Firefox adds shake to summarize AI feature on iPhone

☕ Good morning,

There's something magical about morning energy before the world gets too noisy. It's when the simplest ideas feel revolutionary and the biggest challenges seem surprisingly manageable. Maybe today's breakthrough is hiding in plain sight, waiting for you to notice it.

_{—Here’s to the first sip.}

TOOL BAR

1. TwinMind: AI that understands your life and helps you proactively

2. Genstore.ai: Start selling online with just a prompt

3. 10X PPC: Launch PPC Campaigns 10X Faster

TODAY IN AI
Firefox adds “shake to summarize” AI feature on iPhone

Firefox is about to get a quirky new trick: shake your iPhone and it’ll spit out an AI summary of the page you’re on. The feature rolls out this week, powered by Apple’s on-device AI for iPhone 15 Pro and newer, once iOS 26 drops.

If you’re on an older device, Firefox will lean on Mozilla’s own cloud AI to generate the summary, which you can also trigger by tapping the thunderbolt icon or the menu. It only works on pages under 5,000 words, and you can switch it off if you don’t want accidental shakes turning into summaries.

This makes Firefox one of the first major apps to plug into Apple Intelligence, with plans to expand beyond the US, add more languages, and eventually bring the feature to Android. It’s a small but clever way to make AI feel like part of everyday browsing.

 FAST BARISTA

● AMD has enabled FSR 4 AI upscaling and frame-gen for most FSR 3.1 DirectX 12 games. The tech uses ML to predict missing pixels, promising sharper visuals and higher performance. With the new driver, gamers can force FSR 4 on 85+ titles via the Adrenalin app toggle. FSR 4 already powers big games like Cyberpunk 2077, Ghost of Tsushima, and F1 25.

● Nvidia will launch its new Rubin CPX AI chip in late 2026, built for heavy tasks like video and software generation. The design splits input processing from response generation, making inference more efficient. Nvidia claims $100M of CPX hardware could drive $5B in revenue for customers. CEO Jensen Huang says CPX is the first chip made for AI models handling millions of tokens at once.

● Google has upgraded its Veo 3 AI video generator with 1080p resolution and 9:16 vertical video support. The update lets developers create mobile-friendly videos for apps like TikTok, Reels, and Shorts. Pricing is now cheaper too — Veo 3 dropped to $0.40/sec and Veo 3 Fast to $0.15/sec. Google says both models are now stable and ready for large-scale use in the Gemini API.

TECH BARISTA
Spotify brings lossless audio to Premium

Spotify is finally rolling out lossless audio for Premium users after years of delays. Starting this month, subscribers in 50 markets, including the US and UK, will see new quality settings from Low to Very High and Lossless.

The feature won’t cost extra and works on phones, desktops, and tablets, with the best experience on Wi-Fi and wired headphones. Spotify’s lossless caps at 24-bit/44.1 kHz FLAC, not as hi-res as Apple or Tidal, but still a big step up from today’s streams.

For most listeners, this upgrade is less about audiophile specs and more about finally getting the sound quality Spotify has been promising since 2017.

MORE TO KNOW

● Google Cloud says it has $106B in customer contracts lined up, with $58B set to become revenue in the next two years. CEO Thomas Kurian said demand is growing even faster than current revenue, driven by AI services. He pointed to Google’s custom chips and software as strengths against Amazon and Microsoft. The business made $13.6B in Q2, up 32%, and now runs at over $50B in annual revenue.

● Jaguar Land Rover says a cyberattack forced it to halt production and confirmed “some data” was stolen. The breach has disrupted sales, supply chains, and even vehicle repairs. U.K. regulators must be notified within three days, but it’s unclear whose data was taken. Officials warn recovery could take weeks, raising concerns over economic fallout.

GADGETS BARISTA
Apple event highlight

Apple’s fall event was packed, but the spotlight went to the iPhone Air, Apple’s thinnest phone yet at just 5.6mm. It has a wide “camera plateau,” a 6.5-inch ProMotion display, and runs on the new A19 Pro chip. Starting at $999, it lands September 19th alongside accessories like a MagSafe battery pack and translucent cases.

The regular iPhone 17 finally gets ProMotion on its 6.3-inch OLED and packs the A19 chip, while the iPhone 17 Pro goes all out with three 48MP cameras, a new heat-dissipating aluminium unibody, and Apple’s biggest iPhone battery yet. The Pro starts at $1,099, and the Pro Max can stretch to 2TB of storage.

On wearables, Apple introduced the Watch Series 11 with 5G and hypertension alerts, the Watch Ultra 3 with Apple’s largest watch display and satellite texting, and the Watch SE 3 with new sensors and always-on display at $249.

The lineup closed with AirPods Pro 3, same look, but upgraded ANC, better bass, live translation, and even a built-in heart rate sensor. Like everything else announced, they ship September 19th.

_{^{Written by Gadget Face}}

FAST FLASH

● Nvidia has released its latest Game Ready driver (581.29) ahead of Borderlands 4. The update adds DLSS 4, Multi Frame Generation, and Reflex support for the new game. Dying Light: The Beast also gets optimizations with full DLSS features before launch. Nvidia updated RTX Remix too, letting modders enhance smoke and fire in 165+ classic games.

● Arm has unveiled its next-gen Lumex chip designs, built to run AI directly on phones and wearables. The lineup includes four versions, from ultra-efficient chips for watches to high-power ones for flagship smartphones. Lumex enables large AI models to run on-device without relying on the cloud. The designs, made for 3nm manufacturing, aim to speed up adoption by handset makers worldwide.

STARTUP BAR
TwinMind raises $5.7M for AI “second brain” app

Three former Google X researchers are building TwinMind, an AI app that acts like a “second brain.” It runs quietly in the background, capturing your conversations (with permission) and turning them into notes, tasks, and answers.

The startup just raised $5.7M at a $60M valuation in seed funding and rolled out a new speech model, Ear-3, which supports 140+ languages. TwinMind can run 16–17 hours on-device without draining your battery, deletes raw audio instantly, and keeps only text locally for privacy.

With over 30,000 users already, TwinMind goes beyond meeting note apps by passively building context all day, blending offline conversations with browser activity to create a structured memory you can tap into.