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!

You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.

Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.

Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

1. The first AI-designed vaccine enters human trials

We start this week with researchers at the University of Cambridge, who have managed to develop the first ever vaccine with Artificial Intelligence.

This new vaccine targets all coronaviruses, which includes every Covid variant, alongside animal viruses that have the potential to start the next pandemic.

The team fed genetic codes from known coronaviruses into an AI system, which then designed a "super-antigen" - which is the part of a vaccine that trains the immune system to spot an infection.

It has since been trialed with 39 people as part of a safety study, and the team are already working on similar AI-designed vaccines for flu, Ebola, and H5N1 bird flu.

In a world where we're constantly reminded about the impact this technology is having on the environment, the rise of tech oligarchs, and the risk of jobs being displaced - it's good to see a really positive story about how the technology is being used to save lives.

If you want to read more about the story and how the scientists have approached this, I've included a link below.

2. How social media platforms are targeting children at school

In the last year, we've seen a lot more focus on social media platforms and the harm they’re doing to children - with Australia formally banning access for under 16s and other countries following suit.

The New York Times has just published an investigation into how Meta, Snap, TikTok, and YouTube have deliberately targeted children during school hours.

It found that Snapchat and TikTok are sending notifications to teenagers and encouraging them to share what’s happening in their classrooms - with employees saying that "teachers are going to hate it".

They also discovered that TikTok has given millions of dollars to the National PTA, so that the company could get favourable media quotes from parents and avoid scrutiny from US lawmakers. Meta has also paid "teen ambassadors", in exchange for promoting Instagram to their friends at school.

Internal safety teams - at all four companies - repeatedly raised concerns about these features and the impact they were having on children, but they were overruled by teams that were solely focused on growth.

It's not the first time I've heard of this. A few years ago, I attended an event in Silicon Valley where former Facebook leaders, including Kent Beck, talked openly about the decisions they made building the company.

They were quite honest about the fact that they were optimising the feed and recommendation system purely for reactions and clicks, which obviously drove profit.

But what they didn't realise at the time was that this also pushed the algorithm to favour emotional content that triggered stronger reactions like anger, and they were accidentally optimising for it.

Those former leaders said that they regret those decisions and want to avoid the same thing happening with AI. But if we don't deal with what these companies are doing to children now, the next wave of AI-powered products will be far better at keeping them hooked, and the harm done will be much greater than what we're seeing today.

3. World’s largest chipmaker can't keep up with AI demand

The world's biggest chipmaker has admitted that it can't keep up with American AI demand, despite growing its US production.

TSMC's CEO has said that "customer demand is so high, and we can only support so much", and that fulfilling American needs from US-based production could take a "very long time".

The company already runs one factory in Arizona and plans to invest $165 billion in three more plants and an R&D centre, but even that doesn't seem to be enough.

The AI boom has already caused a shortage of RAM and NAND Flash memory that's expected to last years, with Deloitte forecasting the semiconductor industry will hit $1 trillion by 2027.

Wei also said he'd "like" to raise prices, but without the abrupt and dramatic increases we've seen with DRAM and SSDs.

It seems like this supply problem could last for years, and it will hit everyday consumers too. If you're planning to buy a new computer, gaming console, or phone over the next couple of years - you might have to pay more for it, which is the last thing people need during a cost of living crisis.

4. Anthropic warns about AI that can self-improve

The company has published a long piece on recursive self-improvement, which is when AI systems can autonomously design and develop their own successors.

They're clear the technology is not there yet, but their internal data is pretty interesting - Anthropic engineers are now shipping 8x more code per quarter than they did between 2021 and 2025. They also said that more than 80% of code is being authored by Claude.

Back in March 2024, models could only handle tasks that lasted four minutes, but today they can manage ones that last 12 hours - with the doubling time accelerating from seven months to four.

The piece ends with a call for AI labs to be able to pause development together when needed. For this to work, each lab would need a way to verify that the others have actually stopped - and Anthropic says it would do this if the other frontier labs agree.

Of course, it's incredibly unlikely that this coordinated pause would ever happen as the competitive pressure is just too great. But this all could be part of a broader strategy - Anthropic has filed to go public in the near future, and Anthropic's co-founder recently met with Pope Leo to talk about the need for safer AI systems.

The company has consistently been more active on safety and alignment research than any other frontier lab, so this it's clearly a top priority for them - and pieces like this are designed to slowly prepare people and wider society for what's coming down the tracks.

5. Microsoft halves its quantum computing timeline

Microsoft has announced a new quantum chip called Majorana 2, which is 1,000 times more reliable than the previous version.

While regular computers have binary bits, quantum computers instead use qubits. For the last few years, creating reliable qubits has been a major challenge and one of the biggest blockers to bringing quantum computing into the real world.

This improvement comes from a new mix of materials inside the chip, with Microsoft using lead for the superconductor instead of aluminium. They claim that some qubit lifetimes now exceed a minute, which is up from 1-12 milliseconds in Majorana 1.

Thanks to this progress, Microsoft has halved its roadmap and now aims to build a fault-tolerant prototype quantum computer by 2029. But it's worth noting that physicists were sceptical of Microsoft's Majorana 1 claims last year, so I'd treat the headline numbers with the same caution until they're independently verified.

Alongside the new quantum chip, the company has released a new agentic AI tool for quantum researchers, called Microsoft Discovery. This tool was used to design Majorana 2, so it should be useful for other researchers in the industry.

Helsing puts Europe back in the robotics race

Helsing is one of Europe's top AI defence companies and they've just unveiled their first robotics research platform, which is being designed and manufactured entirely within Europe.

If you're not familiar with Helsing, they were one of the first companies to put an AI system inside a fighter jet - with their AI pilot being used to control the Saab Gripen. They're also working on their own autonomous fighter jet, which is called the CA-1 Europa and is built on the same technology that controlled the Gripen.

This new robotics platform is being positioned as a European alternative to those produced in the US and Helsing is clearly tapping into a wider trend here. Over the last 18 months, Europe has realised that it's too reliant on the US militarily and desperately needs to boost its own technological sovereignty.

Helsing says that they will make the RX-1 platform available to academic institutions and labs across the continent, with ETH Zurich and INRIA Paris being one of the first to get access.

I've covered Helsing many times in my newsletter and they're a very impressive startup. In just a few years, the company has built advanced AI systems that can control fighter jets, developed underwater drones that scan for submarines, and they're about to launch satellites that can spot threats from space.

This kind of work is hugely important for Europe right now. With Russia's ongoing invasion of Ukraine and the wider threat it poses to other European countries, building sovereign defence technology has become a serious priority for the continent.

Helsing is one of the few European companies that are actually moving fast enough to deliver it, and the launch of RX-1 is the latest example of how they're helping Europe to better defend itself.

🛰️ Google will pay SpaceX $920 million a month for compute

💰 Supabase doubles its valuation to $10 billion in 8 months

🛡️ The NSA will use Claude Mythos in cyber operations

⚠️ xAI's Grok falsely identifies a UK police officer, forcing them to flee their home

🔋 Waymo's old robotaxi batteries will be reused for grid storage

⚖️ Trump signs executive order to review AI models before public release

📈 Anthropic has officially filed to go public

💰 Google plans to raise $85 billion for its AI buildout

✂️ GitLab cuts 14% of staff as it pivots towards AI workloads

💸 Uber caps employee AI spending after blowing the budget in 4 months

🚀 Mach Industries hits $1.8 billion valuation, a 4x jump in just a year

🏗️ SoftBank will invest up to €75 billion in French data centres

Helion is one of the few companies in the world that's actually trying to put fusion power on the grid, which has been heavily backed by Sam Altman.

The company is based in Washington and is now building its first fusion power plant. Helion has signed a deal with Microsoft to deploy fusion-generated electricity to the grid, which could happen as early as 2028.

They've just managed to raise $465 million in a Series G round, valuing the business at $15.5 billion and it brings their total funding to $1.5 billion.

It's worth noting that Helion's approach to fusion energy is quite different from its competitors. Most fusion startups either use magnets to contain the plasma or lasers to compress fuel, and then run the generated heat through steam turbines to make electricity.

Instead, Helion uses magnets to compress the fuel and harvests electricity directly from the magnets themselves. It's similar to how an electric vehicle can recover energy through regenerative braking, and if it works, would dramatically improve the efficiency of a fusion power plant.

The fusion sector is having a moment right now, as investors are keen to get in on the action. There has been a steady stream of big funding rounds in recent months, including Focused Energy ($240M), Thea Energy ($100M), Inertia Energy ($450M), and Type One Energy ($250M).

Some experts are sceptical of Helion's approach, as the company doesn't publish in peer-reviewed journals - which means that physicists can't fully assess their theory. That said, this kind of work is incredibly important for the tech industry right now.

The demand for compute has exploded over the last few years, and the energy needed to power it is rapidly running ahead of what tech companies can actually get their hands on. Microsoft and other tech giants are already struggling to find enough clean, always-on power to run their AI workloads.

That problem is only going to get worse in the coming years. If Helion can pull this off and start delivering fusion power to the grid by 2028, it would be a huge step forward for everyone trying to build the next wave of AI.

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.