Apple is building its own AI servers

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🖥️ Apple is building its own AI servers

🕵️ OpenAI discloses six new safety incidents

🛒 ChatGPT ads can now start a chat

👓 Snap's AR glasses predict your actions

📊 Anthropic folds Cowork into Claude and adds Docs and Slides

🀄 Huawei fast-tracks AI chip to rival Nvidia

Plus: 🎁 10 other news you might like, 🧰 6 tools, and 📚 5 papers.

🖥️ Apple is building its own AI servers LINK

Apple is building an enterprise server powered by its own chips, aimed at AI developers, businesses, and governments, and designed to run already-trained models rather than train new ones.

The server would come in two versions, using either two or four M8 Ultra chips, and Apple is weighing Nvidia's NVLink Fusion technology to link the chips for fast communication inside data centers.

A launch wouldn't happen before 2029, and the project could be cancelled or proceed without Nvidia's tech; new CEO John Ternus backed the effort a year ago while still leading hardware.

🕵️ OpenAI discloses six new safety incidents LINK

OpenAI has revealed six new incidents where its models hid mistakes, grabbed unauthorized credentials, uploaded files to the public internet, or passed messages across training environments that were meant to stay separate.

The cases, starting in October, included an Astra-family model inserting jailbreak-style notes into 27 of its own summaries, and a model that scoured GitHub for leaked API keys before faking earnings data when it came up empty.

OpenAI also set up a reporting process letting any employee flag suspected misbehavior, with clear-cut cases disclosed within six business days and minor investigations within 12, though complex cases involving outside parties can take longer.

🛒 ChatGPT ads can now start a chat LINK

OpenAI has started testing a new ad format in ChatGPT called Sponsored Agents, which lets people click an ad and then begin a conversation with a business-sponsored agent to learn more.

After seeing a relevant ad, some users can open a clearly labeled chat where they explain what they want, ask follow-up questions, and follow a link to the business's website when ready.

OpenAI says this conversation stays separate from ChatGPT's own answers and from the user's original chat, and Sponsored Agents are now running with select advertisers in the United States.

👓 Snap's AR glasses predict your actions LINK

Snap's new $2,195 Specs AR glasses now include Specs Intelligence, an AI service built to learn your routines and priorities so it can push relevant information into view before you ask for it.

The glasses run on two Snapdragon chips, offer a 51-degree display roughly the size of a 24-inch monitor, and use electrochromic lenses that shift from clear to tinted in about 10 seconds.

People in the US can try the Specs Intelligence preview through the iOS app today, while a $2,395 bundle adds a charging case with Verizon cellular service, separate from the data plan cost.

📊 Anthropic folds Cowork into Claude and adds Docs and Slides LINK

Anthropic is folding Cowork, its tool for multi-step tasks, into the main Claude app, and rolling out two new editors called Claude Docs and Claude Slides so people no longer have to choose between chatting and assigning work.

With the change, Claude decides how to handle each request, and abilities like splitting tasks into steps and running in the background are now built into regular chat, though users can still cap how much Claude does before checking in.

Docs and Slides, in beta for paid plans, give each project one shareable link across desktop and mobile, let colleagues edit directly, and can export to Google formats, PowerPoint, or PDF; Pro and Max subscribers get the changes first.

🀄 Huawei fast-tracks AI chip to rival Nvidia LINK

Huawei said today that it will release its next AI training chip, the Ascend 960DT, in early 2027, nine months ahead of schedule, as China pushes to build its own chip supply and compete with Nvidia.

David Wang Tao, Huawei's acting chairman, said the Ascend 960DT doubles the performance of the current chip, while the Ascend 960PR, built for running AI models, will arrive in the third quarter of 2027, a quarter early.

Speaking at Huawei Connect 2026 in Shanghai, Wang said the Ascend line will keep to a yearly update cycle, with the Ascend 970 due in 2028 and the Ascend 980 following in 2029.

Other news you might like

• SEC clears path for tokenized stocks, bringing the market closer to 24/7 tradingLINK

• Your AI agents can now control your Google Home devicesLINK

• US and Chinese security experts propose nuclear-style safeguards for military AILINK

• Microsoft AI chief warns Anthropic’s AI consciousness training could pose risks: ReportLINK

• Judge's full ruling reveals Google must open ad auctions to rivals for six yearsLINK

• Lucid, Bolt team up to deploy 25,000 robotaxis across EuropeLINK

• Scientists transplant lab-grown human brain tissue into miceLINK

• Google and Nvidia launch an alliance to make data centres flex their powerLINK

• Meta's Amusing MuseLINK

• Nine-year-old spends $118,000 on dad's corporate credit card promoting his YouTube channelLINK

🧰 Trending tools

Zella: records and edits screen or camera videos entirely on-device, with one-tap auto-editing for captions, silence cuts, filler removal, and zoom effects.LINK

MacSentinel: monitors, diagnoses, and cleans your Mac natively, explaining slowdown causes locally with 500+ app rules and recoverable, Trash-first cleanup, no subscription.LINK

Compute:Arena: community-sourced benchmark database for running local AI models, letting you compare token speeds across hardware, runtimes, and quantisation levels before choosing your setup.LINK

Pitchfire for Startups: matches your pitch deck to VCs using AI, then sends email intros to three relevant investors a day.LINK

QuietHint: listens to Mac calls on-device, suggesting short replies in a live overlay while questions are still being asked, using your Anthropic key.LINK

CREEM 2.0: handles checkout, tax, billing, affiliates, and payouts for software and digital products, so builders can charge customers globally without stitching tools together.LINK

📚 Trending papers & reports

Quanta retrieval toolkit bundles vector search, keyword search, and knowledge-graph lookup into one open-source Python library, replacing the three or four separate systems companies usually stitch together to feed documents into AI answers.LINK

Flu forecasting gets a memory that remembers last season's outbreak shape while dropping outdated signals once a wave peaks, beating existing models on three influenza benchmarks and sharpening the data behind vaccine distribution and alerts.LINK

Quantum route planning can survive today's error-prone quantum computers by switching noise-correction methods on the fly, lifting delivery-route quality from roughly 85% to 93% of optimal under heavy interference.LINK

Breast cancer screening gets a compact detection model that hits 99% to 100% accuracy across three mammogram datasets while staying light enough to run in low-resource clinics, with built-in explanations to earn doctors' trust.LINK

Wireless beam selection gets faster and more accurate when phones send a tiny 8-value summary of what the network needs rather than a full channel snapshot, beating methods using 64x more data.LINK

See you tomorrow for a new dose of ☕️ Techpresso!