☕️ China debuts biggest AI model trained on local chips

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🇨🇳 China debuts biggest AI model trained on local chips

🧠 Meta's new AI turns your thoughts into text

🔒 Apple rushes security updates to beat AI hackers

⚖️ Supreme Court reins in police phone tracking

💬 WhatsApp now lets you reserve usernames

💳 UK may force Apple to allow Apple Pay rivals

🎁 + 16 other news you might like

🧰 + 6 trending tools

📚 + 5 trending papers

🇨🇳 China debuts biggest AI model trained on local chips LINK

Meituan launched LongCat-2.0 on Tuesday, a 1.6-trillion-parameter model the company says is the first of its size to be trained entirely on domestically developed chips, a direct response to US export controls.

The model carries a context window of one million tokens, completed end-to-end training and inference on a 50,000-chip domestic compute cluster, and Meituan says its performance matches Google's Gemini 3.1 Pro, released in February.

Meituan open-sourced the model, releasing the weights so the community can run LongCat-2.0 against the cited benchmarks, though the training-hardware claim is harder for outsiders to confirm since it rests on Meituan's own account.

🧠 Meta's new AI turns your thoughts into text LINK

Meta has introduced Brain2Qwerty v2, an AI system that turns brain activity into text using non-invasive brain recordings, aiming to help people who lost the ability to communicate because of brain lesions.

The system records brain activity with a helmet-like magnetoencephalography (MEG) scanner, then feeds the raw neural signals into an end-to-end AI model that reconstructs sentences a person is trying to type.

Trained on roughly 22,000 sentences from nine volunteers, Brain2Qwerty reached 61% average word accuracy, far above the 8% of earlier non-invasive methods, and Meta is releasing the code and dataset openly.

🔒 Apple rushes security updates to beat AI hackers LINK

Apple is releasing security fixes earlier than its usual schedule, pulling them out of the annual iOS cycle into standalone updates, because AI is shortening the time attackers need to weaponise a known weakness.

Apple told Reuters the change is preventive, arguing that the gap between a vulnerability becoming public and being exploited has narrowed, since AI tools can read code and summarise a diff to build a working exploit faster.

The company said there was no evidence the newly patched vulnerabilities had been exploited, and it did not specify how much earlier updates would arrive or which categories of fix would be pulled forward.

⚖️ Supreme Court reins in police phone tracking LINK

The US Supreme Court ruled that police need a warrant to obtain location history and other phone data that people share with companies like Google and Apple.

The 6-3 decision in Chatrie v. United States found that people keep a reasonable expectation of privacy from the government, extending Fourth Amendment protections to data handed over to tech companies.

The case stemmed from a 2019 bank robbery in which police used a geofence warrant to find all devices located near the scene of the crime.

💬 WhatsApp now lets you reserve usernames LINK

WhatsApp is letting you reserve a username this week ahead of a full launch later this year, so you can chat with new people and join group conversations without sharing your personal phone number.

With over three billion people on WhatsApp causing many names to overlap, reservations are opening early, and creators, small businesses, and organizations can claim their existing Instagram or Facebook username on the app.

There's no directory or suggestions, so people need your exact username to contact you, and an optional username key lets you control who can message you while hiding your phone number from first-time contacts.

💳 UK may force Apple to allow Apple Pay rivals LINK

Britain’s Competition and Markets Authority has proposed forcing Apple to let third-party rivals to the Apple Wallet app handle contactless payments, opening the door to competitors to Apple Pay on iPhones.

Such payments require access to the NFC chip built into modern iPhones, which Apple currently blocks banks and other financial institutions from using, and the CMA wants to change that restriction.

Apple opposes the proposal, and the article argues the case is weak because Apple’s fees are very small and users are unlikely to abandon having all their contactless payment options inside one app.

Other news you might like

• Apple’s biggest leak in years shows the iPhone 18 Pro surviving a drop testLINK

• Tidal won't monetize AI slop music, company saysLINK

• OpenClaw app for Android puts AI agents in your pocket and… looks like thatLINK

• Gemini’s personalized AI image generation is now free for US usersLINK

• Cursor now has a mobile app for guiding your coding agent on the goLINK

• UK finalizes sweeping crypto rules, raising the bar for global market standardsLINK

• Anthropic and Gov. Newsom forge deal allowing California government to use Claude at half priceLINK

• Amazon seeks cheaper AI alternatives as Anthropic shifts to token-based pricingLINK

• RAM prices expected to rise another 40-50% in Q3 2026, and then 30% more in Q4 as AI demand outpaces supplyLINK

• Marc Andreessen joins the Pentagon’s board, and the old line blursLINK

• Apple Design award winner acquired by Apple for new Swift toolsLINK

• Meta restricts use of Claude Code and Codex to keep rival AI out of its training dataLINK

• In San Francisco’s A.I. Era, Even $180,000 Tech Salaries Are No Longer EnoughLINK

• Drivers Are Not Happy With Their Cars' ScreensLINK

• Meta secretly tested ChatGPT, Gemini, and Character.AI with thousands of minor-perspective crisis promptsLINK

🧰 Trending tools

Cursor for iOS: a native mobile app that lets you launch cloud-based coding agents, monitor progress, and merge pull requests from your phone.LINK

Skills Marketplace by Databox: connect with vetted Databox experts who can build dashboards, configure analytics, and handle reporting setup on your behalf.LINK

Foresight by Lightning Rod: a calibrated forecasting API that gives developers accurate, scored predictions on future events, purpose-built to replace general-purpose LLMs in prediction workflows.LINK

Akiflow: a unified task and calendar manager that consolidates tasks from multiple tools, helping busy professionals plan and prioritize their day efficiently.LINK

Supafax: an email-native assistant that automatically prioritizes your inbox, drafts replies, and handles meeting scheduling without leaving your email.LINK

v0 Design Systems 2.0: lets you turn ideas into production-ready UI components instantly, using AI-assisted generation with built-in secure infrastructure.LINK

📚 Trending papers & reports

Pregnancy and period tracking apps were caught sending sensitive health data to advertising services, though some apps managed to show ads without leaking user data, proving better privacy is possible.LINK

Memory inside AI agents is fragmented and poorly defined, so this survey maps the whole landscape, giving builders a clearer shared vocabulary and practical guide to how agents store, update, and retrieve information.LINK

Long-document memory in language models gets faster and more accurate with a new shortcut method that summarizes text into compact markers, outperforming full reading by over 5.7 points on one benchmark.LINK

Building autonomous AI agents now has a single end-to-end practitioner's guide, covering everything from core model design to multi-agent coordination, memory systems, and real-world deployment in one place.LINK

Memory systems for AI agents have no single best design, a large benchmark study finds, showing that what works depends entirely on the specific task, with no architecture winning across all situations.LINK

Want to get the latest news differently? Find us on:

See you tomorrow for a new dose of ☕️ Techpresso!