AITLDR
Siri's back, Cornelis bets $205M against Nvidia, and Microsoft draws a line

Siri's back, Cornelis bets $205M against Nvidia, and Microsoft draws a line

iOS 27 resurrects Siri, a $205M startup targets Nvidia's AI networking crown, and Microsoft tells its models to stop being evil.

Subscribe free →
Siri Didn't Die — It Just Took Five Years to Get Good

Siri Didn't Die — It Just Took Five Years to Get Good

Apple shipped iOS 27 and macOS 27 Golden Gate today, and the headline feature is a fully overhauled Siri AI that reviewers are actually recommending people use. The new Siri can answer real questions, handle cross-app tasks, and integrates deeply enough that third-party apps like fashion discovery tool Daydream can let users search products by camera roll photo — without opening the app. The update drops Intel Mac support entirely; Apple Silicon only from here.

Why it matters: Apple has 1.5 billion active iPhones — if even a fraction of those users start leaning on Siri daily, Apple's on-device AI play becomes one of the largest AI deployments on earth overnight.


Cornelis Just Raised $205M to Fix the Part of AI Infrastructure Nobody Talks About

Cornelis Just Raised $205M to Fix the Part of AI Infrastructure Nobody Talks About

Cornelis Networks, which builds high-speed networking fabric that lets AI chips talk to each other faster, announced a $205 million funding round. The problem it's solving — chip interconnects — is increasingly the bottleneck as clusters scale beyond what Nvidia's own InfiniBand and NVLink can elegantly handle. More chips means more time wasted on chips waiting for each other.

Why it matters: Nvidia dominates the GPU market, but the interconnect layer is a real vulnerability — specialized networking startups that crack this could become critical infrastructure for every hyperscaler building out the next generation of training clusters.


Microsoft Published a "Code of Conduct" for Its AI Models — Yes, Really

Microsoft Published a "Code of Conduct" for Its AI Models — Yes, Really

Microsoft released a formal AI code of conduct that tells its models to not hack systems, not deceive users, and to support humans rather than replace them. It's a mix of high-minded principles ("accelerate human flourishing") and concrete safety guardrails baked into how the models are supposed to behave at a policy level.

Why it matters: A published code of conduct is only as strong as its enforcement mechanism, but it does create a paper trail that regulators, enterprise customers, and competitors can hold Microsoft to — which may be exactly the point.


Superhuman Buys Meeting Notetaker Fathom to Build an Agentic Productivity Stack

Superhuman Buys Meeting Notetaker Fathom to Build an Agentic Productivity Stack

Email startup Superhuman acquired Fathom, a YC-backed AI meeting notetaker with 400,000 monthly active users. The deal is a direct bet that the next battleground for productivity software isn't features — it's owning the full loop of communication: email, meetings, and whatever autonomous tasks come next.

Why it matters: Standalone AI notetakers are already being commoditized by Zoom, Google Meet, and Teams; Fathom's best shot at survival was getting acquired by a product that could plug meeting context into a broader agentic workflow, and Superhuman needed exactly that data layer.


Waymo Now Operates in 15 Cities After Las Vegas Launch

Waymo Now Operates in 15 Cities After Las Vegas Launch

Waymo opened commercial robotaxi operations in Las Vegas, making it the company's 15th market. Vegas is a strategically interesting choice — high tourist volume, grid-style streets, and a city government that's been aggressively courting autonomous vehicle operators for years.

Why it matters: Waymo's methodical city-by-city expansion is quietly building the operational data and regulatory playbook that any competitor trying to catch up will need years to replicate.


Jensen Huang Took Trump's Call Live Onstage — Then Talked Robots

Jensen Huang Took Trump's Call Live Onstage — Then Talked Robots

Nvidia CEO Jensen Huang put President Trump on speakerphone during the All-In Summit, in what has become a running motif of the Huang era. Beyond the theater, Huang used his time onstage to argue that robots won't take over the world — a message that conveniently aligns with Nvidia's pitch that more AI infrastructure spending is good, actually.

Why it matters: When the most important chip CEO in the world is taking presidential calls live and opining on robot risk, it's a sign of how thoroughly AI infrastructure has become a geopolitical and political asset, not just a tech story.

Quick Hits

  • →Insight Partners' Deven Parekh explains why his $90B firm is staying diversified while rivals pile into OpenAI and Anthropic — a contrarian read worth your time. TechCrunch
  • →Larry Ellison canceled a planned $7.5B Oracle stock sale — no explanation given, which is the kind of move that makes analysts very curious. TechCrunch
  • →VCs picked their 9 buzziest startups from YC's latest Demo Day batch, including brain chips and floating nuclear reactors. TechCrunch
  • →Fashion app Daydream is one of the first third-party apps to ship a real Apple Intelligence integration, turning camera roll photos into shoppable results via Siri. TechCrunch

AI TLDR

Get AI Daily in your inbox every morning — free

Free forever·Unsubscribe anytime·5 min read
Share this edition: