Memory Savings, Physical Attacks, and Gaps in Platform Access

Cloudflare’s 100TB fleet memory saving and a roughly 50% Ryzen speedup over two years put performance gains at two scales in focus, alongside a hardware security bypass and changes affecting platform developers.

Performance gains at different scales

Cloudflare describes saving another 100TB of RAM by applying math to memory use. The result comes from algorithmic and data-structure improvements across its fleet, where small savings per node accumulate into a large infrastructure reduction. For memory-heavy services, the example makes a concrete case for examining how data is represented: the aggregate benefit can be substantial even when each machine contributes only a small saving.

Daniel Lemire’s examination of how AMD Ryzen became roughly 50% faster in two years approaches performance from the processor side, looking at microarchitectural improvements and which gains appear in real code. That makes hardware generation an important part of benchmark context. Cloudflare’s work concerns fleet memory consumption, while Lemire’s concerns execution speed on individual machines; each points to a different constraint that a single headline performance number can obscure.

A physical route past debug locks

Ledger’s Donjon team used photon-emission analysis to guide laser fault injection and enable secure debug on the RP2350. The demonstrated bypass matters for products that store secrets on this microcontroller: a debug lock alone does not establish protection against an attacker with physical access. The specific result concerns the Raspberry Pi RP2350, but it gives teams using that chip a concrete attack to consider when assessing what their hardware protections actually cover.

Platform changes depend on source access

GrapheneOS says Android 17 adds new APIs without releasing them to AOSP, the first such release since Android 3.x. For alternative ROM developers and teams tracking platform changes through open source, that leaves the new API surface without the corresponding AOSP release. The concern is access to implementation changes, with consequences for developers whose work depends on following Android through its public source.

Apple has separately published release notes for the Xcode 27.1 beta, providing a review point for teams considering an Apple toolchain update. Its beta status matters when planning a CI upgrade: compiler, SDK and known-issue changes are the relevant areas to examine before adoption. This is a toolchain evaluation step, while the Android development concerns whether newly added platform APIs are available through the open-source release at all.