The team behind the open-source Bomfather eBPF security agent found that the costliest part of their file-access enforcement wasn't applying allow/deny decisions, but repeatedly walking parent directories to determine which path-based policy applied to a given file. By caching the resolved policy per inode instead of recalculating it on every file open, they cut kernel-side CPU overhead by roughly 90%, particularly benefiting workloads like databases that repeatedly access the same file paths.
An independent developer known as Speedstu built a PowerShell-based toolkit that combines ZLUDA with AMD's HIP/ROCm SDK, allowing CUDA-only applications to run on AMD GPUs in Windows without dual-booting or virtual machines. The setup successfully mapped CUDA's driver API and libraries like cuBLAS, cuSPARSE, and cuFFT to AMD equivalents, and was tested by training a small reinforcement-learning model on a Radeon RX 9060 XT using unmodified CUDA code.
The 2026 Emmy Awards will take place on Monday, September 14, at the Peacock Theater in Los Angeles, honoring the year's best in television. Details on how audiences can watch the ceremony live have been outlined ahead of the event.
IBM is hosting an interactive webinar aimed at higher education students, particularly IEEE chapter members, covering the basics of AI types, common concerns, and IBM's approach to Responsible AI. Attendees will apply these principles to a real use case involving IBM Bob, a software development life cycle agent, and can participate via live polling and Q&A.
Google has released XLS, an Apache 2.0-licensed toolchain that converts high-level functional descriptions into synthesizable Verilog and SystemVerilog hardware designs. The tool lets engineers write logic once and run it both as native host software and as generated hardware, with correctness guarantees ensuring the two versions behave identically.
A researcher used AI to scan thousands of digitized construction trade journal issues dating back to the 1850s, tracking every historical attempt to automate building tasks. The analysis found that successful automation clusters into two categories: off-site factory processes like steel fabrication and precast concrete, and on-site tasks simple enough to be performed by a machine repeating the same motion along a fixed path. Nearly everything outside these two categories failed to gain lasting adoption.
U.S. District Judge Carl J. Nichols issued a preliminary injunction stopping the Postal Service from implementing new mail-voting regulations tied to Trump's executive order, mirroring an earlier block by a Boston judge. Nichols ruled that without the injunction, many valid absentee ballots could go uncounted in the midterms. The case comes as the Supreme Court is already reviewing a similar injunction from a separate lawsuit.
New federal rules, including a proposed $100,000 fee on H-1B visa applications introduced in September 2025, are adding to a string of immigration restrictions on skilled foreign workers. IEEE Spectrum found that U.S. tech companies are responding by expanding hiring and opening offices abroad rather than sponsoring visas domestically, though none of the 25 firms contacted would discuss the shift publicly.
Security researcher Bill Swearingen showcased a reinforcement-learning system at DEF CON that generates geometric patterns capable of fooling object-detection AI, including systems built on the popular YOLO framework. He tested the patterns against 11 face- and person-detection models, successfully lowering confidence scores or evading detection entirely, while companies like Cap_able and Urban Privacy already sell clothing woven with similar disruptive designs.
Cultured Code released Things 3.24, adding compatibility with iOS 27, iPadOS 27, watchOS 27, and macOS 27 Golden Gate on the same day Apple rolled out those updates. The release brings full integration with the revamped Siri and Spotlight search, an extra-tall widget option, and expanded notification snooze durations.
A newly published technical handbook explains FlashAttention, the exact attention algorithm that speeds up Transformer models by minimizing memory traffic rather than reducing computation. It walks through the standard attention implementation's repeated reads and writes to GPU high-bandwidth memory, then details how tiling, online softmax, and recomputation eliminate most of that overhead across FlashAttention-1 through FlashAttention-4.
Actor Liam McIntyre, who voices Wolverine, and narrative director Walt Williams discussed how they shaped the character for Insomniac's upcoming Marvel's Wolverine game. They described collaborating on performance and story choices to create a fresh take on Logan distinct from prior adaptations.
GoKawiil