What do you think of the new look?

Spent most of the week in Lisbon speaking to a B2B manufacturing giant about why disruption is an invitation versus a threat. Wrapped a five-year outlook on blockchain (spoiler: fragmentation, consumer hype decline), prepped for SXSW London, and had sponsor talks for the upcoming AI salon (there are still opps if you are interested in some thought leadership). I also put two TNN sessions publicly live. Book now as they’re going to fly; Google / NotebookLM (July 17) and Shopify (September 18).

Hit reply and let me know what you think of the new look, what you’re going to see at SXSW, and who you want to see next at TNN.

Apple shipped the iOS 19 beta with Gemini, GPT-4o, and Claude integrations ahead of WWDC. Siri’s overhaul was confirmed as an interface layer over third-party models, while Apple’s in-house LLMs will open to developers. Tim Cook privately lobbied Texas lawmakers to block child safety legislation, and Trump renewed pressure to stop Apple’s India manufacturing shift. OpenAI launched a Mac app and expanded GPT-4o access across tiers. Reka AI debuted, launched by ex-Stability execs targeting Claude and ChatGPT users. Anthropic restricted Claude 3 Opus access and launched Claude 3 Haiku in the EU with local data compliance — a strategic regulatory play.

Google rolled out Gemini to Chrome, gave students free access to Gemini Advanced. A joint DOJ–FTC antitrust investigation into Google’s AI dominance marked a new regulatory era for foundation models, and the company will find out its antitrust fate in August. YouTube banned monetisation of AI-generated voice clones, targeting impersonations of celebrities and artists, a defining move in AI content moderation.

Tesla cut full self driving to $99/month to encourage adoption. Nvidia delayed Blackwell shipments to 2026 due to production limits. AMD surged past Intel in valuation, a historic moment for both companies. The U.S. military ran GPT-4 in war games, testing its use in high-pressure decision environments. X sued Bright Data over scraping, aiming to claim ownership over publicly accessible training data. Kenya scrapped an AI infrastructure deal with the UAE after internal political blowback. WormGPT, a black-market LLM used for fraud and phishing, was taken offline, highlighting growing pressure to police rogue models.

DID APPLE JUST SCREW OVER OPENAI?

While the CEO of Anthropic was predicting a bloodbath for 50% of white-collar jobs due to AI, Apple was quietly rewriting how billions will use it, via an iOS 19 beta update that revealed native integrations for Gemini, GPT-4o, and Claude. All of which confirmed Apple’s shift to a multi-model strategy ahead of WWDC (previously stated, though many assumed it was years away). Siri won’t just be powered by an Apple model, it will act as a smart interface, routing tasks to different foundation models behind the scenes. Not brilliant news for OpenAI as it’s more competition, but considering last week’s $6.5 billion Ives purchase, not unexpected either.

Far from leader, Apple is positioning itself as the broker between users and the model layer, with Siri becoming the control surface for daily AI use. Part smart move, but more a necessity. Apple just doesn’t have the AI chops yet. Code suggestions might come from GPT-4o, travel recaps from Claude, and visual answers from Gemini, all surfaced through Apple’s interface, framed by its privacy guarantees, and optimised for its hardware. Consumers trust Apple and Tim Cook, so it’s not a terrible move for all parties.

Apple isn’t trying to win the AI race on model performance. Instead, Cook and co are aiming to own the layer that decides which model gets used, and how. That turns the competitive logic on its head. Where OpenAI, Google, and Anthropic are battling over capability, Apple wants the interface, and by extension, the user relationship.

SO WHAT?

Apple wants an interface war not a model war which it knows it would instantly lose. OpenAI and Ives are clearly thinking no-phone which is equally unattractive to Apple as it nuked their +$200 billion per annum revenue money-maker. So instead, they’ll work with everyone. Foundation models are not user-friendly. What matters now is who controls the user-facing layer, how trust is established, and where the invocation (i.e. ‘Hey Siri’ or ‘Alexa’) happens. Going with multiple providers positions Apple not as a model company but as an AI broker, a gateway that selects and packages model output in a way that feels cohesive, private, and aligned with platform norms that is seamless to the user (aka the ‘magic’ Apple loves). The company is leaning into what it already does best: now be first, own the customer relationship, manage privacy expectations, and deliver polish at scale.

The move is also interesting when it comes to antitrust, just how Apple remains unbiased and transparent while Google pushes $20 billion a year to them will be watched closely. Being “the best model” may not matter if Apple routes only specific queries your way and masks attribution behind Siri branding.

[DO] Optimise for low-friction invocation. Siri is being redesigned as the entry point for AI across iOS. Developers who reduce steps between intent and outcome now will fit better into Apple’s UX vision later.

[DON’T] Assume Apple is quietly building a foundational model to compete with OpenAI or Google. The company is playing a different game entirely. Underestimating how quickly Apple can change the power dynamics in AI simply by re-routing demand would be a mistake.

WD_DTW? is a weekly intelligence series tracking the moves of OpenAI, Amazon, Google and Meta (coming soon); what they shipped, what it signals, and why it matters. £15pm (per title), or £49.99 a month for all titles, plus you get TBD+ membership too.

Electronic tattoos are getting used more and more for mental stress. /12 mins

Knowledge work is dying, but here’s what comes next. /18 mins

Internal docs show Meta plans to use AI to automate up to 90% of its privacy and integrity risk assessments, including in sensitive areas like youth risk. /8 mins

Can a photograph help predict who will survive cancer treatment?. /6 mins

Four AI agents raised $2000 for charity, and scientists are . /9 mins

In a world first, Brazilians will soon be able to sell their digital data. /8 mins

A superbug that can feed on plastic is becoming a problem for hospitals. /8 mins

Mary Meeker’s +350 page report used the word unprecidented +50x. /60 mins

Return-To-Office rates have remained steady at 40% for a year, companies are looking to offload 30m sq ft by 2030. /5 mins

How design makes you feel things. /15 mins

Getting value here? TBD+ gives you more: real access, real people, real events. Join us.

TBD+ is available on desktop and mobile app (iOS + Android)

// C_NCENTRATE is written and curated by Paul Armstrong

I deliver actionable insights that keep companies ahead of the disruption that’s coming. As the founder of TBD Group and author of Disruptive Technologies, I am trusted by global brands, agencies, and when breaking news hits the FT, WSJ, BBC, and CNN ask for my analysis. Find out more and connect with me on LinkedIn and Bluesky.