GoKawiil
Tech News
Android Authority found Spotify developing a 'Cache Limit' feature for its Android app, letting users set a maximum size for the app's temporary storage cache. The tool would warn that smaller limits increase mobile data use and could affect playback, while suggesting a default size for balance. Downloaded music and podcasts stored offline would remain unaffected by this cache limit.
Testing by Bilibili channel Homolab found that the 1TB iPhone 18 Pro Max, which uses QLC NAND, suffers major performance drops compared to the 512GB TLC-based iPhone 18 Pro. Once its SLC and secondary TLC caches are exhausted, sustained writes fall to about 79.4 MB/s and can dip as low as 25.6 MB/s, worsening to just 1.1 MB/s once the drive is 60% full.
DeepSeek released V4.1 Flash, a model initially mistaken for a minor update but revealed via its technical report to be a significant architectural overhaul, effectively a V5-class release. It achieves near 420 tokens/second throughput while compressing KV cache by 4x through techniques including cross-layer compression, sparse attention indexing optimizations, and FP4 precision, alongside a YOCO-inspired prefill design that only activates 8B parameters during prefill versus 16B during decode across its 40 layers.
CloudX built and released cloudx-io/setup-go, a drop-in replacement for GitHub's official actions/setup-go, after finding that the official action caused parallel jobs to interfere with each other and repeatedly pull stale cache data. In backtests on CloudX's monorepo running lint, test, and build jobs in parallel, the new action eliminated an estimated 86% of unnecessary work performed by the default setup, cutting test job runtimes by 69%.
Microsoft Outlook users encountering the 'Your message can't be displayed right now' error can resolve it by clearing their cache, whether they use the web version or the desktop client. The fix forces Outlook to rebuild fresh copies of emails without deleting any messages, calendar events, or contacts.
A new textbook titled Foundation Model Engineering has been released, aimed at AI engineers and research-minded readers who want to understand foundation models beyond basic API use. It ties together topics like attention, mixture-of-experts, RLHF, multimodality, long-context inference, retrieval-augmented generation, and agents into a single engineering narrative, using PyTorch examples, quizzes, and interactive visualizers.
Newegg is selling AMD's Ryzen 7 9850X3D, currently the top gaming CPU on the market, for $484, roughly $15 under its original launch price. The deal throws in a 240mm Cooler Master AIO liquid cooler worth $79.99 and a copy of Onimusha: Way of the Sword worth $69.99 at no extra cost. The chip, an upgraded version of the 9800X3D announced at CES 2026, keeps the same 8-core/16-thread design and 3D V-Cache but adds a higher boost clock for a modest gaming performance gain.
Newegg is offering a combo deal pairing AMD's Ryzen 7 9800X3D processor with an Asus TUF Gaming X870E-Plus motherboard and 32GB of Corsair Vengeance RGB DDR5-6000 memory for $1,109.99. The package also throws in a 240mm Cooler Master AIO liquid cooler and a copy of Onimusha: Way of the Sword at no extra cost, saving buyers roughly $150 versus buying the parts separately.
A developer walks through optimizing a basic spin-lock implementation in C++, starting from a naive atomic exchange loop that slows dramatically under contention. Benchmarks show the naive version taking 3.14 ns uncontended but ballooning to 246 ns with four threads, driven by cache-line contention and branch mispredictions. Through iterative refinements, the author achieves a version that is 5.7 times faster and consumes 5.4 times less energy than the original.
OpenJDK's JEP 544 proposes extending HotSpot's existing AOT cache mechanism to store precompiled native code generated during a training run, making it instantly usable when an application starts in production. If workload behavior shifts, the JVM falls back to dynamic JIT compilation to maintain performance, blending AOT and JIT approaches without requiring any code or configuration changes from developers.
A researcher replayed 68,266 requests from 393 Claude Code sessions and 23,608 Mooncake requests through a custom prefix-cache simulator, testing three approaches meant to outperform the standard LRU eviction policy used by vLLM, SGLang, LMCache and Mooncake Store. None of the alternatives beat the default LRU-over-radix-leaves baseline, and the analysis found that cache misses mostly come from rapid, seconds-apart tool-calling loops rather than sessions timing out after going idle.
DeepSeek confirmed it will officially release its V4.1 Flash model on September 10, 2026 (Beijing Time), stating internal testing shows it outperforms V4 Pro on performance, cost, speed, and task completion time. Until V4.1 Pro arrives, all Pro-model requests will be automatically redirected to V4.1 Flash and charged at the cheaper Flash pricing tier.