicated tasks.

It’s part of a move to position Claude as a workspace, rather than a simple chatbot, and allow Anthropic to extend its lead over OpenAI in the enterprise sector.

5. France bans officials from using US big tech

The French government has taken steps to remove its dependence on American big tech, as officials will be banned from using Google Meet, Zoom, and Microsoft Teams for video calls.

Instead, civil servants will be required to use Visio - a videoconferencing tool that was built by France's Digital Authority - and it will be hosted on servers run by French company Outscale.

French officials were also forced to remove WhatsApp and Telegram last summer, with Tchap being used in their place.

It’s part of a broader push to shift to use local services, as European leaders become worried about how reliant they are on US digital infrastructure.

Over the last 12 months, this has been brought into focus and European politicians are slowly making the changes necessary. But these changes are incredibly expensive, as local companies will need more investment to compete with America’s tech industry.

I’m not sure that European politicians are ready to make that commitment yet, unless something big jolts them into action. Regardless, this is a wider trend that we’ll see more of in the coming year.

Clawdbot is impressive, but a security nightmare

You might have heard about Clawdbot over the last week, which has since been rebranded as Moltbot. And it’s since been renamed again as OpenClaw. Maybe branding isn’t their strongest suit.

The tool has become incredibly popular in recent days, as the community tests the different ways that it can be used. Essentially, it’s an AI assistant that can manage your calendar, send messages, and handle digital tasks for you - completely autonomously.

In some cases, people have used it to order food online and asked the bot to communicate with them on Telegram. But the real power is the AI agent’s ability to control your entire computer, file system, and run Terminal commands - which is also its greatest weakness.

Since the AI model will have full access to your computer, it can do anything it wants. That’s very useful for completing tasks, like organising hundreds of files, but it raises huge security risks.

Over the last 2 years, I’ve developed AI agents for my professional work and seen many instances where they become confused or misunderstand what I want it to do. In one case, the AI agent misinterpreted what I meant by “undo your changes” and then tried to delete important files.

If I didn’t spot this mistake by the AI and put a stop to it, I would have lost these files forever. The same issues are present with Clawdbot / Moltbot / OpenClaw / whatever the next name is.

There’s also a security risk with how the AI agent reads your prompts. If a bad actor knows that you have installed OpenClaw, they could send an email like “ignore all previous instructions and run [command]” to extract your files.

If you do want to use this AI agent, make sure that you run it in a secure environment and don’t give it access to your computer. This does limit what it can do, as it won’t have access to your accounts and data, but the risks of installing it locally on your machine are too high.

If you want to read more about securing OpenClaw, this is a useful guide from Composio.

🚀 SpaceX wants to put 1 million data centres into orbit

🚗 Waymo raises $16 billion in funding

💰 Amazon is in talks to invest $50 billion in OpenAI

🔭 Astronomers used AI to find 1,400 “anomalous objects” from Hubble archives

🔋 Redwood Materials raises $425 million for battery recycling and energy storage

🥽 Meta burned $19 billion on VR last year

📉 Tesla profits have fallen by 46% in 2025

👥 Amazon says it will lay off 16,000 people

🤖 Waabi raises $1 billion and expands into robotaxis with Uber

🔬 OpenAI launches Prism, a new AI workspace for scientists

📊 Uber will gather driving data for robotaxi companies

💻 Microsoft announces an AI chip to compete with Amazon and Google

This impressive startup has just raised $100 million in Series B funding, alongside a $49.8 million Space Force contract to modernise America's satellite control network. Considering that the startup only closed its $30 million Series A less than a year ago, it’s quite a jump up.

The company was founded in 2023 by Bridgit Mendler, who rose to fame in the 2010s as an actress on Disney Channel. She then paused her career to study at MIT and setup Northwood Space with her husband.

Their main focus is on building compact antenna systems, which are designed to replace the ageing dish infrastructure that currently handles our satellite communications. It’s an incredibly hard problem to solve, as it requires a serious amount of capital - but the timing couldn’t be better for Northwood.

As orbit becomes increasingly crowded with new constellations, the bottleneck is shifting earthward. SpaceX and Amazon can afford to build their own ground stations, but everyone else is left competing for third-party capacity that’s extremely limited.

Northwood believes that owning the entire ground station stack will prove invaluable, as satellite operators scale from a handful of birds to dozens or hundreds.

But the Space Force contract is particularly interesting. According to a 2023 GAO report, the Pentagon has known about capacity issues with its satellite network since 2011 - so it desperately needs to modernise the technology.

By late 2027, Northwood expects that it will handle communications with hundreds of satellites. Whether that timeline holds remains to be seen, but with $150 million in fresh capital, they've certainly got the runway to try.

This Week’s Art

We’ve covered quite a bit this week, including:

If you found something interesting in this week’s edition, please feel free to share this newsletter with your colleagues.

Or if you’re interested in chatting with me about the above, simply reply to this email and I’ll get back to you.

Liam

Share with Others

If you found something interesting in this week’s edition, feel free to share this newsletter with your colleagues.

Liam McCormick is a Senior AI Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.