The Arch User Repository has now been targeted by supply-chain attackers three times in 2026. The third wave, which began with a malicious commit pushed to openconnect-sso on July 29, prompted Arch Linux to pause all AUR package adoptions on July 30 and freeze all AUR pushes on August 1. It was the same structural vulnerability each time: an attacker adopts an orphaned package through the AUR's standard process, then immediately pushes a malicious PKGBUILD.

The campaign — tracked by researchers as "Atomic Arch" — initially compromised over 400 packages in its first wave (June 2026), later expanding to more than 1,500 affected packages across all three incidents. The malicious PKGBUILDs silently fetched two rogue npm packages — atomic-lockfile and js-digest — during the standard build process, deploying a Rust-based infostealer and, in some cases, an eBPF rootkit targeting developer environments and CI/CD infrastructure.

What Changed on August 11

After eleven days of frozen pushes, Arch Linux restored AUR functionality on August 11 with new structural controls. New users are now subject to a review gate before they can adopt orphaned packages — closing the specific path exploited in all three waves. The team reverted malicious commits, permanently banned the associated accounts, and published a checklist of affected package names so users could audit their systems.

Arch's official repositories — [core], [extra], and [multilib] — remained unaffected throughout all three incidents. The AUR has always operated under a different trust model: packages there are community-maintained, and Arch has long advised users to read PKGBUILDs before building. The attacks exploited the gap between that advice and actual user behavior at scale.

The Harder Question the AUR Now Has to Answer

The review gates are a structural improvement, but they don't resolve the underlying tension in the AUR's design. The repository's strength — that anyone can contribute, and packages can be adopted without friction — is also its weakness. Three attack waves in a single year involving the same adoption vector is a pattern, not an anomaly.

The Arch team has acknowledged they're still working toward a more permanent fix for orphaned package transfers. In the meantime, if you run Arch and install AUR packages, the checklist of affected names is the first place to look. And if you build AUR packages in a CI environment, the window between June and August 2026 warrants a careful review of build logs.

The AUR isn't going away, and neither are the people who want to compromise it. What August 11 bought is time — and, hopefully, the momentum to use it.

Migrating a live container has always been painful. The traditional approach — stop the container, copy its filesystem, restart it on the destination — means downtime measured in minutes, not seconds, especially when the container's data set is large. For production environments, this friction is a real operational cost.

Incus 7.4, released August 30, 2026, changes that for users on ZFS or Btrfs storage. The new near-live migration approach keeps the container running while Incus repeatedly transfers incremental filesystem snapshots to the destination. When the delta between snapshots is small enough, Incus performs a final sync — the container is paused only for that last transfer — then immediately restarts on the new host. The window of actual downtime shrinks from minutes to seconds.

How to Use It

The feature is surfaced through the --refresh flag on incus move. The same approach applies both within a single Incus cluster and between separate remote servers. A companion option, refresh-migrate, was also added to cluster.evacuate, meaning you can use the same near-live method when draining a cluster node for maintenance.

One important limitation: the feature only works with ZFS and Btrfs. Both filesystems support efficient snapshot-and-diff operations that make incremental transfer practical. Other storage backends lack the same mechanism, so the conventional migration path remains unchanged for those setups.

Security Fixes Bundled in This Release

Incus 7.4 also patches two medium-severity CVEs. CVE-2026-81500 addressed a client-side path traversal issue, and CVE-2026-81501 fixed a problem with private image imports. Users running earlier versions should update regardless of whether they intend to use the new migration feature.

The release also extends UEFI Secure Boot tooling with a new incus low-level secureboot command set for managing database entries, and adds burst limit controls for VM disk I/O — useful for environments where multiple VMs share storage and one workload can saturate the available throughput.

What Incus Is

For context: Incus is a community fork of LXD that emerged after Canonical moved LXD under exclusive control in 2023. The project is maintained under the Linux Containers umbrella by Stéphane Graber and has moved at a steady release cadence, adding significant features since the fork. The 7.0 LTS branch released earlier in 2026 cemented its position as the de facto open-source choice for system container and VM management on Linux.

Near-live migration is the kind of feature that doesn't make headlines for most users — until the day you need it. Incus 7.4 makes that day significantly less painful.

When Rebel Wolves announced that The Blood of Dawnwalker would launch September 3, 2026, expectations were high. The studio is staffed by veterans of CD Projekt RED, including former Witcher 3 game director Konrad Tomaszkiewicz, and their debut pitch was ambitious: a single-player, open-world action RPG set in 14th-century Europe where you play as a man cursed to be human by day and vampire by night — with only 30 in-game days to save your family from a powerful vampire court.

The reviews are in, and they largely deliver on that promise. On Metacritic, the game sits at 83 based on 53 critic reviews, with 91% of those reviews in the positive column. Destructoid went further, calling it the best game of 2026 — no small claim in a year that has already seen several major releases.

The Day/Night Split Is the Game's Real Engine

The core mechanic that critics keep returning to is the dual-form system. During the day, Coen fights exclusively as a human — sword-based combat, stamina management, and the vulnerability that comes with mortality. At night, he shifts into a vrakhir, gaining wall-running, a claw-based attack set, and significantly more power against other undead. The result is effectively two different gameplay modes layered over the same open world.

The 30-day countdown adds a layer of tension that most open-world RPGs deliberately avoid. You can't do everything. Side quests, alliances, and exploration all cost time, and time spent means days ticked off your counter. Critics have noted this forces a kind of intentionality that's increasingly rare in the genre — you can't simply explore every corner of the map without cost, because every choice has a real in-game consequence.

Where It Stumbles

Not every review is glowing. The combat system — built around a directional parry mechanic that rewards blocking at the right moment to break an enemy's guard — feels repetitive according to Game Informer, with enemy encounters reusing movesets beyond the main boss fights. Several reviewers flagged that the various systems, while each individually strong, don't fully integrate into a coherent whole.

It's a debut-studio issue that's hard to escape: the ambition is clearly present, but so is the roughness that comes from building something this complex for the first time.

Built on Unreal Engine 5, Published by Bandai Namco

The game runs on Unreal Engine 5 and is published by Bandai Namco Entertainment for PS5, Xbox Series X|S, and PC. The technical execution has been praised in several reviews, with the medieval European world and its day/night transformations receiving particular attention for visual fidelity.

For a first game from a new studio, an 83 Metacritic with near-universal praise for its ambition is a strong start. Rebel Wolves described this as the first chapter of a new RPG saga. If they can tighten the systems in a sequel, the foundation they've built here is something to watch.

AMD's ambitions in the AI infrastructure market took a concrete step forward on August 21st with the official launch of Helios — a rack-scale AI system designed to take on NVIDIA's NVL-series racks at every level: compute density, memory bandwidth, and scale-out networking. The numbers are staggering, and the customer list already includes Microsoft Azure and OpenAI.

What's Inside a Helios Rack

A single Helios rack packs 72 liquid-cooled AMD Instinct MI455X GPUs, built on the CDNA 5 architecture and equipped with HBM4 memory. These are joined by 6th Gen AMD EPYC 9006 (Venice) CPUs, AMD Pensando networking silicon, and the full ROCm software stack — all from a single vendor, in a single rack, under a single support contract. The system fits into 18 compute trays and 6 switch trays in a 4OU ORW rack.

The specs that matter most for AI workloads:

• 2.9 EFLOPS of AI compute per rack (FP8)
• 31 TB of total HBM4 memory
• 1.7 PB/s of total HBM4 bandwidth
• 260 TB/s of intra-rack GPU-to-GPU bandwidth
• 43 TB/s of server-to-server bandwidth via Pensando networking

NVIDIA Is the Clear Target

AMD has been building toward this moment for years. The MI300X was competitive in certain inference workloads; Helios is AMD's answer to NVIDIA's GB200 NVL72 at the rack scale. The architecture mirrors the thermal approach NVIDIA has adopted — liquid cooling, extreme density — while betting that a single-vendor integrated stack simplifies procurement and support in ways a best-of-breed ecosystem approach cannot.

Who's Actually Buying It

AMD announced that Helios deployments are beginning in H2 2026. Microsoft Azure and OpenAI are both on record with Helios commitments — which is the most meaningful signal possible. OpenAI running workloads on AMD hardware signals that ROCm has reached a point where production AI training and inference on AMD is viable at scale. If that holds up in practice, the competitive dynamics of the AI accelerator market shift meaningfully.

The ROCm Question

Hardware specs only matter if the software works. AMD's ROCm ecosystem has historically trailed CUDA in tooling maturity, framework compatibility, and debugging experience. The Helios launch coincides with what AMD describes as a substantially improved ROCm release — but the proof will be in sustained production deployments. If Azure and OpenAI report smooth operations over the next two quarters, AMD's AI narrative changes substantially. If they hit friction, the specs become a footnote.

Helios is available now in production. More details on the system architecture can be found at AMD's official blog.

On August 27th, over 100 technology companies — including OpenAI, Anthropic, Google, and Microsoft — published an open letter warning that AI-enabled cyberattacks are poised to become dramatically more widespread and dangerous in the coming months. The letter calls for coordinated action across the private and public sectors before the window to prepare closes.

What the Letter Says

"In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable," the letter states. "We have a limited window to strengthen cyber defenses."

The signatories are unusually broad. Cybersecurity firms like CrowdStrike, Okta, and Fortinet are present, but so are Mastercard, Visa, General Motors, and Shopify. The breadth signals this isn't positioning from the security industry — the concern has spread to virtually every sector that depends on digital infrastructure.

The Threat: AI as a Force Multiplier for Attackers

The letter's core argument is that AI lowers the barrier to sophisticated cyberattacks. Tasks that previously required deep technical expertise — crafting convincing phishing campaigns, finding exploitable vulnerabilities in code, automating lateral movement inside networks — can increasingly be partially or fully automated by capable AI models. As models improve, so does the attack surface.

The sectors identified as most at risk include hospitals, water treatment plants, and the infrastructure that powers the internet itself. These are systems where a successful attack can translate directly into physical harm — not just data breaches or financial loss.

What the Companies Are Asking For

The letter makes a specific and notable request: that frontier AI companies give vetted defenders early access to advanced models before public release. The idea is that security researchers and critical infrastructure operators need to understand a model's attack-assist capabilities — and develop countermeasures — before adversaries can exploit them at scale.

The letter also calls for broader collective action across industry and government, though it stops short of prescribing specific regulatory frameworks. It's framed as an alarm, not a policy proposal, and the signatories are careful to position defense — not restriction — as the goal.

Reading Between the Lines

This letter arrives at an interesting moment. AI labs have faced growing pressure over dual-use risks, and an industry-led initiative like this can be read as a preemptive move to shape the regulatory conversation. Whether it leads to concrete changes — or remains a statement of concern — depends on whether the "collective action" it calls for actually materializes in practice.

Either way, the convergence of OpenAI, Anthropic, Google, and Microsoft on a single document about AI danger is worth noting. These are companies that compete intensely for model supremacy. The fact that they found common ground here suggests the threat assessment is genuine.

Linus Torvalds opened the Linux 7.3 merge window on August 30th with the release of 7.3-rc1 — and it's already one of the largest opening RCs in recent memory, clocking in at just under 41 million lines of code. A significant chunk of that is courtesy of a massive AMD GPU register file dump, but the real story is the breadth of improvements landing across the entire stack.

AMD Gets Big Wins

AMD dominates the diff stats for 7.3-rc1. The AMD kernel graphics driver ballooned to roughly 6.5 million lines, largely due to new DCN 6.0 display engine register headers and other upcoming GPU IP blocks that AMD needs in-tree ahead of its next hardware generation. Phoronix's breakdown confirms that AMD's changes account for around a third of all rc1 delta — which is extraordinary even by the standards of AMD's historically large graphics submissions. Additional Zen 6-era platform enabling is also in the mix for users who want early support for upcoming hardware.

The Steam Controller Finally Gets a Native Driver

Buried in the gaming section is a long-overdue win: Linux 7.3 includes a native HID driver for the original Valve Steam Controller. Up until now, using the Steam Controller on Linux without the Steam client required either SDL2 or third-party workarounds. The new in-kernel driver changes that — the controller will work as a standard HID input device, regardless of whether Steam is running. A small but meaningful quality-of-life win for Linux gamers.

Apple Silicon Gets Thunderbolt and Better Power Management

Apple hardware support in Linux continues its steady progress. 7.3-rc1 introduces initial USB4 and Thunderbolt support for Apple M1 through M3 chips, alongside a new Apple PMGR (Power Manager) driver that handles power-gating of on-chip hardware blocks. This opens the door for more peripherals to work reliably on Apple silicon machines running Linux, and lays groundwork for future Thunderbolt device compatibility.

KVM, Scheduler, and Other Highlights

A major overhaul of KVM guest memory conversion is part of rc1, improving how the kernel manages memory transitions for virtual machines. On the scheduler side, improvements targeting gaming workloads, hybrid CPU topologies, and lower scheduler latency have been landing throughout the 7.3 cycle.

• Intel Xe3P (Nova Lake) graphics — enabled out-of-the-box
• FailFS upstreamed — a minimal filesystem that always returns errors, useful for testing error-handling paths
• Low-latency mode for USB4STREAM — benefits real-time audio and data over USB4
• Nouveau + NVK Vulkan Video — open-source NVIDIA driver gains Vulkan Video decode support
• Initial Intel Starfire support begins landing

The final release of Linux 7.3 is expected in the second half of October 2026. If you want to test now, rc1 is available from kernel.org.

At 7:26 a.m. EDT this morning, NASA's Nancy Grace Roman Space Telescope received its final launch clearance and lifted off on a SpaceX Falcon Heavy from Launch Complex 39A at Kennedy Space Center. The launch window had a 60% favorable weather probability — good enough to proceed. Roman is now en route to its operational orbit at the Earth-Sun L2 Lagrange point, roughly 1.5 million kilometers from Earth, the same neighborhood as the James Webb Space Telescope.

Named after Nancy Grace Roman, NASA's first chief of astronomy, the telescope spent years in development and carries capabilities that represent a genuine step-change in how much sky we can observe at Hubble-level resolution.

The numbers that matter

Roman's wide-field imager has a field of view 100 times larger than Hubble's. At the same resolution. That's not an incremental improvement — it changes the economics of observation entirely. Roman will map the sky a thousand times faster than Hubble at equivalent detail, producing roughly a thousand years' worth of Hubble imaging data every ten months of operation.

In practical terms: surveys that would have taken decades to complete become manageable multi-year projects. The telescope is designed to observe in near-infrared light, which penetrates dust clouds that block optical wavelengths and stretches with the expansion of the universe, making it ideal for studying objects at cosmological distances.

What it's going after

Roman has three primary science programs. The first is dark energy — specifically, using the distribution of billions of galaxies across time and the behavior of Type Ia supernovae as standard candles to constrain the equation of state of dark energy and understand why the universe's expansion is accelerating. This is one of the most significant open problems in physics.

The second is exoplanets. Roman will use gravitational microlensing to detect planets that are extremely difficult to find with other methods: cold planets far from their stars, free-floating rogue planets, and Earth-mass bodies. The goal is a statistical census — not just finding individual systems, but understanding what kinds of planetary architectures are common across the galaxy.

The third is a general infrared sky survey of extraordinary depth and breadth. This data will be publicly available and will underpin follow-on science for decades, the same way Hubble's archival data continues to produce new discoveries long after any particular observation campaign ends.

The coronagraph experiment

Roman also carries a technology demonstrator called the Coronagraph Instrument — a device designed to block starlight so precisely that planets orbiting nearby stars can be directly imaged in reflected light. If it works as designed, it will achieve contrasts roughly 1,000 times better than any coronagraph previously flown in space. This is a pathfinder for future missions designed to photograph Earth-like planets around Sun-like stars.

Why today matters

The launch of a telescope of this scale is genuinely rare. Webb launched in 2021 and is still returning science that reshapes entire fields. Roman is designed to complement Webb, covering wide areas that Webb cannot efficiently survey while Webb follows up on the most interesting targets at higher resolution. Together, they form the core of NASA's next decade of space-based astrophysics.

The first science observations are expected to begin several months after launch, following mirror alignment, instrument calibration, and systems checkout. Data from the Wide Field Survey will begin flowing publicly through the Barbara A. Mikulski Archive for Space Telescopes at STScI. If the commissioning goes cleanly, Roman will be doing science before the end of 2026.

The International Mathematical Olympiad is a six-problem, nine-hour proof marathon that humbles professional mathematicians. The gold-medal threshold has historically sat around 29 out of 42 points. This year, Claude Opus 5 scored 42 out of 42 — without writing a single line of Python, calling an external solver, or using any agent scaffolding.

The result was announced by researchers tracking AI benchmark performance shortly after the IMO 2026 results were finalized. Anthropic generated four independent solutions to each of the six problems, had all 24 checked by an automated judge, and then had a human grader score a selected solution for each problem. All six received full marks.

How the run worked

Anthropic prompted the model to produce a rigorous, self-contained proof for each problem with a 256,000-token output limit and adaptive thinking set to maximum. One solution ran out of output space and was resampled at lower thinking effort; that was the only intervention. No code execution, no symbolic math libraries, no tool calls — just long-form mathematical reasoning in natural language.

The IMO 2026 leaderboard at BenchLM.ai shows Opus 5 at the top with 42/42. Multiple competing frontier models also attempted the problems: Claude Fable 5, Google's Sol, and others all achieved perfect scores as well, suggesting the 2026 IMO problems have effectively been solved as a benchmark category. The bar has moved.

What this actually means

The first honest answer is: this benchmark is now saturated. IMO problems, once considered a credible ceiling test for AI mathematical reasoning, no longer distinguish between the leading models. The community will need harder targets — ongoing research competitions, unsolved conjectures, or novel proof-generation tasks.

The second answer is more practical. Opus 5 was released on July 24, 2026, positioned as Anthropic's everyday enterprise model — a step below the company's flagship Fable 5 in cost ($5 per million input tokens vs. Fable's higher tier) but designed for extended agentic work, with a 1M-token context window and up to 128K output tokens. That a model in this price bracket can now solve every IMO problem should change how developers think about deploying mathematical reasoning tasks. Problems that previously required the most expensive tier may not anymore.

The third answer is the one that sits with you: solving all six IMO problems on the first pass, with multiple independent correct proofs per problem, places current AI performance "roughly five standard deviations above average human capability" on these tasks, according to MindStudio's benchmark analysis. Mathematicians who spend careers working on olympiad-style problems would score somewhere around the gold threshold. The model cleared it by 13 points.

What comes next

The IMO benchmark joins a growing list of once-formidable tests that AI systems have fully saturated: coding competitions, bar exams, professional licensing tests. Each saturation point reshapes the conversation about what to measure next. For mathematical reasoning, the frontier is now graduate-level open problems and real-time research collaboration — domains where correctness is harder to verify and the search space is genuinely open.

For now, the headline stands: a model you can access today via API solved every problem the world's best high-school mathematicians attempted this year. Without a calculator.

Serious Sam: Shatterverse, the multiverse-hopping co-op spin-off developed by Behaviour Interactive and published by Devolver Digital, launches on August 31, 2026 for PlayStation 5, Xbox Series X/S, and PC via Steam and the Microsoft Store. If you've been looking for a reason to get five people in a voice chat and shoot aliens at absurd velocity, your window is now.

The setup

Über Mental — the series' perpetual big bad — has done something new this time: fractured the walls between dimensions. The result is the Shatterverse, and the solution is a team of Sams pulled from alternate timelines who have to work together to repair it. It's a deliberately maximalist premise that gives Behaviour Interactive room to do whatever they want with the visual identity of each run.

The tone is consistent with what Dead by Daylight's studio has shown it does well: tight cooperative loop design and systematic escalation. Shatterverse is a 1-to-5-player co-op roguelite FPS, and every structural choice is designed to make repeated runs feel different.

How the runs work

Each run is a set of wave-based fights across procedurally generated map layouts. Between waves, players choose upgrades for their arsenal — boons that stack and interact with each other, plus optional run modifiers that can bend core rules of the game. The classic Serious Sam weapons are all here (the double-barreled shotgun, the minigun, the rocket launcher in all its variants), joined by new armaments specific to the Shatterverse setting.

Death ends a run, but progression continues: unlocks and meta-progression carry over, and each attempt can be approached with a different build philosophy. Behaviour Interactive has described the design goal as making sure no two runs feel identical — a reasonable promise given the combination of procedural maps, boon interactions, and modifier stacking.

The studio fit

Behaviour Interactive is best known for asymmetric horror in Dead by Daylight, which makes Shatterverse a noticeable genre pivot. But the underlying competency — designing systems that generate replayability through player choice rather than purely random generation — transfers well. Roguelite shooters live and die on the quality of their upgrade interactions, and Behaviour's track record of managing complex player-versus-environment loops is encouraging on that axis.

The Devolver Digital connection also matters. Their Steam page has been running a public playtest ahead of launch, and the feedback loop from those sessions appears to have influenced pacing and onboarding. Early playtest impressions ranged from enthusiastic to cautiously optimistic, with most friction points centering on the learning curve for new players unfamiliar with the series.

What to expect at launch

The game ships tomorrow. Devolver titles typically avoid aggressive day-one monetization — the expectation is a complete experience at the base price, with potential DLC runs and additional content added post-launch if the game finds its audience. The co-op focus means the experience scales meaningfully with the number of friends you bring: solo runs are viable but demonstrably harder than the full five-player configuration.

If the roguelite FPS category interests you at all, Serious Sam: Shatterverse is worth watching on launch day. The source material has always been loud, fast, and unconcerned with subtlety. That's exactly the right DNA for this genre.

Pimax has a reputation for big promises and difficult shipping timelines. That made their July commitment to clear the entire Dream Air preorder backlog by August all the more noteworthy — and this month, they followed through. Units are now in backers' hands, and the reviews are in: the Dream Air is redefining what desktop PC VR looks like in 2026.

The Hardware

The Dream Air's headline specification is its display: 3,840 × 3,552 pixels per eye on Sony micro-OLED panels, running at 90Hz. That's an enormous leap over the pixel density most headsets ship with, and it largely eliminates the screen-door effect that has haunted PC VR since its early days. Combine that with a 110° field of view through pancake optics and you get an image that reviewers have consistently described as closest-to-real-life among current shipping headsets.

The hardware package also includes eye tracking, automatic IPD adjustment, spatial audio, and DisplayLink connectivity — and the whole thing weighs under 180 grams. That last number is significant: weight and balance have been the persistent comfort complaint for high-end PC VR headsets, and the Dream Air addresses it directly.

What Reviewers Found

SlashGear's review, published August 24, praised the visual fidelity as best-in-class for a PC VR headset and highlighted the comfort improvement over previous Pimax products. SimRacingSetups gave it an 8.2/10 overall with a 9.2/10 for gameplay — a vertical where high pixel density directly translates to cockpit clarity and instrument legibility.

The recurring theme across reviews is that the micro-OLED display is genuinely transformative compared to LCD or standard OLED panels. Colors are more saturated, blacks are deep, and the combination of resolution and panel quality makes the headset feel like a generational step rather than an incremental spec bump.

The Dream Air SE

Alongside the flagship Dream Air, the Dream Air SE is also shipping in weekly batches, offering a slightly more accessible entry point into the same micro-OLED platform. Both units share the core Sony display technology and the under-180g target weight.

Context

The Dream Air arrives at a moment when the premium PC VR space is unusually active. Valve's Steam Frame is cleared for launch but hasn't shipped yet, and the Meta Quest line owns the standalone segment. Pimax's decision to target the tethered, high-fidelity end of the market — and to actually deliver on its shipping promise — gives serious sim racers, VR content creators, and enthusiasts a genuinely compelling new option.

If you've been tracking the Dream Air since it was announced and wondering whether Pimax would actually ship it on time, the answer is yes. And if you haven't been tracking it, the review scores suggest it's worth a closer look.

The Astro team just shipped version 7.0 of their static site framework, and the headline feature isn't a new API or a richer component model — it's a quiet swap under the hood that makes building content-heavy sites dramatically faster. Meet Sätteri, the Rust-powered Markdown and MDX processor that ships as Astro 7's new default pipeline.

What Sätteri Is

Sätteri is a Markdown processing engine built in Rust by Astro core team member Erika. It sits on top of two well-regarded Rust libraries: pulldown-cmark for Markdown parsing and Oxc for JavaScript/MDX handling. The key shift from the previous setup is that features like GitHub Flavored Markdown, smart punctuation, heading IDs, math support, and wikilinks are now built in rather than being plugins stacked on top of the unified/remark/rehype JavaScript pipeline.

That's a significant structural change. Where the old approach assembled a chain of JavaScript plugins at build time, Sätteri handles those steps natively in compiled Rust code, eliminating the overhead of passing content through multiple JS transform layers.

The Performance Numbers

In Astro's own benchmarks, switching a site's Markdown pipeline from the old unified-based system to Sätteri shaved over a minute off build times for both the Astro documentation site and the Cloudflare docs site. Combined with Astro 7's second major upgrade — Vite 8, which ships with the new Rolldown bundler replacing esbuild — the overall build speed improvement lands between 15% and 61% depending on the project's content volume and structure.

For a framework whose users are frequently content-heavy sites running hundreds or thousands of Markdown files, those numbers matter in practice.

What Stays Compatible

The transition isn't a hard break from the unified ecosystem. Sätteri still supports flexible JavaScript plugins, which means existing remark and rehype plugins can be adapted or wrapped rather than discarded outright. The Astro team designed the processor with a pluggable architecture precisely to preserve compatibility for teams that have invested in custom transformations.

The Bigger Picture

Astro 7 continues a pattern we've seen across the JavaScript tooling ecosystem over the past couple of years: performance-critical internals moving to Rust. Biome replaced ESLint and Prettier for many teams. Oxc is eating away at Babel's role as a parser. Sätteri follows the same logic applied specifically to the Markdown processing bottleneck that content sites hit at scale.

For most developers, the upgrade path to Astro 7 should be straightforward. The defaults are faster, the common features are still there, and the places where behavior changes are documented. If you run a site with hundreds of content files and have been watching your CI build times creep upward, this is the release worth upgrading for.

After more than a year of anticipation — and one delay that pushed the release past its original 2025 target — FromSoftware and Bandai Namco launched Elden Ring: Tarnished Edition on the Nintendo Switch 2 on August 28, 2026. The verdict from critics has been immediate and emphatic: this is one of the best console ports in recent memory.

What's in the Box

Tarnished Edition is an all-in-one package that bundles the complete base game with the Shadow of the Erdtree expansion, along with bonus armor sets and a steed attire exclusive to the Switch 2 version. At $79.99 on the Nintendo eShop (41 GB download), it's priced at a premium — but it also ships with the full Elden Ring experience out of the gate, no separate DLC purchase required.

Performance: Better Than Expected

The big question going into launch was whether the Switch 2's hardware could handle a game of this scope. The answer turned out to be yes — comfortably. The port runs at 1080p in both docked and handheld mode, targeting 30fps with consistent results. The extra year of optimization work that caused the delay from 2025 was evidently well spent.

FromSoftware has always been precise about performance targets over raw frame rates for its titles, and that philosophy works in the Switch 2's favor here. The 30fps cap is stable, and the visual fidelity — textures, lighting, draw distance — holds up impressively on portable hardware.

What Critics Are Saying

The reception has been near-universal praise. Metacritic has the port sitting at 91 based on 40 critic reviews, with OpenCritic's tally at 88 and a 100% recommended rate. Reviewers have consistently highlighted the stability of the port and how well the open-world traversal translates to the handheld form factor.

The ability to drop into the Lands Between — or the Realm of Shadow — during a commute or from the couch without docking a dedicated console is exactly the kind of proposition that made the original Switch so compelling for game-of-the-generation titles.

Why This Matters

Elden Ring on Switch 2 is more than a checkbox port. It demonstrates that the Switch 2 can run demanding open-world games at a quality level that doesn't require compromise. For players who have been holding out on the title, or who want a portable version of a game they already finished on another platform, Nintendo's official launch page confirms this is the definitive portable edition — base game, expansion, and extras, all in one.

If you've been waiting for the right moment to step into the Elden Ring, portable 1080p with a 91 Metacritic might just be it.