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!

Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:

All in seconds. No searching, no switching tabs, no manual updates.

Ready to scale faster?

Start now →

1. Jeff Bezos wants $100 billion to buy manufacturing companies and automate them

We start this week with Jeff Bezos, who plans to raise $100 billion and buy up dozens of manufacturing companies across chipmaking, defence, and aerospace - then automate them with AI.

He's been meeting sovereign wealth funds in the Middle East and Singapore to drum up backing for a "manufacturing transformation vehicle," which aims to rival SoftBank's Vision Fund in scale.

The strategy hinges on a startup called Project Prometheus, which is building AI models that can simulate the physical world. This could be incredibly useful for the manufacturing sector, as you can use these AI models to predict airflow around a plane’s wings or where metal parts will crack under stress.

Jeff Bezos setup the company back in November and now serves as their co-CEO. Ultimately, he wants to apply this technology across the $100 billion fund’s portfolio to make the manufacturing companies more efficient and profitable.

It’s a playbook that some investment firms are already running in sectors like accounting and property management, but we haven’t seen it happen in such a dramatic scale for the manufacturing sector.

While there’s huge ambition here, and Bezos is one of the few people who can realistically attempt something at this scale, the gap between simulating physics in a lab and transforming legacy manufacturing businesses is enormous.

Whether Prometheus can deliver and automate factory floors - rather than just impressing investors with demos - remains to be seen, but with this kind of runway, they've certainly got time to try.

2. Polymarket is posing a real threat to journalists

Over the last few years, you might have heard about “prediction markets” called Polymarket and Kalshi. Both websites have become incredibly popular over recent months, as they allow people to essentially bet on anything.

For example, you can place money on who the next US President will be in 2028, when the US/Iran conflict will end, or who will become the next leader of Vietnam. It’s led to billions of dollars being spent on these wagers, which has exposed a really nasty side when people lose these bets.

In Israel, people have even targeted journalists like Emanuel Fabian, who reported that an Iranian ballistic missile struck on March 10 - a fact that was confirmed by the Israeli military.

More than $14 million had been wagered on whether Iran would strike Israel that day, and the bet's rules state that intercepted missiles don't count, so strangers immediately messaged him demanding he change the report to say the missile was intercepted.

What started as polite emails escalated into fabricated screenshots of conversations he never had, a fake lawyer calling him, and eventually explicit threats referencing his home, his parents, and his family - with one person claiming that they would spend $900,000 to "end" him if he didn't change the article.

Fabian reported the threats to police, and Polymarket has since banned the accounts involved, although the company didn't address whether it could actually prevent this kind of interference in future.

It’s an incredibly alarming story, as these new technology platforms could pose a threat to our society at large. If people are placing millions of dollars on an outcome, and can use insider knowledge to win big, it raises big questions for honesty and corruption.

3. Spanish company is able to shrink AI models by 95%

This is a startup that I highlighted back in June and is working to dramatically slash AI operating costs. The company has already raised over $215 million and is aiming to raise another $580 million.

Essentially, the company uses a compression system that’s inspired by some of the principles in quantum computing. They claim that their technology can reduce the size of LLMs by 95%, without impacting the model’s performance.

Multiverse has recently launched its own AI chatbot, which is powered by a model that’s small enough to run directly on your device - without having to be run in the cloud. They’ve also released their own API portal for developers, allowing them to use their compressed models.

Of course, you can only run these models locally if your device has enough RAM - otherwise the app will have to fall back and use cloud-based processing instead. But the real value is in enterprise and edge use cases, where an internet connection can't be guaranteed - including drones, satellites, and critical infrastructure.

If your team is keen on using small models, but aren't sure where to start, Multiverse's API is worth a look.

4. Sam Altman thinks iris scanning will prevent fraud with AI agents

Tools for Humanity has released a new tool called AgentKit, which is designed to prove that a real human is behind an AI agent's purchasing decisions. The startup was founded by Sam Altman in 2019 and he currently serves as their chairman.

If you're not familiar with agentic commerce, it's when we use AI agents to browse the web and buy things for us. As you’d expect, there’s quite a bit of buzz around this area and plenty of companies are keen to get in on the action - with Amazon, Mastercard, and Google starting to build the technology into their platforms.

But there are obvious concerns here about fraud and how people can really trust the technology. Until that’s resolved, I highly doubt that people will be comfortable giving these AI agents access to their bank accounts.

To tackle this, Sam Altman’s company has created a device that scans your iris and can verify that you’re a real person. This ties in nicely with the release of AgentKit, which allows other companies to see that you approved the AI agent’s purchase - rather than someone else.

Of course, the irony isn't lost that Altman's other company, OpenAI, has been widely blamed for creating the very problem World is now trying to solve. And the adoption barrier here is significant - this only works if consumers have scanned their irises with an Orb, which is far from mainstream.

In a world where we’re being asked to hand over more and more data, I’m incredibly skeptical that people will want to be doing this. Who really wants to scan their iris to approve a transaction?

5. Gecko Robotics lands the largest US Navy robotics deal yet

The robotics company has signed a five-year deal with the US Navy, which is worth up to $71 million, and will create digital replicas of their ships. These are commonly referred to as a "digital twin” and manufacturing companies are already using them to improve efficiency and spot issues early.

The idea is that Gecko's robots will crawl through every part of a ship and create a detailed digital twin - allowing the Navy to monitor the condition of its fleet and get ahead of maintenance issues before they become serious.

It will start with 18 ships in the US Pacific Fleet, but the ambition is to expand the program further. It’s also worth giving context here, as around 40% of the Navy's fleet is unavailable at any given time due to incredibly long maintenance cycles.

They currently spend around $13-20 billion to upkeep the entire fleet, every single year. But the Navy wants to hit 80% ship readiness by 2027, which is a significant jump from where they are today.

If the digital twin approach works well at scale, it could meaningfully reduce the time ships spend out of service - which matters a lot given the current geopolitical climate. But scaling from 18 ships to an entire fleet is a big step, and it'll be interesting to see how quickly that happens.

Regardless, this is a great example of robotics and AI being applied to a genuinely practical problem - rather than the flashier use cases we usually hear about.

Google is betting on “vibe design”

Google has made a big update to Stitch, allowing you to “vibe design” and create a high-quality UI in just a few seconds. Similar to other designer tools, like Figma, you can setup an infinite canvas and explore different ideas.

You can also use a design agent and have it reason across your entire project, use voice mode to talk through the changes in real-time, or click one button and turn your designs into a working prototype.

The tool also supports DESIGN.md files, which can be used to upload your company’s design rules and share them across different tools - such as Claude Code.

Google is clearly targeting two audiences here - professional designers who want to explore different ideas and work faster, and non-designers who want to skip the design phase entirely and go straight from idea to prototype.

I'm a bit skeptical about how well "vibe designing" actually works in practice, as the gap between a nice-looking prototype and something that's genuinely usable tends to be where most projects struggle.

Accessibility, responsive layouts, edge cases in user flows, and consistent spacing are all things that take real design expertise to get right, and I'm not convinced that describing what you want in plain English is enough to handle that level of detail.

But the MCP integration is worth paying attention to, as it allows Stitch to plug into existing developer workflows rather than sitting in isolation. It'll be interesting to see whether designers adopt this for their own work.

🤖 Bot traffic will overtake human traffic by 2027, according to Cloudflare

🚫 Pinterest's CEO wants social media to be banned for under-16s

✍️ WordPress allows AI agents to write and publish posts

🚀 Jeff Bezos' Blue Origin enters the race for space data centres

🖥️ OpenAI is building a desktop “superapp” to rival Google

🚘 Waymo passes 170 million autonomous miles, still safer than human drivers

🩺 Fitbit’s AI health coach can read your medical records

🔒 Nvidia launches NemoClaw, a more secure version of OpenClaw

💰 DoorDash and AI companies are paying people to train AI models

📍 FBI is buying location data to track US citizens, its director confirms

🧰 Mistral bets on “build-your-own AI” to compete with OpenAI and Anthropic

🎥 ByteDance pauses the global rollout of its Seedance 2.0 video model

This startup was spun out of Rivian just a few months ago by its CEO RJ Scaringe, with a focus on building industrial robots for factories.

The company plans to use data from Rivian's own EV factory and train robots that can do the fiddly work that traditional factory automation still struggles with - things like handling small parts, adjusting grip pressure, or working in tight spaces.

To make that happen, Mind Robotics has just raised $500 million in Series A funding - bringing its total to $615 million and its valuation to around $2 billion, which is a lot of money for a company that's only a few months old.

But what sets them apart from many of their competitors is that they're explicitly steering away from the humanoid robot designs that have dominated the conversation over the last year.

There's also a potential hardware angle, as Rivian has been developing its own custom silicon for autonomous vehicles, and Scaringe has hinted that those chips could end up powering Mind Robotics' systems as well.

Having Rivian as both a source of real factory training data and a venue to deploy the robots gives them a meaningful advantage over competitors starting from scratch. That said, Scaringe's claim that there will be a large number of robots deployed by the end of this year feels ambitious for a company that was only setup in November.

Given the funding they’ve raised already, and the amount of data available to the team, Mind Robotics is certainly one to watch in the coming year.

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.