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!

Power users figured this out early: speaking a prompt gives you 10x more context in half the time. You include the edge cases, the examples, the tone you want — because talking is fast enough that you don't skip them.

Wispr Flow captures everything you say and turns it into clean, structured text for any AI tool. Speak messy. Get polished input. Paste into ChatGPT, Claude, Cursor, or wherever you work.

89% of messages sent with zero edits. 4x faster than typing. Works system-wide on Mac, Windows, and iPhone.

Start flowing free

1. DeepSeek's new model matches those from OpenAI and Google

We start this week with DeepSeek, the Chinese lab that wiped almost $1 trillion off US tech stocks last year - and is back with a new advanced model.

While Anthropic and OpenAI are raising hundreds of billions to stay at the frontier of AI, DeepSeek has just released a new model that comes close to matching them - at just a fraction of the price.

The V4 Pro model has 1.6 trillion parameters, which makes it the biggest open-weight model available, and DeepSeek claims that it can beat GPT-5.2 and Gemini 3.0 Pro on some reasoning tasks.

The Chinese company does admit that it still trails the leading labs by 3-6 months on knowledge tests, but it's catching up fast. Given the eye-watering sums that Western investors are spending to stay ahead, that lead is about to get very expensive to maintain.

The timing is pretty awkward though. This launch came a day after the US accused China of stealing American IP on an industrial scale, and both Anthropic and OpenAI have previously accused DeepSeek of "distilling" their models.

2. OpenAI launches workspace agents that automate work for businesses

OpenAI has launched workspace agents in ChatGPT, which make it much easier for companies to automate their workflows and share these AI agents with other employees.

The agents are powered by Codex, which is OpenAI's coding model and gives them the ability to actually take actions, rather than simply answering questions.

You can quickly describe a workflow that your team often does - like drafting sales follow-ups, generating weekly reports, or converting product feedback into tickets - and ChatGPT will help you create an AI agent that anyone on your team can use.

For example, Rippling has already built one of these agents to research sales accounts and post deal briefs directly into Slack. The company says it's now saving its sales team around 5-6 hours per week.

It's a direct response to Anthropic's Claude Cowork, which has become very popular in recent months and has a similar set of features.

The bigger question here is whether enterprises actually want to bet everything on one AI company like OpenAI - or whether they'd rather use a more model-agnostic platform that gives them some flexibility.

Notion has been pitching its own agent platform pretty aggressively in recent months. It could quickly become a real alternative for businesses, as you can use almost any model with their own AI agents - not just OpenAI's.

3. Tim Cook is stepping down as Apple CEO

After 15 years of leading Apple, Tim Cook has confirmed that he's stepping down as CEO. He'll be replaced in September by John Ternus, who currently oversees Apple's hardware team.

Cook took over from Steve Jobs back in 2011, at a time when many people thought nobody could really replace him. Looking back, he's done a fantastic job of growing Apple and expanding into new markets.

The Apple Watch and AirPods are great examples of this. AirPods now dominate the global audio market and the Apple Watch makes up around 25% of all smartwatch sales. He also pushed Apple's services business hard, which has grown into a $100 billion annual revenue stream on its own.

The Mac lineup has also had quite the comeback. There was a point where it looked like laptops were being replaced by smaller devices, but Cook's team has turned the Mac into an incredibly powerful device - with global sales rebounding strongly.

Of course, the Cook era hasn't been without its missteps. The Vision Pro was a flop, and Apple has fallen significantly behind on AI - which feels like the bigger problem for Ternus to inherit.

Ternus is a slightly different kind of leader to Cook. He's much more product-focused, having been involved in pretty much every major Apple product over the past decade - including the iPhone, Mac, AirPods, and Apple Watch. Cook, on the other hand, came up through supply chains and operations.

That product-led approach could be exactly what Apple needs right now. The next era of the company will be defined by AI, and Ternus will have to move quickly if Apple wants to catch up with Google, OpenAI, and Anthropic.

4. Meta is tracking what its employees do to train AI agents

Imagine working a job where every keystroke you make is being recorded to train an AI on how to do your job. That's now the reality at Meta.

This new tool is called the Model Capability Initiative, which will allow advanced AI models to understand how people actually work and learn how to complete tasks.

The timing couldn't be much worse. Meta has just announced that it's laying off another 10% of its workforce - around 8,000 employees - on top of the 2,000 cuts already made earlier this year.

The company is also scrapping 6,000 open roles, which is why public job listings have dropped from 800 in March to just seven today.

Workers are basically being asked to train their own replacements, which one Meta employee called "very dystopian" - and I don’t disagree at all.

5. Google says 75% of its new code is now AI-generated

There's a pretty fundamental shift happening in software development right now, as developers are increasingly relying on AI agents to write code for them.

Google says that 75% of all new code at the company is now AI-generated, which is up from 50% last fall and just 25% a year ago.

Sundar Pichai also claims that a recent complex code migration was completed six times faster than it would have been a year ago.

Of course, "AI-generated code" is doing a lot of work in these stats. From my own experience using these tools, it covers everything from autonomous agents writing whole functions to engineers simply accepting autocomplete suggestions.

The headline numbers might be a bit inflated, but the underlying trend is real. Every major tech company is now leaning heavily on AI to write their code, and it's quickly becoming the default way to develop software.

The bigger shift is what this means for software engineering as a profession. Engineers are increasingly moving from writing code line-by-line to directing AI agents and then reviewing their work - which is a fundamental change in the job.

Microsoft's CTO has predicted that 95% of all code will be AI-generated within five years, and if that holds, the skills that have defined software engineering for decades might not be the ones that matter most going forward.

Anthropic adds live dashboards to Claude Cowork

Imagine opening Claude every morning and seeing a dashboard that's already updated with today's calendar, your open tickets, and your team's progress. That's basically what Anthropic is now offering with a new feature, called "live artifacts" in Claude Cowork.

You simply describe what you want and Claude will build a HTML page that lives in its own tab, with version history saved each time you iterate on it.

For example, you could ask for a project tracker that pulls data from Linear and Slack, or a morning brief that finds upcoming meetings in your calendar.

It's actually a pretty clever idea. Most AI tools today force you to ask for the same information over and over again - what's on my calendar today, what tickets are open, what mentions need replying to. With these live artifacts, you can just build the dashboard once and Claude will refresh the data in the background.

If you're using these tools every day, the time savings will really start to add up. I'd recommend starting with something simple, like a weekly summary that pulls in updates from a few of your most-used apps - so you don't have to keep checking five different apps every morning.

There are some caveats though. Live artifacts stay on your computer, can't be shared with teammates yet, and unlike normal Cowork sessions, they use your connectors without asking - so you'll want to be careful with any connectors that can write or modify your data.

But it's a meaningful step forward and makes it much easier for people to interact with these AI tools. Anthropic is steadily turning Claude into a proper productivity platform, rather than just another chat tool. If you're using Cowork already, this is definitely worth a try.

🚫 Norway could ban social media for under-16s

👁️ Six-year-old girl has her sight restored by gene therapy

🔌 Claude can now connect directly to apps like Spotify and Uber Eats

🧠 OpenAI launches GPT-5.5 with better coding and improved efficiency

💼 Google updates Workspace and adds new AI features for businesses

🚀 SpaceX has an option to buy AI coding startup Cursor for $60 billion

🚚 Amazon partners with Sweden's Einride for its electric trucks

🕵️ The NSA is reportedly using Anthropic's Mythos AI model

🎵 Deezer says 44% of songs uploaded daily are AI-generated

🖼️ ChatGPT's Images 2.0 is surprisingly good at generating text

💰 Google invests up to $40 billion in Anthropic, with Amazon adding another $5 billion

☢️ Nuclear startup X-energy raises $1 billion in its data centre IPO

This startup is building tools that give creators much more control over AI-generated images, videos, and audio. Now that creative professionals are really starting to adopt these tools, ComfyUI has just raised $30 million at a $500 million valuation.

They originally started as an open-source project in 2023, at a time when diffusion models were still adding extra fingers to hands. To fix these mistakes, creatives had to constantly regenerate images and had no real control over what they looked like.

To tackle this problem, ComfyUI's founders built a node-based interface that allows creators to connect each step of the generation process. You can then control specific parts of an output, without losing the bits that are already working.

The tool now has over 4 million users across visual effects, animation, advertising, and industrial design. Several studios are now listing "ComfyUI artist" or "ComfyUI engineer" as a formal job title - which just shows how essential their technology has become.

Their pitch is pretty simple. Image generators from companies like Midjourney and OpenAI can get you 80% of the way there, but the last 20% is what really matters - and traditional prompt-based tools make that almost impossible.

ComfyUI's biggest competitor was Weavy, which Figma acquired last year, so this round looks well-timed. In a world that's about to be flooded with AI-generated content, tools that help creators to stand out will be in real demand.

It's great to see ComfyUI secure this funding, and with 4 million users and growing demand from creative studios, they're 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.