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!

Most AI creative tools fall short for one simple reason. You can generate tons of ads, but they aren’t up to par.

Refining copy, adjusting layouts, or nudging a CTA into place shouldn’t require rewriting prompts over and over. It slows teams down and breaks the creative process.

With Hightouch Ad Studio, AI gets you 90% of the way there. For the final 10%, use a built-in editor to quickly refine copy and design, or export directly to Figma for seamless collaboration with your design team.

Move faster without losing control. Every ad, exactly how you want it.

See how Ad Studio works

1. Isomorphic Labs raises billions to create new medicines with AI

We start this week with news from London, where Isomorphic Labs has raised a whopping $2.1 billion in Series B funding - allowing it to scale its work and use AI to design new medicines.

The company was spun out of DeepMind back in 2021, and is being led by Sir Demis Hassabis - who also runs Google DeepMind.

But what's interesting here, aside from the huge amount of money they've raised, is the fact that the UK Government has taken part in this funding round. Isomorphic Labs has received new backing from the UK Sovereign AI Fund, which signals that the UK government wants to keep this kind of research close to home and prevent it being moved to the US.

For decades, successive UK Governments helped talented companies to be setup in the country - then watched as they moved to the US for the funding and support network they needed, or are bought by foreign firms. But this has changed in recent years, following Peter Kyle's push for tech sovereignty.

Isomorphic Labs now have strategic partnerships with Novartis, Lilly, and Johnson & Johnson, and this new funding will be used to push their first medicines towards clinical trials.

If anyone can pull this off though, it's probably Isomorphic. They've assembled a top team from DeepMind and other AI labs, and Hassabis's track record over a decade at DeepMind really speaks for itself. With $2.1 billion in fresh funding behind them, they've got a serious shot at making it work.

2. Notion becomes a hub for AI agents

Notion has launched a new Developer Platform, which allows teams to turn its note-taking workspace into a hub for AI agents.

The company launched Custom Agents back in February, essentially acting as AI teammates that can autonomously handle tasks - like writing weekly reports, answering questions in Slack, or routing tasks between teammates.

They've now released something called Workers, which allows developers to write their own custom code and have it run directly inside Notion's infrastructure. So in practice, this means that you can create a Worker that pulls customer data out of an internal database, or one that gives an AI agent the ability to update records in your CRM.

If you've got data scattered across half a dozen different tools, you can use Notion Workers to bring them together in one place and get clearer insights. Previously, doing this kind of thing meant that you had to use third-party automation tools or hosting the code yourself - which was expensive and hard to maintain.

The platform also supports external agents now, allowing you to chat with Claude Code, Cursor, and Codex directly inside Notion - rather than having to rely on the MCP server.

For teams that are using lots of different AI tools, this is a big shift, as everything can now happen in one place rather than constantly switching between chat windows and copy-pasting data across your apps.

It's a pretty big bet from Notion too, who are now positioning themselves less as a note-taking app and more as the place where your team's agents and tools all come together. They've become the go-to platform for huge startups in Silicon Valley, like OpenAI and Ramp, so this will help those teams to really increase their productivity.

3. OpenAI launches a consulting company valued at $14 billion

OpenAI has its eyes on the consulting sector, following the launch of a new business called the OpenAI Deployment Company (or DeployCo). This new venture will help companies to build and roll out AI systems across their operations.

To kick things off, OpenAI has acquired Tomoro - an AI consulting firm that's based in the UK, whose clients include Tesco, Virgin Atlantic, and Supercell.

DeployCo is launching with $4 billion in funding and is now valued at $14 billion, but notable investors include McKinsey, Bain & Company, and Capgemini. It's a pretty strange dynamic, given the fact that DeployCo will be directly competing with these companies.

OpenAI plans to embed Forward Deployed Engineers inside customer organisations - an approach that it's borrowed from Palantir - to work alongside business teams, identify where AI can help, and then build the systems that will deliver it.

I have to say, this is a pretty smart move from OpenAI. Most enterprise AI projects don't fail because the model isn't powerful enough, they fail because companies don't know how to properly integrate the technology into their existing tools and workflows.

It also signals that OpenAI is no longer just an AI lab. They're now competing with the big consulting firms directly, which is going to make for an interesting few years.

4. Anthropic targets small businesses with new tools

Anthropic has launched a new suite of AI workflows, which are specifically designed for small businesses and can plug directly into the tools they already use - like QuickBooks, PayPal, HubSpot, Canva, and Docusign.

The idea is that business owners can switch this on inside Claude Cowork and pick from 15 ready-to-run workflows. These can handle a wide range of tasks, including planning payroll, chasing invoices, closing the books, or running a marketing campaign.

A nice touch is that you have to approve anything before it sends, posts, or pays - which directly addresses the trust issue that people have with AI systems. It's especially important for small business owners, as half of them told Anthropic that data security was their biggest concern with AI.

Anthropic is also rolling out free AI training for small business owners, both online and in-person - with a plan to tour 10 cities across the US.

Going downmarket seems to be a pretty solid move for Anthropic. There are 36 million small businesses in the US alone, compared to a few thousand Fortune 500 companies - and most of them need that extra help to adopt AI properly within their business.

5. AI model cracks Apple's most advanced security in 5 days

Security researchers have used Anthropic's Mythos AI to find a serious exploit in MacOS, which could allow hackers to take control of Macs around the world. They were able to do this by chaining together two new bugs into a privilege escalation attack, known to be one of the most dangerous types of security flaw.

With help from Claude, the research team broke through Apple's most advanced security technology in just five days. That's despite Apple spending half a decade building the Memory Integrity Enforcement technology, which was designed specifically to stop the kind of memory corruption attack that Calif just pulled off.

The same AI model has also found over 100 high-severity vulnerabilities in Firefox, in just two weeks. This is something that usually takes dedicated teams months to do and shows that the technology will become a powerful tool for spotting cyber vulnerabilities - and exploiting them.

Security experts are now warning of a flood of new vulnerabilities that could overwhelm the security teams around the world. Critical infrastructure and smaller businesses are likely to be hit hardest when this happens, as their security teams don't have the resources to patch bugs at the pace that AI can find them.

The White House has opposed Anthropic expanding access to Mythos, as they're worried that they will be blamed if a devastating AI-enabled cyberattack happens. As I mentioned last week, federal officials are now considering an executive order that would give the government oversight of the most-advanced AI models.

China’s leading robotics firm just launched a $650,000 mecha

As if AI-powered cyber attacks weren't terrifying enough, China's leading robotics company has revealed a huge robot that can punch through walls and walk on all fours.

Unitree has revealed the GD01, weighs around 500kg with a passenger inside and costs $650,000. The robot can switch from standing on two legs to walking on all fours, and is capable of punching through cinderblock walls.

Unitree are unlikely to find lots of buyers for this in the immediate term. The target market is a bit unclear, as the company doesn't really explain what the GD01 is actually for.

There are also plenty of issues with the product. When the robot switches from two legs to four, as the cockpit seat doesn't reorient for them and it means that the passenger is left staring up at the sky.

Regardless, Unitree is using this opportunity to show that it's racing ahead of rivals in the West - with its humanoid G1 robots already going viral for doing martial arts, backflips, and playing tennis.

But it's not just about publicity, there are other, more worrying ways that this technology could be used by militaries. It's easy to imagine weapons being mounted to a 500kg robot, which can walk on rough terrain and punch through walls.

Equally, robots like this could be used to evacuate wounded soldiers from the front lines - similar to the tracked robots that Ukraine has been using recently to evacuate casualties.

Either way, it seems that science fiction is slowly becoming our reality.

🏘️ 70% of Americans don't want AI data centres in their backyards

📱 OpenAI is bringing Codex to your phone

⚖️ OpenAI is preparing legal action against Apple as their partnership frays

📑 Anthropic connects Claude to a range of legal tools

🎬 Netflix is building an AI animation studio

🛰️ SpaceX and Google are in talks to launch data centres in orbit

✂️ Cisco cuts 4,000 jobs to invest more in AI, despite record revenue

💰 Helsing is raising $1.2 billion at an $18 billion valuation

🛡️ Anduril raises $5 billion and doubles its valuation to $61 billion

📊 Anthropic now has more business customers than OpenAI, according to Ramp

💻 Google unveils Googlebook, its new line of AI-native laptops

🗣️ Thinking Machines wants to build an AI that listens while it talks

This startup is based in San Francisco and it's creating AI tools that allow models to train themselves. The company was founded by Sara Hooker, who previously led AI research at Cohere.

Adaption is part of a new wave of AI startups, which are mostly focused on research work and have been able to raise significant funding to compete with the big AI labs.

The company has just launched a new product called AutoScientist, which automates the way AI models are fine-tuned for new tasks - making the process much faster than traditional methods. It builds on their existing product lineup, which allows teams to create higher-quality datasets for training.

Adaption claims that AutoScientist has more than doubled "win rates" across different models. This measures how often its trained models are picked as the better answer, when compared directly with other versions.

Those numbers are tricky to verify though, as the product is designed to adapt models for specific tasks - which means that conventional benchmarks like SWE-Bench or ARC-AGI aren't really applicable.

But being able to fine-tune AI models faster is important, as more teams can adapt the technology for their own specific work - rather than being stuck with what the big labs offer and working around their limitations.

If Adaption can deliver what they claim, AutoScientist could put advanced AI training within reach of smaller teams - not just the big labs.

If you want to learn more about the company and what they're working on, 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.