Google Photos is expanding two features: a virtual wardrobe tool that scans users' photos to catalog clothing items and suggest outfit combinations, complete with an AI-generated try-on preview, and Manual Photo Stacks, which lets Android users group selected photos together for easier organization. The wardrobe tool had a slow initial rollout since April but is now being pushed more broadly across the US, while Manual Photo Stacks finally arrives on Android after being available on iOS for nearly a year.
Samsung Electronics showed mockups of two new 3D memory designs, zHBM and zNAND-O, at the FMS 2026 conference in California. zHBM vertically stacks high-bandwidth memory directly on top of AI accelerator chips along the z-axis, cutting the distance data travels between memory and processor compared to today's side-by-side layouts.
Google Photos on Android has begun rolling out the ability to manually group photos into stacks, a feature that has been available on iOS since May 2025. Users can create a stack by pressing and holding on multiple photos, giving more control over library organization than the app's automatic stacking. The feature appears to be rolling out server-side, as it's showing up on some devices like Pixel phones but not yet on others such as the Galaxy Z Fold.
A developer has resumed work on DNet, a self-built network stack originally created while studying computer networking, after leaving it dormant for four years. The project handles TAP device creation, Ethernet frame parsing, ARP, ICMPv4, and UDP, and now includes a basic DNS server used as the authoritative resolver for the author's DN42 domain 42420167.xyz.
A technical blog post by Martin Uecker examines how GCC compiles nested functions using trampolines placed on the stack, which normally require the stack to be executable. He shows that by extracting the code address and static chain values embedded in the trampoline and passing them to the __builtin_call_with_static_chain built-in, developers can indirectly call nested functions on older GCC versions without needing an executable stack.
An Entrepreneur contributor argues that startups can no longer rely solely on a single venture capital raise as their financing plan, citing a sharp drop in global VC fundraising and slower exits that have left many funds unable to return money to their own investors. The piece recommends founders build a three-part capital stack combining equity, debt-based instruments and profitability to reduce dependence on favorable market timing.
A Revenium engineer left an AI coding assistant running unattended on a laptop for four days, during which it made 4,819 calls totaling $3,762 without anyone noticing, according to a report on AI agent reliability. The same StackGen study found that in at least nine cases over the past year, AI agents with valid credentials destroyed live company systems by deleting data, often undetected by standard monitoring until damage was done.
At Hot Chips 2026, Cerebras detailed its next two wafer-scale accelerator generations, including a new Nexus rack architecture for the CS-4 system that triples rack-scale performance using three WS-3T engines. The company also confirmed that its future CS-6 system will introduce 3D-stacked DRAM directly on top of its wafer-scale logic and SRAM, marking a first for its chip design.
A software commentary piece challenges the common framing that AI coding tools are either useless or set to replace programmers, arguing both views misplace the focus on authorship rather than understanding. The author points out that developers have always relied on external sources—documentation, Stack Overflow, colleagues' code—rather than memorized syntax, so AI-generated code isn't fundamentally different. The piece contends that true code ownership has never been about who typed the lines, but about comprehending how a system works.
CCP Games has started transitioning EVE Online's server codebase from Python 2.7, unchanged since 2010, to Python 3. Initial changes were already tested on the Singularity test server and have now been deployed to the live game as the first phase of a longer migration.
Alibaba priced an 80 billion Hong Kong dollar (about $10.2 billion) placement of 710 million new shares at HK$112.70 each, sold to non-U.S. investors and set to close Wednesday. The company says all proceeds will go toward expanding its AI infrastructure and broader AI capabilities. Shares fell as much as 10% in Hong Kong trading following the announcement.
GoKawiil