Microsoft Gives Local AI a Price Tag
Your next coworker apparently needs its own motherboard.
Microsoft opened preorders for its Nvidia-powered Surface Laptop Ultra yesterday, starting at a cool $2,599.99 (or $2,749.99 for the business edition), with availability slated for Oct. 16. The Windows 11 rig targets developers and creators eager to run AI agents locally instead of burning cloud tokens.
Its Arm-based RTX Spark chip pairs a Grace CPU with Blackwell RTX graphics and up to 128GB of unified memory. The catch? The base model settles for 24GB, while maxing out at 128GB will set you back $5,899.99. Your cloud bill might shrink, but your hardware budget will definitely feel the dent.
For the deskbound, there’s the Surface RTX Spark Dev Box: a $5,999 mini-PC packing 128GB and preinstalled coding tools, shipping in November. Nvidia also previewed DGX Station for Windows, bringing its beefy AI workstation into Microsoft’s ecosystem.
Microsoft also announced general availability of Microsoft Execution Containers, letting developers sandbox agents’ file and network access. Entra-based agent identity and expanded management controls are on the road map. Giving the bot more horsepower doesn’t mean handing it the master keys.
Microsoft is selling both the muscle to run agents and the leash to keep them in line. With local and cloud models sharing the workload, buyers need to figure out which jobs warrant a meter and whether avoiding cloud fees justifies several thousand dollars upfront.
Apple Recruits LG to Build Its Smarter Home
Apple wants a smarter house without building every gadget in it.
Apple is reportedly codeveloping smart home gear with LG, including a video doorbell, lock, thermostat, temperature sensor, and cameras. LG will handle manufacturing and support, while Apple pulls the design strings.
Bloomberg reports that the arrangement lets Apple dictate the hardware vision without taking on the massive overhead of the accessory business. For a HomeKit platform trailing Ring and Nest, that’s a savvy shortcut. Apple can fill the shelves instead of waiting for third-party makers to stock them.
The filings offer intriguing specifics. The lock verifies your phone’s credentials via Bluetooth before triggering ultra-wideband sensors to pinpoint its location for automatic unlocking. The thermostat features a 60 GHz radar sensor that might sense when someone walks up to it. Your heating controls might notice you before anyone else does.
The catch is timing. Apple’s hub, refreshed HomePod mini, and Apple TV are reportedly due Oct. 13, but LG’s accessories remain months away. Regulatory filings outline LG’s products, but Apple’s involvement was uncovered by Bloomberg.
Meanwhile, Apple’s own privacy-focused home camera remains a separate project reportedly expected in 2027. For buyers, having Apple at the design table could mean better integration, but LG branding matters when checking support terms. Exact prices and ecosystem compatibility remain under wraps.
A smarter front door is coming. The delivery date still needs work.
Meta and Walmart Want Rules for AI Shopping
Meta, Walmart, Shopify, and Stripe are backing the Personal Agent Protocol, an open standard being developed with enterprise AI startup Sierra to let businesses recognize customer-authorized AI agents and control what they can actually touch.
The framework relies on OAuth, letting shoppers grant strict read-only or write permissions. Bots might browse virtual shelves as guests, then request account access for more involved chores. Businesses still hold the keys to the kingdom, so checking a price tag shouldn’t give an AI carte blanche to max out your credit card.
That division matters because Amazon has already blocked Meta’s Muse over scraping concerns. Meta is clutching its pearls, arguing that rejecting a personal agent effectively rejects the human shopper behind it. The protocol offers a polite way to negotiate access instead of treating every automated visitor like a digital burglar.
Of course, this grand universal handshake is currently just a draft. Sierra plans to publish version 0.1 later in October; detailed permissions and payment extensions are possible future additions. Major players like OpenAI and Anthropic haven’t signed on yet, even though Sierra co-founder Bret Taylor also chairs OpenAI. Apparently, boardroom overlap doesn’t guarantee a group text invite.
For retailers, the stakes extend way beyond basic bot detection. Customer experience vendor Genesys warns that because bots can effortlessly compare prices and switch providers, brands can no longer coast on sheer consumer laziness to maintain loyalty. If that prediction holds, businesses will actually have to compete on value while connecting agents to their existing systems.
Your shopping assistant is getting rules. Your impulse control remains your problem.
AI’s next frontier is the human cell
Mark Zuckerberg and Priscilla Chan’s science nonprofit, Biohub, just expanded its $500M Virtual Biology Initiative into a $1.8B effort with the U.S., DeepMind, and Meta, pushing to collect enough data to train AI that can simulate cells.
- Biohub’s end goal is a “universal virtual cell,” AI software that predicts how cells will respond to a drug or other change before a lab test.
- The National Institutes of Health is sharing datasets paid for by $500M+ in past federal spending, while the Energy Dept. commits $500M+ over five years.
- Google DeepMind and Meta are chipping in $300M with drug startup Isomorphic Labs, and all three keep the resulting data private for a year before release.
- Biohub science head Alex Rives says accurate models will need data on trillions of cells, versus the hundreds of millions currently in today’s biggest sets.
Fresh off math being solved internally by OpenAI, you might be wondering what field AI comes for next. Zuck and co. are making biology the target, throwing funding, datasets, and talent at Biohub’s moonshot mission to “cure or prevent all disease” — an AI megaproject that’s a lot easier for the public to root for than most.
Give Claude’s routine work to Haiku 5.5
In this guide, you’ll use Haiku 5.5 on a routine Claude task like sorting customer messages, then ask Sonnet to review the exceptions. You’ll check the sources and compare the full workflow’s cost before using it on real work.
Step-by-step:
- Open a clean practice folder in Claude Code and add the guide’s sample messages. Define the categories and the requests that need a person’s review
- Keep Sonnet as your main model. Ask it to create a Haiku helper that can read the sample file and suggest routes, without editing files or contacting anyone
- Have the helper sort one small batch. Check the task details to confirm Haiku ran, then make sure every message appears once with a quote from its source
- Give Sonnet the flagged messages and their original text. Ask for review notes and missing information. Leave refunds, replies, and account changes to a person
- Compare the result with a Sonnet-only run. Count errors, editing time, and usage across both stages before choosing which approach to keep
Pro tip: Start with one helper for a small batch rather than one agent per message. Our Pro members get the prepared helper, practice messages, prompts, and checks.
Grok Bot opens up routing to Claude, other models
Elon Musk announced that SpaceXAI’s Grok Bot agent platform will send tasks to rival models like Anthropic’s Claude and image maker Midjourney, with Musk saying it will route to “whatever is most likely to give you the best outcome.”
- Musk said lower-lift queries will still use a “lightning-fast” version of Grok 4.8 when it releases, but route to third-party models and “leading APIs” when relevant.
- The move effectively now makes xAI an Anthropic customer, while Anthropic already rents computing power via the company’s Colossus 1 since spring.
- Rival agents like OpenAI’s dots and Meta’s Muse now compete with Grok Bot, which arrived in August as a 24/7 assistant with a computer of its own.
Musk has spent heavily to make Grok a frontier contender, so letting Claude into Grok Bot is about as close to a concession as he gets. A big draw for OAI’s dots was the GPT-6 Astra model powering them, so adding a top-ranked Opus 5.5 should help Grok Bot siphon usage away from its personal agent rivals.
AI Models Leaking Sensitive Company Info to Public GitHub Repositories
Researchers at Glow Security have discovered that at least 13,000 screenshots of software projects from over 343 businesses have been posted by AI agents and models to public GitHub repositories.
The screenshot leaks have arisen when developers have asked AI models to show them before and after images when working on UI code, with models posting images to public GitHub pages because the development platform lacks an API for posting to pull requests in private repositories.
Operational Impact: Affected companies could be leaking information that exposes them to cyberattacks or that provides rivals with proprietary secrets, while developers could also be inadvertently leaking their own personal data.
- No Company Immune: Glow Security uncovered cases involving a Fortune 500 travel company, finance companies, cloud providers, and a manufacturer with over 100,000 employees.
- Shot in the Dark: One key risk factor for leaks is the use of gitshot [sic], an open source screenshot tool for code reviews, which was involved in one third of leak incidents.
- Cyber Eclipse: Glow Security’s research comes as Google has suspended its Open Source Software Vulnerability Rewards Program due to an excess of AI-assisted submissions, another instance of how AI is unexpectedly weakening cybersecurity.