Blogroll

SAVE $170: As of Sept. 11, shop Staub's pumpkin-shaped dutch ovens for 46% off. Get the classic Staub Pumpkin Dutch Oven in burnt orange for $199.99 and save $170 for 46% off.

Opens in a new window Credit: Staub Staub 3.5-quart Pumpkin Dutch Oven $199.99 at Zwilling
$369.99 Save $170 Get Deal

Is that a chill I sense in the air? Fall weather is creeping in, and if you're ready to haul out your sweater collection, then you're also probably thinking about fall cooking. Maybe you're dreaming of a hearty beef stew or rich pot of French onion soup? Whatever you've got up your sleeve, you'll need the right vessel, and what's more festive than a pumpkin-shaped Staub? What's even better? A pumpkin-shaped Staub that's on sale.

As of Sept. 11, Staub has marked down all of their pumpkin cast iron cocottes. Both the .75-quart mini ovens and the 3.5-quart full-sized dutch ovens are on sale, and they're up to $170 off. The best deal is on the 3.5-quart Staub, which is normally priced at $369.99, but currently marked down to $199.99. That gives you 46% off for $170 in savings. Savings on the pumpkin dutch oven vary by color but the classic burnt orange is down to $199.99.

If you're not looking for the best deal but just a taste of fall, the petite pumpkin dutch oven is 16% off, bringing it down to $49.99.

While the outside of the Staub pumpkin dutch oven is decorative, the inside is covered in enameled cast iron and the lid comes in handy for long simmering stews and braises. It's dishwasher, microwave, broiler, and oven safe, too. The oven has the pumpkin shape on the inside, perfect for creating a pumpkin-shaped sourdough. Grab the Staub pumpkin dutch ovens for $199.99 and save 46% off the festive cookware.

Navigating an email inbox isn’t fun, and Google is constantly trying to improve how that works in Gmail. The latest updates will make it easier to find and track all your purchases, as well as find good deals among the piles of clutter.

In the landscape of prestige television, few shows are called "the greatest ever made" as frequently as The Wire, but this is not just for its brilliant writing—it's a realistic portrayal of a modern American city. Here's why you shouldn't miss this show.

This year we’ve seen remarkable advances in agentic AI, including systems that conduct deep research, operate computers, complete substantial software engineering tasks, and tackle a range of other complex, multi-step goals. In each case, the industry relied on careful vertical integration: tools and agents were co-designed, co-trained, and tested together for peak performance. For example, OpenAI’s recent models presume the availability of web search and document retrieval tools (opens in new tab). Likewise, the prompts and actions of Magentic-One are set up to make hand-offs easy—for example, allowing the WebSurfer agent to pass downloaded files to the Coder agent.  But as agents proliferate, we anticipate strategies relying heavily on vertical integration will not age well. Agents from different developers or companies will increasingly encounter each other and must work together to complete tasks, in what we refer to as a society of agents. These systems can vary in how coordinated they are, how aligned their goals are, and how much information they share. Can heterogenous agents and tools cooperate in this setting, or will they hinder one another and slow progress?

Early clues have emerged from an unexpected source: namely, Model Context Protocol (opens in new tab) (MCP). Since January 2025, MCP has grown from a promising spec to a thriving market of tool servers. As an example, Zapier boasts a catalog of 30,000 tools (opens in new tab) across 7,000 services. Composio provide over 100 managed MCP servers (opens in new tab), surfacing hundreds of tools. Hugging Face is now serving many Spaces apps over MCP (opens in new tab), and Shopify has enabled MCP for millions of storefronts (opens in new tab). A society of tools is already here, and it promises to extend agent capabilities through cross-provider horizontal integration. 

So, what does MCP have to say about horizontal integration? As catalogs grow, we expect some new failure modes to surface. This blog post introduces these as tool-space interference, and sketches both early observations and some pragmatic interventions to keep the society we’re building from stepping on its own feet. 

Tool-space interference describes situations where otherwise reasonable tools or agents, when co-present, reduce end-to-end effectiveness. This can look like longer action sequences, higher token cost, brittle recovery from errors, or, in some cases, task failure.

A framing example

Consider MCP as a means for extending Magentic-One, a generalist multi-agent system we released last year, to cover more software engineering tasks. Magentic-One ships with agents to write code, interact with the computer terminal, browse the web, and access local files. To help Magentic-One navigate version control, find issues to solve, and make pull requests, we could add an agent equipped with the GitHub MCP Server. However, now each time the team encounters a task involving GitHub, it must choose whether to visit github.com in the browser, execute a git command at the command line, or engage the GitHub MCP server. As the task progresses, agent understanding of state can also diverge: changing the branch in the browser won’t change the branch in the terminal, and an authorized MCP tool does not imply authorization in the browser. Thus, while any single agent might complete the task efficiently, the larger set of agents might misunderstand or interfere with one another, leading to additional rounds of debugging, or even complete task failure.

Figure 1: We can extend Magentic-One by adding an agent that equips the GitHub MCP server. However, on every turn involving a git-related task, the orchestrator will need to decide between messaging the Computer Terminal agent (with access to the git command line interface), WebSurfer agent (with access to github.com), and the agent with the GitHub MCP server. This overlap raises the possibility that they will interfere with one another.   Tool-space interference, through the lens of MCP

To better understand the potential interference patterns and the current state of the MCP ecosystem, we conducted a survey of MCP servers listed on two registries: smithery.ai (opens in new tab) and Docker MCP Hub (opens in new tab). Smithery is an MCP Server registry with over 7,000 first-party and community-contributed servers, which we sampled from the Smithery API. Likewise, Docker MCP Hub is a registry that distributes MCP servers as Docker images, and we manually collected popular entries. We then launched each server for inspection. After excluding servers that were empty or failed to launch, and deduplicating servers with identical features, 1,470 servers remained in our catalog.

To automate the inspection of running MCP servers, we developed an MCP Interviewer tool. The MCP Interviewer begins by cataloging the server’s tools, prompts, resources, resource templates, and capabilities. From this catalog we can compute descriptive statistics such as the number of tools, or the depth of the parameter schemas.  Then, given the list of available tools, the interviewer uses an LLM (in our case, OpenAI’s GPT-4.1) to construct a functional testing plan that calls each tool at least once, collecting outputs, errors, and statistics along the way. Finally, the interviewer can also grade more qualitative criteria by using an LLM to apply purpose-built rubrics to tool schemas and tool call outputs.  We are excited to release the MCP Interviewer as an open-source CLI tool (opens in new tab), so server developers can automatically evaluate their MCP servers with agent usability in mind, and users can validate new servers. 

While our survey provides informative initial results, it also faces significant limitations, the most obvious of which is authorization: many of the most popular MCP servers provide access to services that require authorization to use, hindering automated analysis. We are often still able to collect static features from these servers but are limited in the functional testing that can be done.

One-size fits all (but some more than others)

So, what does our survey of MCP servers tell us about the MCP ecosystem? We will get into the numbers in a moment, but as we contemplate the statistics, there is one overarching theme to keep in mind: MCP servers do not know which clients or models they are working with, and present one common set of tools, prompts, and resources to everyone. However, some models handle long contexts and large tool spaces better than others (with diverging hard limits), and respond quite differently to common prompting patterns. For example, OpenAI’s guide on function calling (opens in new tab) advises developers to:

“Include examples and edge cases, especially to rectify any recurring failures. (Note: Adding examples may hurt performance for reasoning models).”

So already, this places MCP at a disadvantage over vertical integrations that optimize to the operating environment. And with that, let’s dive into more numbers.

Tool count

While models generally vary in their proficiency for tool calling, the general trend has been that performance drops as the number of tools increases. For example, OpenAI limits developers to 128 tools, but recommends (opens in new tab) that developers:

“Keep the number of functions small for higher accuracy. Evaluate your performance with different numbers of functions. Aim for fewer than 20 functions at any one time, though this is just a soft suggestion.”

While we expect this to improve with each new model generation, at present, large tool spaces can lower performance by up to 85% for some models (opens in new tab). Thankfully, the majority of servers in our survey contain four or fewer tools. But there are outliers: the largest MCP server we cataloged adds 256 distinct tools, while the 10 next-largest servers add more than 100 tools each. Further down the list we find popular servers like Playwright-MCP (opens in new tab) (29 tools, at the time of this writing), and GitHub MCP (91 tools, with subsets available at alternative endpoint URLs), which might be too large for some models.

Figure 2: The number of tools listed by each catalogued server directly after initialization. Note: servers can change the tools they list at any time, but only 226 servers in our catalog declare this capability. Response length

Tools are generally called in agentic loops, where the output is then fed back into the model as input context. Models have hard limits on input context, but even within these limits, large contexts can drive costs up and performance down, so practical limits can be much lower (opens in new tab). MCP offers no guidance on how many tokens a tool call can produce, and the size of some responses can come as a surprise. In our analysis, we consider the 2,443 tool calls across 1,312 unique tools that the MCP Interviewer was able to call successfully during the active testing phase of server inspection. While a majority of tools produced 98 or fewer tokens (opens in new tab), some tools are extraordinarily heavyweight: the top tool returned an average of 557,766 tokens, which is enough to swamp the context windows of many popular models like GPT-5. Further down the list, we find that 16 tools produce more than 128,000 tokens, swamping GPT-4o and other popular models. Even when responses fit into the context window length, overly long responses can significantly degrade performance (up to 91% in one study (opens in new tab)), and limit the number of future calls that can be made. Of course, agents are free to implement their own context management strategies, but this behavior is left undefined in the MCP specification and server developers cannot count on any particular client behavior or strategy.

# of tools that would overflow context inModelContext Window1 call2 calls3-5 calls6-10 callsGPT 4.11,000,00001711GPT 5400,000171525GPT-4o, Llama 3.1,128,00016153340Qwen 332,00056378690Phi-416,0009360116109 Figure 3: Tool call response length averages, in tokens, as observed by the MCP Interviewer’s functional test plan. Only successful tool calls are considered. Horizontal lines indicate context window limits for GPT-4o and GPT-5. Tool parameter complexity

Mirroring the challenges from increasing the number of tools, increasing the complexity of a tool’s parameter space can also lead to degradation. For example, while MCP tools can take complex object types and structures as parameters, composio (opens in new tab) found that flattening the parameter space could improve tool-calling performance by 47% compared to baseline performance.  In our analysis, we find numerous examples of deeply nested structure—in one case, going 20 levels deep.

Figure 4: The maximum depth of each tool’s input properties schema. A depth of 0 indicates a tool with no properties. A depth of 1 indicates a tool with named properties but no annotations (e.g., no description or type). A depth of 2 indicates a tool with named and annotated properties.  A depth of 3+ indicates a tool with structured properties that have additional nested annotations.  Namespacing issues and naming ambiguity

Another often-cited issue with the current MCP specification is the lack of a formal namespace mechanism (opens in new tab). If two servers are registered to the same agent or application, and the servers have tool names in common, then disambiguation becomes impossible. Libraries like the OpenAI Agents SDK raise an error (opens in new tab) under this circumstance. Clients, like Claude Code, prefix tool names with unique identifiers to work around this issue. In our analysis of MCP servers, we found name collisions between 775 tools. The most common collision was “search”, which appears across 32 distinct MCP servers. The following table lists the top 10 collisions.

Tool NameNumber of Instancessearch32get_user11execute_query11list_tables10update_task9generate_image9send_message9execute_command8list_tasks8search_files8

Even when names are unique, they can be semantically similar. If these tools behave similarly, then the redundancy may not be immediately problematic, but if you are expecting to call a particular tool then the name similarities raise the potential for confusion. The following table lists some examples of semantically similar tool names relating to web search:

websearchbrave_web_searchsearch-webtavily_web_searchweb_searchgoogle_news_searchsearch_webgoogle-play-searchsearch_webkrgoogle_search_parsedgoogle_searchsearch_google_imagessearch_googleget_webset_search_exaai_web_searchsearch_google_scholarweb_search_exaduckduckgo_web_searchsearch_web_toolgoogle_search_scraperweb_search_agentanswer_query_websearchbatch-web-search  Errors and error messages

Like all software libraries, MCP will occasionally encounter error conditions. In these cases, it is important to provide sufficient information for the agent to handle the error and plan next steps. In our analysis, we found this was not always the case. While MCP provides an “IsError” flag to signal errors, we found that it was common for servers to handle errors by returning strings while leaving this flag set to false, signaling a normal exit. Out of 5,983 tool call results with no error flag, GPT-4.1 judged that 3,536 indicated errors in their content. More worrisome: the error messages were often of low quality. For instance, one tool providing web search capabilities failed with the string “error: job,” while another tool providing academic search returned “Please retry with 0 or fewer IDs.”

Resource sharing conventions

Finally, in addition to tools, MCP allows servers to share resources and resource templates with clients. In our survey, only 112 (7.6%) servers reported any resources, while 74 (5%) provided templates. One potential reason for low adoption is that the current MCP specification provides limited guidance for when resources are retrieved, or how they are incorporated into context. One clearcut situation where a client might retrieve a resource is in response to a tool returning a resource_link (opens in new tab) as a result — but only 4 tools exhibited this behavior in our survey (arguably, this would be the ideal behavior for tools that return very long, document-like responses, as outlined earlier).

Conversely, a whole different set of issues arises when there is a need to share resources from the client to the server. Consider for example a tool that provides some analysis of a local PDF file. In the case of a local MCP server utilizing STDIO transport, a local file path can be provided as an argument to the tool, but no similar conventions exist for delivering a local file to a remote MCP server. These issues are challenging enough when implementing a single server. When multiple tools or servers need to interact within the same system, the risk of interoperability errors compounds.

Recommendations

On balance, along any given dimension, the average MCP server is quite reasonable—but, as we have seen, outliers and diverging assumptions can introduce trouble. While we expect many of these challenges to improve with time, we are comfortable making small recommendations that we feel are evergreen. We organize them below by audience.

Protocol developers

We recognize the advantages of keeping MCP relatively lightweight, avoiding being overly prescriptive in an environment where AI models and use cases are rapidly changing. However, a few small recommendations are warranted. First, we believe MCP should be extended to include a specification for client-provided resources so that tools on remote servers have a mechanism for operating on specified local files or documents. This would more effectively position MCP as a clearinghouse for resources passed between steps of agentic workflows. The MCP specification would also benefit from taking a more opinionated stance on when resources are retrieved and used overall.

Likewise, we believe MCP should quickly move to provide formal namespaces to eliminate tool name collisions. If namespaces are hierarchical, then this also provides a way of organizing large catalogs of functions into thematically related tool sets. Tool sets, as an organizing principle, are already showing some promise in GitHub MCP Server’s dynamic tool discovery, (opens in new tab) and VS Code’s tool grouping (with virtual tools) (opens in new tab), where agents or users can enable and disable tools as needed.  In the future, a standardized mechanism for grouping tools would allow clients to engage in hierarchical tool-calling, where they first select a category, then select a tool, without needing to keep all possible tools in context.

Server developers

While our MCP Interviewer tool can catalog many outward-facing properties of MCP servers, developers are often in a much better position to characterize the nature of their tools. To this end, we believe developers should publish an MCP Server card alongside their servers or services, clearly outlining the runtime characteristics of the tools (e.g., the expected number of tokens generated, or expected latency of a tool call). Ideally developers should also indicate which models, agents and clients the server was tested with, how the tools were tested (e.g., provide sample tasks), list any known incompatibilities, and be mindful of limitations of various models throughout development.

Client developers

Client developers have the opportunity to experiment with various mitigations or optimizations that might help the average MCP server work better for a given system or environment. For example, clients could cache tool schemas, serving them as targets for prompt optimizations, or as an index for RAG-like tool selection approaches. To this end, Anthropic recently reported using a tool testing agent (opens in new tab) to rewrite the prompts of defective MCP servers, improving task completion time by 40%. Likewise, rather than waiting for the protocol to evolve, clients could take proactive steps to resolve name collisions— for example, generating namespaces from server names—and could reduce token outputs by summarizing or paginating long tool results.

Market developers

Finally, we see an opportunity for marketplaces to codify best-practices, spot compatibility issues at a global level, and perhaps centralize the generation and serving of model or agent-specific optimizations. Mirroring how a market like PyPI distributes Python wheels matched to a developer’s operating system or processor (opens in new tab), an MCP marketplace could serve tool schemas optimized for a developer’s chosen LLM, agent or client library. We are already seeing small steps in this direction, with registries like Smithery providing customized launch configurations to match users’ clients.

Conclusion

In summary, the MCP ecosystem offers significant value for AI agent development, despite some early growing pains. Grounded in insights from the MCP Interviewer (opens in new tab) and our survey of live servers, the evidence is clear: horizontal integration is expanding capability, yet it also exposes forms of toolspace interference that can erode end to end effectiveness. Anticipating rapid advances in model capability and growing architectural diversity, the recommendations provided here aim to ensure that protocol, server, client, and marketplace developers are well positioned to adapt and thrive. Key steps include implementing formal namespaces to eliminate collisions, enhancing protocol support for client provided resources, and encouraging transparent server documentation to foster interoperability and robust development practices across the ecosystem. 

By embracing these evergreen recommendations and proactively addressing compatibility, usability, and optimization issues, the AI agent community can create a more reliable, scalable, and efficient infrastructure that benefits both developers and end users. The future of MCP is bright, with ample opportunities for experimentation, standardization, and collective progress.

Opens in a new tab

The post Tool-space interference in the MCP era: Designing for agent compatibility at scale appeared first on Microsoft Research.

Visual Studio Code is releasing a major update this month, version 1.104. This update, also known by its code name "Eclipse," is bringing some great new features, especially for its AI-powered chat and a bunch of quality-of-life improvements.

Just as it seemed like physical media like DVDs and Blu-rays were about to disappear, the failures of the streaming industry has led to the rediscovery of media ownership. The thing is, if you didn't come up with DVDs and Blu-ray discs, you might not be aware of the restrictions that come with them.

SAVE $150: The Dreame AirStyle Pro styler and dryer is on sale at Amazon for $249.99, down from the list price of $399.99. That's a 38% discount.

Opens in a new window Credit: Dreame Dreame AirStyle Pro $249.99 at Amazon
$399.99 Save $150 Get Deal

Now that we're in the middle of September, the temperatures are less suitable for walking outside with wet hair. Grandma would not be pleased to see anyone head off to school or work in the morning with wet hair. To keep her from worrying, get that hair dry before leaving the house thanks to this great deal at Amazon.

As of Sept. 11, the Dreame AirStyle Pro styler and dryer is on sale for $249.99 at Amazon, marked down from the standard price of $399.99. That's a 38% discount that takes a sweet $150 off.

Dreame is making a name for itself by producing attention-worthy dupes, namely coming after Dyson and Shark. The Dreame Airstyle Pro looks a whole lot like the Dyson Airwrap but comes in at a much lower price, especially thanks to today's deal. Currently, the Dyson Airwrap is selling for $599.99, which means you'll save a whopping $350 by going with the Dreame AirStyle Pro.

Dreame considers the AirStyle Pro to be a seven-in-one hair styler and high-speed dryer, and it comes with tons of attachments to make this happen. There's the fast dryer attachment, the flyaway attachment, auto-wrap barrel, hard smoothing brush, soft smoothing brush, volumizing brush, and even more. To top it all off, it comes with a lux storage box.

SEE ALSO: Mashable's favorite e-reader, the Kobo Clara Colour, now comes in white

You also get three temperature settings and three wind speed settings on the Dreame AirStyle Pro. Dreame uses JetAirflow technology to help keep hair protected from overheating which is also said to reduce frizz.

Dreame also added user-friendly features to the AirStyle Pro like an anti-tangle cord that rotates 360 degrees and a filter that's easy to clean. The body of the styler itself weighs just 0.64 pounds, making it a lightweight styler.

If you tend to leave home with wet hair, make grandma happy by snagging the Dreame AirStyle Pro and drying your hair before heading out. It's on sale today at Amazon for under $250.

It’s been a busy week at Apple with the drop of the latest iPhone 17 and the new Apple Watch 11. Amid the bustle of activity, Apple is also officially dropping watchOS 26. There are plenty of changes coming down the pipeline, and tucked in among all the other new features are four brand new watch faces.

Per Apple’s release notes when installing the update, the four new watch faces include nods to classic watches, and the incorporation of Apple's new Liquid Glass design language:

Exactograph is a modern re-imagination of a traditional regular watch which separates out hours, minutes, and seconds for more precise timekeeping. 

The new Exactograph watch face. Credit: Apple

Flow uses Liquid Glass numerals that refract a fluid orb of color that responds to your wrist movements. 

The Flow watch face, which incorporates Liquid Glass elements. Credit: Apple

Waypoint acts as a live compass that shows where you are in relation to saved locations in (Apple) Maps or points of interest (Available on the Apple Watch Ultra).

The Waypoint face, exclusive to Apple Watch Ultra. Credit: Apple

Hermès Faubourg Party features animated shorts that appear based on the time of day or your movements, with a series of characters (Available on Apple Watch Hermès).

An Apple Watch Hermès-exclusive face, featuring animated characters. Credit: Apple

As you may have surmised, the Waypoint watch face is exclusive to the Apple Watch Ultra while the Hermès Faubourg Party is exclusive to the Apple Watch Hermès. In addition to the new watch faces above, the Photos watch face has received a Liquid Glass makeover, so it’ll look new as well. 

The updated Photos watch face. Credit: Apple

All Apple Watch users who upgrade to watchOS 26 will get the Flow and Exactograph faces. According to Apple’s website, that will include all Apple Watches from the Series 6 onward and all three Apple Watch Ultra generations. The Apple Watch SE 2 is also invited to the party. 

The update will begin rolling out to the appropriate devices on Monday, September 15. Beta users should already have access to it now. 

WatchOS 26 was initially announced at WWDC, so most of the new features have been announced already. Apple Watch users will be treated to a new Liquid Glass design, the new AI-powered Workout Buddy, and the inclusion of Live Translation in addition to a number of improvements of things like the aforementioned Photos watch face and the addition of a Notes app. 

SAVE UP TO 53%: As of Sept. 11, Soundcore headphones are marked down up to 53%. Find the Soundcore V20i open-ear headphones for $23.49, knocking 53% off for $26.50 in savings.

Best Soundcore headphone deals Best open-ear headphone deal Soundcore V20i open-ear headphones $23.49 (Save $26.50) Get Deal Best headphone deal Soundcore Life Q20 headphones $29.99 (Save $30) Get Deal Best earbud deal Soundcore P20i earbuds $19.98 (Save $20.01) Get Deal

Dreaming of long fall walks? Wearing a sweater in the cool crisp air and listening to the new Taylor Swift album. If you're ready to upgrade your headphones for the new season, Soundcore has some great deals for you.

Soundcore is our go-to for budget-friendly headphones. They start under $100, but still deliver high-quality sound. And when they're on sale, you can score a pair of headphones or earbuds with active noise cancellation for under $50.

Soundcore's headphones are especially good for students who don't want to spend an entire paycheck on a pair of headphones but still want to be able to block out the noise while studying in the library.

Right now there are a ton of deals to be had on Soundcore headphones, offering savings up to 53% off. Here are all of the best deals now.

Best Soundcore deal Opens in a new window Credit: Soundcore Soundcore V20i open-ear headphones $23.49 at Amazon
$49.99 Save $26.50 Get Deal Why we like it

Headphone trends come and go, but right now sleep earbuds and open earbuds are the buzziest. Sleep earbuds provide comfort while blocking out noise, but open earbuds are a safe option that allows you to be aware of your surroundings while still listening to music.

Right now, you can find Soundcore's open-ear headphones for their lowest price ever. As of Sept. 11, the Soundcore V20i open-ear headphones are down to $23.49, saving you $26.50 for 53% off. That's officially their lowest price ever, from its previous low of $24.99.

Open-ear headphones are ideal for working out. When you're on a run or on a bike, and want to listen to your pump-up playlist but still want to be aware of traffic, open-ear headphones allow for the best of both worlds. The Soundcore V20i, hook around the ears for a comfortable yet snug fit. Plus, you can take calls with the V20i, which have four mics in them.

Looking for a more traditional pair of headphones? Don't worry, there are still a ton more Soundcore deals to explore.

More Soundcore deals

• Soundcore P20i earbuds — $19.98 $39.99 (save $20.01)

• Soundcore Life Q20 headphones — $29.99 $59.99 (save $30)

• Soundcore Space A40 — $44.58 $79.99 (save $35.41)

• Soundcore Q20i earbuds — $44.99 $69.99 (save $25)

• Soundcore Life Q30 headphones — $55.98 $79.99 (save $24.01)

Microsoft is shutting down its Outlook Lite app for Android, a lightweight version of its email client that was designed for low-end phones and slower networks. The app will be retired starting on October 6, 2025, and after that date, new installations will be blocked.

This episode of How It Hits breaks down The Long Walk and its place in a long tradition of deadly competition stories, from Battle Royale to The Hunger Games. Why do we keep coming back to these narratives—and why do they so often put kids in the arena?

SAVE $34: As of Sept. 11, you can get both the Anker Soundcore Sleep A30 earbuds and Soundcore Q20i noise-cancelling headphones for just $265.98. That's $34 in total savings and $9 cheaper than buying each pair separately on sale.

Opens in a new window Credit: Anker Anker Soundcore Sleep A30 earbuds and Q20i noise-cancelling headphones $265.98 at Amazon
$299.98 Save $34 Get Deal

Over-ear headphones and compact earbuds both have their time and place. But when it costs hundreds of dollars for each, you often have to make a tough decision. If you're torn between the two, we recommend checking out this bundle deal from Anker Soundcore. For under $270, you can get your hands on both.

As of Sept. 11, you can get the Anker Soundcore Sleep A30 earbuds and Q20i noise-cancelling headphones bundled together for only $265.98. A $300 value, that saves you about $34 total. The Sleep A30 earbuds are $230 on their own, so this bundle deal essentially gets you the Q20i headphones for only $36.

We're huge fans of the Soundcore Sleep A30 earbuds. They're the only pair of sleep earbuds on the market that feature active noise cancellation and it works great. Mashable's Bethany Allard tested them for a few months and had mostly positive things to say. "They bring the best of what I would expect from a premium sleep earbud — active noise cancellation, solid battery life, and plenty of optimization for sound in comfort," she writes. One of their only pitfalls is the price tag — although, with this bundle deal, we can definitely look past that.

The Soundcore Q20i headphones are budget-friendly over-ears that go beyond the basics. They offer active noise cancellation, 40-hour battery life (with ANC on), transparency mode, customizable EQ, and multipoint connectivity. They're not the most advanced headphones on the market by any means, but they'll get the job done.

Not to rag on Apple, but for less than half the price of AirPods Max, you can get a solid pair of noise-cancelling over-ear headphones and premium sleep earbuds from Anker Soundcore. In this economy, that's a major win.

The iPhone 17, fancier Pro version, and an all-new slim model are just slightly better iterations of the iPhone you already have in your pocket. While everyone loves an upgrade, not everyone needs an upgrade.

SAVE $300: As of Sept. 11, get the Roborock S8 MaxV Ultra robot vacuum and mop for $999.99, down from its usual price of $1,299.99 at Amazon. That's a discount of 23% and the lowest price we've seen.

Opens in a new window Credit: Amazon Roborock S8 MaxV Ultra $999.99 at Amazon
$1,299.99 Save $300 Get Deal

Ready to stop spending so much of your time vacuuming and mopping manually? This fall, let the robots handle it with this fantastic deal on a robot vacuum from Amazon. Not only will it save you some serious money, but also the most precious commodity: time.

As of Sept. 11, get the Roborock S8 MaxV Ultra robot vacuum and mop for $999.99, down from its usual price of $1,299.99 at Amazon. That's $300 off and a discount of 23%. It's also the lowest price we've seen.

SEE ALSO: The best robot vacuums in 2025

This premium robot vacuum and mop combo is a great pick for automating your entire cleaning routine. It boasts a self-cleaning, self-emptying docking station as well as AI navigation and 10,000Pa of strong suction so you don't have to worry about going over floors again to suck up every bit of debris.

This model really can do it all, from automatically switching from vacuuming to mopping to save you time, auto-emptying its dust bins, and using intelligent dirt detection to find where all the grime is lurking. It even handles heated air drying for mop messes, so you're good to go minutes after mopping the floor. No more gross puddles to have to worry about drying yourself.

If you're tired of handling all these tasks, this robot vacuum and mop combo can make things significantly easier on you. Just be sure you jump on the deal quickly as it's currently a Lightning Deal, which means it can disappear fast.

Throughout her career, Sydney Sweeney has proven herself to be an accomplished genre hopper, tackling teen dramas like Euphoria, rom-coms like Anyone But You, and horror films like Immaculate. But with Christy, she leaps into a new genre: the Oscar bait biopic.

Directed by David Michôd (The King, Animal Kingdom), Christy stars Sweeney as famed boxer Christy Martin, a pioneer for women in the sport. While the film follows Christy's boxing journey, it also examines her traumatic upbringing with her homophobic mother Joyce (Merritt Wever) and her relationship with her abusive husband and trainer Jim (Ben Foster).

The latter takes up the bulk of Christy's first trailer, with Jim even threatening to kill Christy should she ever leave him. Also on display in the trailer is Sweeney's muscular transformation into Martin, the kind of physical commitment that's catnip for Oscar voters. (See: Christian Bale in The Fighter or Charlize Theron in Monster.) But beyond its Oscar bait-y premise, how does the film actually stack up?

SEE ALSO: 'Christy' review: Sydney Sweeney's Oscar-bait swing is a big miss

Following its world premiere at 2025's Toronto International Film Festival, Christy garnered mixed reviews. Some critics praised Sweeney's performance while pointing out the film's flawed, even bland storytelling.

In her review out of TIFF, Mashable Entertainment Editor Kristy Puchko criticized Sweeney's performance along with the film as a whole, writing, "Despite the premiere buzz, Christy is not the Oscar contender that Sweeney, who also produced the film, seems to have swung for. Its script is a mess, creating clumsy archetypes and hitting on Lifetime movie cliches with no self-awareness."

Christy hits theaters Nov. 7.

Imagine this: you come home, turn on your PC or laptop, and it won’t boot. Now what? You had important work to do. Panic sets in, and you don’t know what to do next.

The Tesla Model Y may still be the best-selling electric crossover in 2025, but cracks are starting to show. Sales have slipped hard over the past few months, and plenty of shoppers are eyeing other options.

X (formerly Twitter) seems to be dominated by videos when you open the app or access the site — and if the news of the day happens to be something violent or disturbing, it can be easy to stumble across content that you don't want to see automatically playing.

The safest way to avoid disturbing social media content is to avoid the apps altogether, but if you do want to access X on a day when unpleasant news is trending, there's a setting you can tweak that will at least stop videos from autoplaying in your feed.

SEE ALSO: Twitter lets you put content warnings on your tweets. Here's how to do it. How to stop videos autoplaying on X Total Time

• 1 minute

What You Need

• X on desktop or smartphone

Step 1: Navigate to "Settings and privacy"

This is under "More" if you're using desktop.

Credit: X

Step 2: Navigate to "Accessibility, display and languages"

Step 3: Click/tap on "Data Usage"

Credit: X

Step 4: Click/tap "Autoplay", then select "Never"

This will stop all videos from autoplaying when you're using X.

Credit: X

It's worth noting that while you can stop autoplaying videos, this won't remove videos from X altogether — and there's still a chance you may see disturbing images while browsing on the app.

If you've seen violent or disturbing content online, and need to talk to somebody, here are some resources:

Crisis Text Line / 988lifeline.org / Samaritans / Find a Helpline

To report harmful content go to the Safer Internet Centre.

Pre-order 'Space Marine 2: Ultima Limited Edition' on Amazon Pre-order for Xbox Space Marine 2: Ultima Limited Edition $99.99 on Amazon Pre-Order Here Pre-order for PS5 Space Marine 2: Ultima Limited Edition $99.99 on Amazon Pre-Order Here

When Warhammer 40,000: Space Marine launched in 2011, it stood apart from the cover-heavy shooters of its era. Where Gears of War and Mass Effect 3 had you crouched behind walls, Space Marine sent you charging into the fray with a chainsword in one hand and a bolt rifle in the other. It was messy, fast, and brutal in a way other shooters weren’t at the time.

The sequel, Warhammer 40,000: Space Marine 2, keeps that same energy but scales it way up. Developed by Saber Interactive, best known for World War Z, it is a third-person action game where you carve through endless Tyranid swarms, swap between six classes in replayable three-player missions, and test your skills in 6v6 PvP. Think God of War’s close combat mixed with World War Z’s swarms, only here you are a genetically engineered supersoldier in bright blue power armor.

SEE ALSO: 'Clair Obscur: Expedition 33 Mirror Edition' comes to PS5 and Xbox Series X: Release date and price info

Now Space Marine 2: Ultima Limited Edition is due Nov. 18 with some physical goodies and all DLC and season pass content released so far.

What's inside the Ultima Limited Edition

The Ultima Limited Edition is the most complete version of the game available. It includes:

• Base game with campaign, co-op missions, and PvP

• SteelBook case with exclusive artwork

• Macragge’s Chosen DLC with an Ultramarines cosmetic pack that adds a chainsword skin, bolt rifle skin, and Crux Terminatus shoulder piece, all usable in campaign, operations, and multiplayer

• Season Pass 1 with a mix of cosmetic packs and champion packs:

  • Ultramarines Cosmetic Pack with over 30 items including heraldry decals, successor chapter markings like Novamarines and Praetors of Orpheus, unique helmets, and new paint colors

  • Dark Angels Chapter Pack with more than 40 items tied to the Dark Angels and their successors

  • Salamanders Champion Pack with a full-body power armor set for the Sniper class and a matching Bolt Sniper Rifle skin

  • Raven Guard Cosmetic Pack with heraldry and paint schemes

  • Imperial Fists Champion Pack with a Tactical class armor skin and weapon

  • Space Wolves Chapter Pack with successor heraldry and colors

  • Blood Angels Cosmetic Pack with heraldry and armor markings

  • White Scars Chapter Pack with chapter-specific decals and paint

• Season Pass 2 with more customization and champion packs:

  • Black Templars Champion Pack with full armor for the Bulwark class and a Power Sword skin

  • Imperial Fists Cosmetic Pack with more than 40 cosmetics across Imperial Fist successors

  • Ultramarines Champion Pack, a class-specific full armor set, coming after launch

  • Additional packs and chapters to be announced

In short, Chapter Packs focus on modular cosmetics like decals, paint colors, helmets, and greave markings, while Champion Packs deliver complete armor sets for one class plus a matching weapon.

Is Space Marine 2 newcomer friendly?

Warhammer 40k is famous for having decades of lore, thousands of pages of novels, and a tabletop game that can be overwhelming if you are not already a fan. The good news is that Space Marine 2 is designed to be approachable even if you know nothing about the setting.

The story follows Captain Titus of the Ultramarines, a character from the first game, and throws him into a clear fight against a terrifying alien swarm. You do not need to know the full history of the Imperium, the Horus Heresy, or the countless Space Marine chapters to enjoy the campaign. The focus is on action, spectacle, and survival rather than deep lore.

Who are the Tyranids?

The main villains in Space Marine 2 are the Tyranids. They are an alien hive species that exists only to consume. Imagine the xenomorphs from Alien mixed with the endless Zerg from Starcraft, then dial the scale up until they devour whole planets. They are not individuals but part of a collective hive mind that overwhelms enemies with sheer numbers. For a newcomer, all you really need to know is that they are terrifying, they never stop coming, and killing them in huge swarms is the core of the game.

SEE ALSO: '007: First Light' could be the best James Bond video game since 'Goldeneye'

For players who want to learn more, the game includes nods and references for fans of the wider universe, but it never requires you to read outside material to understand what is happening. At its core, the story is simple: you are a genetically engineered soldier defending humanity against alien monsters. That makes it an easy entry point for anyone curious about Warhammer without the homework.

What comes after launch

The Ultima Edition and base game release worldwide on Nov. 18, 2025. Focus Entertainment has already confirmed more content will follow:

• Ongoing Season Pass 2 drops with new Champion Packs and Chapter Packs

• Season 3 and Season 4 are planned but details are not yet announced

• Some packs may release on consoles first, with PC players getting them later

Should you preorder the Ultima Limited Edition?

If you only want to play the campaign, the standard edition will be enough. But if you plan to dive into co-op operations or PvP where customizing your Marine matters, the Ultima Edition is a strong value. It includes two full season passes and the Macragge’s Chosen DLC upfront, saving you from buying them separately later.

Warhammer 40,000: Space Marine 2 Ultima Limited Edition is available to preorder now for $99.99 on PlayStation 5 and Xbox Series X. It launches worldwide on Nov. 18, 2025.

While the brand has lost some of its luster in recent years, the average person still associates Marvel movies with the MCU, the cinematic universe that brought us everything from Iron Man to The Thunderbolts. But the MCU wasn’t born until 2008, and there were plenty of Marvel films created before that. And even after that, there have been more films made outside of the MCU umbrella than you can shake an Infinity Gauntlet at.