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!

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

1. Anthropic accidentally leaks the source code for Claude Code

We start this week with Anthropic, which has had a rough few days after the source code for Claude Code was accidentally leaked online.

The company published a new version of the tool earlier in the week, but this accidentally included the source map file - exposing almost 2,000 TypeScript files and more than 512,000 lines of code.

A security researcher was the first to spot the mistake and alerted Anthropic, but it was too late. Users have since re-published the code online, where it’s been copied over 50,000 times.

It’s worth noting that no customer data or credentials were leaked, as they’re stored in different systems. This source code is for the tool itself and allows developers to better understand how Anthropic has built Claude Code.

Developers have already dug through the code and found some hidden features that Anthropic hasn't announced yet. One is a Tamagotchi-like pet that can sit beside your input box and react to your coding. Another is a feature called "KAIROS" that sets up an AI agent and allows it to constantly run in the background.

There had been previous efforts to reverse-engineer Claude Code, but nothing came anywhere close to this level of access. That said, I don’t think this will cause much long-term damage.

Anthropic’s competitors already had a decent idea of how Claude Code works and the techniques it uses. But this will be incredibly useful for smaller companies that are building AI agents for specific sectors - like finance, health, and transport.

2. Stanford claims AI chatbots are making people more self-centred

A new study from Stanford has found that AI chatbots are 49% more likely to tell you that you're correct versus a human - and it's changing how people behave. Users who interacted with these sycophantic models became more self-centred and less likely to apologise.

To test this, the team asked 11 major chatbots for advice on real-world dilemmas - including posts from Reddit where the community overwhelmingly agreed that the person was in the wrong.

But even in those clear-cut cases, chatbots sided with the user 51% of the time.

In a second experiment with over 2,400 participants, people consistently preferred and trusted the sycophantic responses - and said they'd be more likely to come back for advice. In other words, the behaviour that's causing harm is also the behaviour that keeps users engaged.

This isn't particularly surprising if you've spent any time with these tools, but the sheer scale of the problem is alarming. AI companies know that sycophancy is a real issue, but users actively prefer it - which means there's very little commercial incentive to fix the behaviour.

It's hard not to see echoes of social media here. Facebook spent years optimising its platform for engagement before anyone fully understood the cost - and by the time the research caught up, the damage to public discourse and our mental health was already done.

AI companies are now walking the same path and building products that tell users what they want to hear, as it drives engagement and keeps people on their platform.

And with 12% of US teens already turning to chatbots for emotional support, the window for us to get this right is closing fast.

3. OpenAI raises $122 billion, the largest private funding round in history

OpenAI has closed a massive $122 billion funding round and is now valued at $852 billion, making it the largest private funding round in history.

The round was backed by Amazon, Nvidia, and SoftBank, with continued funding from Microsoft. For the first time, OpenAI also raised over $3 billion from individual investors through bank channels.

The numbers are pretty staggering here. ChatGPT now has more than 900 million weekly active users and 50 million subscribers - with OpenAI generating $2 billion in revenue per month.

Enterprise now makes up more than 40% of revenue and it’s on track to match consumer revenue by the end of 2026. Anthropic is making a similar push into the enterprise sector, so we should expect this to be a major battleground over the next year.

OpenAI is also starting to reduce its dependence on Microsoft and Nvidia by striking deals with other cloud and chip providers. The company is even developing its own custom silicon with Broadcom, which could give it more control over costs as it scales.

4. SpaceX files for IPO at more than $1.75 trillion

SpaceX has filed for a US IPO, which could value the company at more than $1.75 trillion and make it the largest stock market listing in history.

To put that into context, it would comfortably surpass Saudi Aramco's 2019 flotation, which still holds the record. SpaceX is reportedly looking to raise more than $50 billion, which would give the company a massive war chest to fund its Starlink expansion and Mars ambitions.

This news comes after SpaceX merged with xAI earlier this year. Much of the valuation is built on Starlink, which now has 9 million subscribers, growing defence contracts, and its own data network.

Unlike most of Musk's ventures, SpaceX is already highly profitable and generated around $8 billion in profit last year.

That said, the way the shares are structured means Musk would likely keep full control of the company even after going public, and the usual questions about his ability to oversee multiple trillion-dollar companies aren't going away.

5. Whoop's valuation just tripled to $10 billion

Whoop has raised $575 million at a $10.1 billion valuation, nearly tripling its previous figure. The round was backed by sovereign wealth funds, medical device giant Abbott, and individual investors - including Cristiano Ronaldo, LeBron James, and Rory McIlroy.

The company says it brought in $1.1 billion in sales last year, more than double the year before. That's a strong number for a hardware company, as Whoop has to manage inventory, shipping, and subscriptions all at once - something a pure software business doesn't have to worry about.

The Abbott partnership is worth watching, as it's one of the biggest medical device companies in the world. By getting them on board, Whoop seems to be moving beyond fitness tracking and into proper health monitoring - which could help to differentiate them from other competitors, like Oura and the Apple Watch.

The big question is whether Whoop will go public. Ahmed says they're doing "a lot of the no-regrets work" to be IPO-ready, and with their rival Oura reportedly already talking to bankers, the pressure is starting to build.

Either way, $10 billion wearable companies were hard to imagine a few years ago - the market has clearly moved on from early Fitbit days.

Baidu’s robotaxis freeze in traffic, trapping passengers

Baidu had a pretty tough week, as hundreds of its robotaxis froze in traffic and caused chaos. An unknown technical problem forced the cars to stop mid-journey, trapping passengers and causing at least three collisions.

Dashcam footage was posted online and showed that cars had to swerve around rows of frozen robotaxis, which were stuck in Wuhan’s fast lanes. In one recording, a driver passed 16 stopped robotaxis in the space of just 90 minutes.

Passengers inside the autonomous vehicles were stuck for over an hour and had no way to get help - as customer support took 30 minutes to reach them, and the SOS button didn't work at all. One passenger angrily wrote online, "so then what exactly is the SOS for?"

It's not a great look for Baidu, which has been expanding its robotaxi fleet aggressively in recent years. The company now operates in over a dozen Chinese cities and has already completed over 20 million rides, but this is exactly the kind of incident that sets the whole industry back.

It's one thing for the technology to fail, but the complete breakdown in customer support and safety systems is harder to explain away. If your emergency button doesn't work when the car stops in a fast lane, that's a problem that goes well beyond a software glitch.

This comes at a critical time for the industry. Waymo is expanding into new US cities and preparing to launch in London, while companies like Uber and WeRide are ramping up their own operations in Dubai.

If these companies want people to trust robotaxis for getting around, incidents like this can't happen - and more importantly, the backup systems can't fail at the same time.

🤖 Claude now supports Microsoft 365 connectors, while Copilot Cowork launches for enterprise

🚀 NASA launches four astronauts toward the Moon on the Artemis II mission

🖥️ Even astronauts face issues with Microsoft Outlook

🧬 Anthropic acquires biotech startup Coefficient Bio for $400 million

🏢 Oracle cuts thousands of jobs as it spends more on AI

🚕 Uber and WeRide launch robotaxis in Dubai

📊 15% of Americans say they'd be willing to work for an AI boss

🤔 Fewer Americans trust AI tools, even as adoption grows

🧪 Mantis Biotech builds digital twins of humans to tackle medical data gaps

🛰️ Starcloud raises $170 million to build data centres in space

👋 Elon Musk's last remaining co-founder leaves xAI

📈 Anthropic's Claude sees a surge in paying consumers

This London-based startup has taken a clever approach to a problem most people have experienced - AI meeting bots that awkwardly join your call and make everyone suddenly self-conscious about what they're saying.

Granola instead sits quietly on your computer, transcribes your meetings locally, and generates notes without anyone else in the call knowing it's there. It's a simple distinction, but it's the core reason the product has taken off.

The company has just raised $125 million in Series C funding led by Index Ventures. That brings its valuation to $1.5 billion - up from $250 million less than a year ago.

What started as a personal productivity tool has been steadily moving into enterprise, with companies like Vanta, Asana, and Cursor now using it. The latest update introduces team workspaces and new APIs that let users plug their meeting notes into other AI tools.

That API launch is worth noting, as it followed a backlash when Granola locked down its local database and broke some of the automations that power users had built on top of it. Granola has since promised to fix this.

The challenge for Granola is that AI meeting notes are rapidly becoming a commodity, with Zoom, Notion, and others all offering similar features.

But the real value is what happens after the notes are taken - drafting follow-ups, updating CRMs, and feeding context into other tools. That's where Granola's integrations could give it an edge.

It's great to see Granola secure this kind of funding, and the growth from $250 million to $1.5 billion in under a year speaks to genuine demand. The bigger question is whether meeting notes alone can sustain a billion-dollar company - or whether Granola needs to become something bigger to justify the valuation.

If you want to try Granola for yourself, I've included a link below.

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.