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.

ICYMI, I recently created a prompt pack that includes 100 use cases for how businesses can adopt AI. Just click a prompt and it opens in ChatGPT, Claude, or Gemini. You can try it here.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

Let's jump in!

1. Waabi and Volvo reveal an autonomous truck

We start this week with Waabi, which is a startup that’s developed an autonomous truck and partnered with Volvo.

The company has equipped Volvo’s VNL truck with its proprietary technology, including sensors, computing hardware, and the Waabi Driver software.

It follows eight months of collaboration between the two companies, with plans to expand across the US in the coming years. The company is already operating its vehicles in Texas.

The partnership builds on Volvo's 2023 investment in Waabi through its venture arm, with Volvo participating again in the startup's $200 million Series B round last year.

Waabi’s nearest competitor is likely Aurora, who are already testing their trucks between Dallas and Houston. They’ve recently hit the milestone of 100,000 driverless miles and are also expanding with a new route between Fort Worth and El Paso.

2. GitHub launches a central hub for AI coding agents

With the new “Agent HQ”, users can more easily adopt coding tools from the top AI companies - including OpenAI's Codex, Anthropic's Claude, Google's Jules, and Cognition's Devin.

The feature centres around a "mission control" dashboard, which allows you to launch multiple AI agents. Developers can even run several AI agents in parallel, which allows you to compare their responses for a specific task and then choose the best option.

It’s an interesting move, as GitHub already offers their own coding tool - known as Copilot. Rather than directly competing with OpenAI and Anthropic, they’ve decided to allow them into the platform and GitHub aims to create a wider ecosystem for AI coding tools.

Ultimately, this is about solidifying GitHub’s dominant position in the market and ensuring that users don’t migrate to other platforms. This is a growing threat for the industry leaders, as startups can now build features much faster than before and are raising millions to take on the big players.

Of course, GitHub isn’t the only company that’s worried about this. Workday has taken similar steps and opened up their platform, so that Microsoft’s agents can interact with its own. By tearing down these walled gardens, they hope that users are more likely to stay with them.

3. OpenAI's AI agent can spot security vulnerabilities and fix them

OpenAI has unveiled an autonomous security research agent, which is powered by GPT-5 and can find security vulnerabilities.

Unlike traditional security tools that rely on fuzzing or composition analysis, Aardvark takes a more human approach and runs its own analysis of the code.

Aardvark will try to build a threat model of the entire codebase, then scan any new commits for vulnerabilities, check them within a sandboxed environment, and finally works with the Codex agent to write code and suggest fixes.

The results seem to be promising, as Aardvark was able to identify 92% of known vulnerabilities. The tool has already been running across OpenAI's internal codebases for several months, with several complex security issues spotted by the agent and flagged to OpenAI staff.

The company is currently offering access through a private beta, but they hope to offer wider access in the future. If you want to apply for access, I’ve included a link below.

4. Stellantis to launch robotaxis by 2028

The company has announced that it will launch its own robotaxi service by 2028 - with Nvidia, Foxconn, and Uber also joining as partners.

Under the deal, Stellantis will manufacture the vehicles, Nvidia and Foxconn will handle the self-driving systems, while Uber will make them available for the public to use.

They're starting with 5,000 robotaxis in the US before expanding globally, although we don't know which cities will get them or when.

It's an interesting moment for self-driving cars. Waymo is ramping up across the US and is now in talks to expand to Australia. Meanwhile, Uber seems to be constantly signing deals.

They've just signed another one with Lucid and Nuro, with plans to launch a premium robotaxi service in San Francisco and compete directly with Waymo.

As I noted in my Q2 trend report, we're seeing companies move much faster on self-driving. After a decade of testing, they're finally confident enough in the technology to scale it up and bring self-driving cars to new cities.

5. Canva wants to become the only platform that businesses need

Canva has released a suite of new tools across marketing and data analysis, as it hopes to build the go-to platform for companies to grow their business. The company says these updates will turn its Visual Suite into a "Creative Operating System".

It’s also making a major push into AI-powered marketing with Canva Grow. This new product allows marketers to design, launch, and track ad campaigns - allowing their AI model to learn from that performance data.

There's also a new Email Design feature, which allows teams to create branded email campaigns. By specifying your brand colours and styling in the Canva platform, their tool can apply the same formatting to your emails.

It’s part of a wider push to eliminate other platforms, like Mailchimp, and run all your workflows within Canva’s ecosystem.

These features are powered by Canva's own AI model, which has been specifically trained to understand design trends and can generate eye-catching visuals.

If you want to collect feedback from your customers, you can use their new Forms tool and automatically import the data into Canva Sheets.

Again, this is all about reducing the need for tools outside of Canva’s ecosystem - like TypeForm or Google Sheets - and centralising that data to take advantage of their AI features.

The message is pretty clear: Canva wants to be more than just a design tool. It's positioning itself as the all-in-one workspace that handles everything from creating content to analysing advertising campaigns.

This $20,000 robot will do your chores (badly)

Would you like a $20,000 robot, which is manually operated by a human and has complete access to your house? Yeah, I didn’t think so.

You might have seen clips of this robot in the last week, after the Wall Street Journal did a report on it and asked the robot to load their dishwasher. Overall, it took the robot 5 minutes to add a fork and two glasses.

The robot itself is about 5'6” tall and has a 4 hour battery. 1X claims that it can help with chores around the house, like taking out your bins, watering plants, and doing laundry.

However, it’s not an autonomous robot at all. While it might look and act like a clumsy robot, some poor human is operating it remotely and struggling with the controls. That’s part of the problem.

1X’s CEO says that you should feel comfortable with this. But should we allow a random person, who we can’t see or know their name, into our house and allow them to walk around?

For most sane people the answer is no. But if you do want that, it’s now possible. You’ll just need to give 1X $20,000 and allow the company to use video footage of your home, so that they can train future robots.

Google’s AI helps blind people to navigate street view

Google’s research team has revealed an interesting prototype called StreetReaderAI. Thanks to recent advances with LLMs, blind and low-vision users can use the tool to explore billions of street view images.

For example, you can ask their AI to describe what’s being shown on the screen. In their demo, the user was looking at Big Ben in London and looking north. You can even ask it to control the view and look to the left or right.

The tool has two different modes - one that’s focused on navigation and general safety, or a "tour guide" version that includes details about a city’s historic and architectural details.

But what’s really special is that the AI has memory, so it can remember what you’ve went past. If you passed a bus stop, walked on for several more metres, and crossed the road - you can ask their AI, “Where was the bus stop?” and allow it to guide you back.

It’s a fantastic way to help people navigate the world, especially those with little or no vision. While it’s currently limited to Google Street View and the virtual world, you can totally see how this could be expanded on and adopted within Google’s mapping system.

If someone with low vision is walking around London and has asked Google Maps to create the route, it could alert them to potential hazards and make the entire journey a lot easier.

🤖 Anthropic is bringing Claude to Excel

🚗 Waymo in talks to launch self-driving taxis in Australia

📄 Perplexity's new AI tool aims to simplify patent research

🎬 Google creates its first ad with AI

📡 O2 partners with Starlink to launch a satellite service in the UK

💼 OpenAI could go public next year at $1 trillion valuation

🖼️ Adobe reveals new AI tools to instantly photoshop images

💻 AMD and Department of Energy will create a $1 billion AI supercomputer

🔄 OpenAI completes its for-profit restructuring and strikes a new deal with Microsoft

⚖️ Australian regulators sue Microsoft over price hikes

⚡ Google will bring a shuttered nuclear power plant back online

📈 Nvidia becomes the first company worth $5 trillion, worth more than the GDP of almost every country on earth. Bubble? What bubble?

This is a very unusual startup and has quickly caught the eye of top AI companies, like OpenAI, Anthropic, and Meta.

Mercor's business model is pretty straightforward: hire ex-employees from investment banks, law firms, and consulting firms, then connect them with the major tech companies.

These contractors are paid up to $200 an hour and given specific tasks - like filling out forms and writing reports - which will be used to train new AI models.

If Goldman Sachs won't share its data with these AI companies, Mercor allows them to hire ex-employees instead and gain their expertise that way.

It's an incredibly grey area, both morally and legally, but their 22-year-old founder insists that they’re not exploiting a "loophole". Instead, he argues that the knowledge in someone's head belongs to them, not their former employer.

The numbers involved here are pretty staggering. The company now pays out over $1.5 million a day to tens of thousands of workers, which shows the scale involved here.

Overall, it’s a business model that’s worked well for Mercor so far. Its valuation has increased 5x in just one year, as it fills the gap created when Meta invested in Scale AI and spooked the other AI companies.

While the company is now valued at $10 billion and is generating a whopping $500 million in annual revenue, I can’t see how this continues for much longer. Companies like JP Morgan, Goldman Sachs, and Accenture will eventually take this to court.

But it’s a sign of the times. AI companies are desperate for training data and they’re willing to hire ex-employees to get it. Even if the move could land them in hot water and is morally questionable, they’re betting that the future profits are worth it.

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.

Have a good week!

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.