Google DeepMind shake-up: Jeff Dean exits, D-Wave quantum leap

Samantha: Welcome to The State of Tech, The European Edition, Thursday August six, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: a major leadership shake-up at Google DeepMind, D-Wave's quantum error correction breakthrough, the widening rift inside Silicon Valley shaping US AI policy, Apple's memory crisis rippling across the tech industry, and to close out, two things you can actually use today, the free open-source screenshot annotator Flameshot and the privacy-friendly weather app Tomorrow.io. Let's start with the shake-up at Google DeepMind.

Google DeepMind loses AI veteran Jeff Dean as Demis Hassabis steps back into a chairman role.

Samantha: Google is going through a serious reshuffle at the top of its AI operation. Jeff Dean, one of the architects of Google's AI strategy going back decades, is leaving to start his own company, and he's taking several senior colleagues with him. At the same time, Demis Hassabis, who has been running Google DeepMind, is moving into a new role as chairman of the lab and chief scientist of Alphabet. Day-to-day control of DeepMind goes to Koray Kavukcuoglu, who has been promoted to senior vice president.

Bob: This is a big deal because Google DeepMind is not just any AI lab. It's the engine behind Gemini, behind a lot of the research that other labs build on, and behind Google's push to keep pace with OpenAI and Anthropic. Losing Jeff Dean is significant on its own, he's been shaping how Google trains large models for years. When you combine that with Hassabis stepping into a more strategic seat, you're looking at a real generational handover. Kavukcuoglu is a known quantity inside DeepMind, but the outside world is going to want to see whether he keeps the same research culture intact.

Samantha: For European researchers and companies, this matters more than it might look. A lot of European AI work sits on top of Google's open research, from published papers to the tools DeepMind releases. If the pace or direction of that research shifts, universities in Amsterdam, Zurich, and Paris feel it quickly. And there's a competitive question too. If Jeff Dean's new startup pulls talent out of Google, some of that talent could look at European hubs like London or Paris, where AI regulation is stricter but the funding environment has been warming up.

Bob: There's also the practical angle for European companies that already use Google's AI products. Enterprise customers signing multi-year deals with Google Cloud want stability at the top. Leadership churn at this scale usually triggers a wait-and-see moment, where big customers pause new commitments until they see the roadmap. Expect Microsoft and Anthropic sales teams to be on the phone this week, using this as an opening. Whether that actually shifts market share depends on how quickly Kavukcuoglu can put his own stamp on things.

D-Wave publishes a quantum error correction breakthrough in Nature, cutting hardware overhead for fault-tolerant systems.

Samantha: D-Wave has published research in the journal Nature showing a fast, high-accuracy two-qubit gate that preserves the error-correction advantages of its superconducting architecture. In plain terms, quantum computers make a lot of mistakes, and correcting those mistakes has always required an enormous amount of extra hardware. D-Wave says its approach substantially reduces that overhead, which is one of the biggest barriers between today's experimental machines and something you could actually use commercially.

Bob: The reason this matters is that quantum has been stuck in a familiar loop for years. Every few months a company announces a milestone, and every few months critics point out we're still nowhere near a machine that solves a real business problem better than a classical computer. Error correction is genuinely one of the harder walls to break through, so a peer-reviewed result in Nature carries more weight than a press release. It doesn't mean commercial quantum is here tomorrow, but it does narrow the gap.

Samantha: For Europe, the timing is interesting. The EU has been pouring money into quantum through its Quantum Flagship programme, and there are serious players in Germany, France, and the Netherlands working on similar problems. A breakthrough from a North American company puts pressure on European labs to show comparable progress, and it also gives European quantum startups a benchmark to argue for more funding. Pharmaceutical companies in Basel and Leiden, which have been experimenting with quantum for drug discovery, are watching this closely.

Bob: The honest caveat here is that Nature publication is a research milestone, not a product launch. Scaling this from a lab demonstration to a machine running real workloads is a multi-year process. But the direction of travel is what markets and policymakers are reading. If error correction gets cheaper faster than expected, the timeline for post-quantum cryptography, meaning the rush to protect data against future quantum attacks, gets more urgent for European banks and governments.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

The AI industry's internal rift over safety and open-source is now driving US policy debates in Washington.

Bob: The long-running fight inside the AI industry, between people who want to move fast and people who want to slow down, and between open-source advocates and closed-model companies, has moved from lab cafeterias into US Congress and the White House. A recent letter signed by Nvidia, Microsoft, Meta, IBM, Dell, and Palantir argued against broad restrictions on open-weight Chinese AI models. That's a striking coalition, and it shows the industry doesn't speak with one voice when lawmakers come asking for advice.

Samantha: The interesting thing is which companies signed that letter and which didn't. You have chipmakers and enterprise software vendors saying, don't cut us off from Chinese open models, we need the ecosystem. Meanwhile, the labs that build proprietary frontier models tend to argue for tighter controls, which not coincidentally also raises the barrier for competitors. This is industrial policy dressed up as safety debate, and Washington is starting to see through it.

Bob: For European regulators, this US split is genuinely useful. Brussels has been criticised for moving fast on AI rules while the US moved slowly, but the EU AI Act is now the reference point that everyone else has to react to. If the US ends up with a patchwork of state-level rules because industry can't agree on a federal framework, that actually strengthens Europe's hand as the place with a single, predictable rulebook. Companies that hate the AI Act in principle often prefer it in practice, because at least they know what they're dealing with.

Samantha: There's also a real question for European companies about which side of this argument to align with. A lot of European AI startups depend on open-weight models, whether from Meta, from Mistral in France, or from Chinese labs. If Washington restricts access to Chinese open models, European firms could benefit from the gap, or they could get caught in the crossfire of export controls. That uncertainty is already showing up in how European venture capital is pricing AI deals.

Apple's memory supply crisis signals a broader tech industry squeeze that could push iPhone prices up by three hundred dollars.

Bob: Apple is dealing with a serious shortage of memory chips, and analysts think it could push new iPhone prices up by as much as three hundred dollars per model. That's not just an Apple problem. A Global Electronics Association report says sixty-two percent of electronics manufacturers are already facing constrained component availability or longer lead times, and eighty-two percent expect prices to rise, with about a third expecting significant increases. This is a supply chain story that touches everyone who buys a phone, a laptop, or a game console.

Samantha: What's driving this is a mix of factors. Memory manufacturers have been redirecting capacity toward high-bandwidth memory for AI servers, which pays much better than the memory that goes into consumer devices. So the same boom that's building giant data centres is now squeezing the supply of chips for regular products. Apple has the margins to absorb higher costs, or to pass them on and still sell every iPhone it makes. Smaller European electronics brands do not have that cushion.

Bob: For European consumers, this shows up in two ways. First, prices on phones, laptops, and even smart home devices are likely to creep up over the next six to twelve months. Second, availability could get patchy, especially for mid-range brands that don't have Apple-scale purchasing power. European retailers are already flagging longer delivery times for certain models, and that's before the holiday shopping season really kicks in.

Samantha: The bottom line is that the AI infrastructure build-out has knock-on effects far outside the AI industry. Every gigawatt of data centre capacity being planned pulls memory, power, and skilled labour away from other parts of the tech economy. European policymakers focused on AI competitiveness need to think about this too, because if European consumers are paying three hundred dollars more for a phone to subsidise American AI training runs, that's a political problem waiting to happen.

The free open-source screenshot tool Flameshot turns any screen grab into an annotated, shareable image in seconds.

Bob: To close out, two things you can actually use today. First up is Flameshot, a free open-source screenshot tool that works on Windows, Mac, and Linux. What makes it stand out is what happens after you take the screenshot. Instead of just saving the image, Flameshot opens an immediate annotation layer where you can draw arrows, highlight areas, blur out sensitive information like email addresses, or add text captions, all before the file even lands on your disk.

Samantha: The reason people recommend it is the workflow. Built-in screenshot tools on most operating systems are fine for capturing an image, but if you want to circle something and send it to a colleague, you have to open a separate editor. Flameshot collapses that into one step. Users report it saves several minutes per screenshot when they're doing something like reporting a bug, walking a family member through a settings screen, or explaining a document to a colleague on Slack.

Bob: A few practical details. Flameshot is genuinely free, there's no premium tier, no account, and no data leaves your machine. Installation takes about a minute from flameshot dot org, or you can grab it from your Linux package manager. You can bind it to a keyboard shortcut, so hitting the key immediately opens the capture-and-annotate mode. There's also a useful option to upload directly to a cloud service if you want a shareable link, but by default everything stays local.

Samantha: For European users, the privacy angle is worth noting. Because Flameshot runs entirely on your own device and doesn't require an account, it's a good fit for people who work with confidential documents, healthcare information, or client data. Reviewers particularly praise it for teams that have moved away from proprietary screenshot tools that phone home. It's the kind of small utility you install once and use every day without thinking about it.

The weather app Tomorrow.io gives you hyperlocal forecasts with radar-style precision, free on iOS and Android.

Bob: Second thing to try today is Tomorrow.io, a weather app available on iOS and Android that has been getting attention for its hyperlocal forecasts. The company runs its own satellites and combines that with ground sensors, which lets it give you minute-by-minute rain predictions for your exact location, not just for your city.

Samantha: What makes it different from the built-in weather app on your phone is precision. Users report that when Tomorrow.io says rain will start at 3:47 PM, it tends to be right within a few minutes. For cyclists, runners, or anyone planning outdoor activities across Europe's famously unpredictable weather, that kind of accuracy is genuinely useful. The free tier gives you forecasts, radar, and severe weather alerts, which covers most everyday needs.

Bob: There is a paid tier for people who want extended forecasts and specialised alerts, but the free version is what most listeners will want. Setting it up takes a couple of minutes, you download the app, allow location access if you want automatic updates for wherever you are, and that's it. Reviewers particularly praise the pollen and air quality layers, which matter for people with allergies during European spring and summer.

Samantha: One caveat: Tomorrow.io is a US company, so if you want a fully European alternative, apps like Meteoblue from Switzerland cover similar ground. But for accuracy in the moment, especially if you're trying to squeeze a bike ride in between showers, Tomorrow.io is the one people keep coming back to.

Bob: Today we covered the leadership shake-up at Google DeepMind, D-Wave's quantum error correction breakthrough, the Silicon Valley rift shaping US AI policy, Apple's memory crisis rippling through the tech industry, and two things to try yourself: Flameshot and Tomorrow.io.

Samantha: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.