Today's spotlights:
Google will now save your search data for AI Training
Claude is quietly winning over consumers of ChatGPT
Tri-Fold War: Samsung Galaxy Z TriFold vs Huawei Mate XT
IBM unveils world's first sub-1nm chip
Apple's "iRing" could be coming to challenge Samsung and Oura
Read time: 5 minutes
Tech Stories This Week
image is AI generated
Google is expanding what it saves from your search activity and using it to train AI. Beyond just text searches, Google will now retain media from your interactions, including:
Images uploaded for Google Lens or reverse image search
Audio/video from Search Live or Translate speaking practice
Voice searches and uploaded files
Why it matters: AI models need diverse data, audio, video, images not just text, to improve. This move gives Google a major data advantage over rivals like OpenAI and Anthropic, thanks to its massive cross-service user base.
The setting: A new "Search Services History" option will appear in Google Account settings enabled by defaultst users. (If you've already turned off Web & App Activity and Search Personalization, you may not see this prompt at all.) for mo
How to opt out: Go to your Google Account → Settings → look for the new pop-up/tab on Search Services History → toggle it off. It can be changed anytime.
My Take: This rollout reflects a broader pattern, platforms increasingly opt users into AI training by default, shifting the burden onto individuals to protect their own data. Critics warn this is fueling growing user fatigue around privacy, making people less likely to actively manage their settings at all.
image is AI generated
New transaction data from Indagari shows everyday consumers are increasingly choosing Claude over other paid AI tools, challenging the assumption that Anthropic's customer base is mostly developers and enterprise users.
The key signals:
"Claude" has overtaken "AI" as the most searched term on DataCamp
Demand for Claude courses is beating ChatGPT 3-to-1 among self-directed learners
Course interest in Claude has spiked 18x in the past 30 days
ChatGPT still leads in total paying users and dominates corporate training but its growth has slowed simply because it's already so large.
The timing is notable: Both OpenAI and Anthropic are inching toward going public, making real revenue signals more valuable than ever.
Wildcard: Anthropic is also navigating fallout from a U.S. government ban restricting international access to its top models, Mythos 5 and Fable 5 forcing the company to pull them from the market entirely for now. How that affects momentum remains to be seen.
Gadgets update
The foldable phone race just added a third fold and Samsung and Huawei are the only two players in the game.
Design: Samsung's Z TriFold only works fully folded or fully open. Huawei's Mate XT offers three usable modes, including a middle dual-panel view Samsung skips entirely.
Display: Samsung wins clearly: 120Hz and 1,600 nits vs Huawei's 90Hz and 850 nits.
Performance: No contest. Samsung runs Snapdragon 8 Elite; Huawei is stuck on an older 7nm chip due to US trade restrictions.
Price: Both phones sit around $2,500–$3,100 squarely in luxury territory, not mass-market.
My Take: Neither phone has a true global release. Samsung is rolling out region by region; Huawei hasn't confirmed international availability at all. Until that changes, this "war" is really a battle for early adopters with deep pockets.
image source: ibm.com
IBM introduced the world's first sub-1nm chip technology, the 0.7nm (7 angstrom) node, pushing past what many considered the physical limits of chip scaling.
Density: Packs nearly 100 billion transistors onto a chip the size of a fingernail, almost 2x denser than IBM's 2nm chip from 2021.
Performance gains: Up to 50% more performance or 70% greater energy efficiency compared to IBM's 2nm chips, a major boost for AI, cloud infrastructure and next-gen devices.
The innovation: A new 3D transistor architecture called "nanostack" the industry's first 3D, nanosheet-based design. It vertically stacks transistors and allows different materials in each layer, optimizing power and performance independently.
Validated, not just theoretical: The architecture was tested via ultra-thin dielectric bonding, dual-channel engineering, and working CMOS inverter operation, proving it can be built and can actually compute.
Bonus win: New research at VLSI 2026 showed 40% SRAM scaling, enabling more efficient chips that can handle AI's heavy data demands.
Why it matters: This marks the first time logic technology has gone below 1nm, entering the angstrom-scaling era and IBM's roadmap projects at least a decade of further scaling ahead.
image is AI ganerated
Apple may finally be entering the smart ring race. According to tipster Kosutami, Apple is developing a smart ring dubbed the "iRing." The leak is thin, no specs, no features, no launch date, suggesting the project is still in early stages.
Classic Apple playbook: The company rarely moves first into new categories, preferring to watch competitors, study what works, then enter with its own polished take. We're seeing the same pattern with its long-rumored foldable iPhone, arriving years after Android rivals already built out the space.
What we know so far:
Earlier patent filings hinted at Apple's smart ring ambitions, suggesting it could include:
Health and fitness tracking sensors
Hand-detection support for Apple Vision Pro
A screen-free, lower-cost alternative to the Apple Watch
The competition it would face:
If real, the iRing would go up against two well-established players:
Oura Ring: The category's most recognized name
Samsung Galaxy Ring: Launched in 2024 at Rs. 38,999, available in 9 sizes, with sleep score tracking, snoring analysis, and IP68 water/dust resistance
For now, the iRing remains speculation but if Apple's history with foldables is any indicator, it likes to arrive late and aim to dominate.