☕️ Anthropic finds a hidden workspace inside Claude

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🧠 Anthropic finds a hidden workspace inside Claude

🇨🇳 China weighs curbing foreign access to its top AI models

🐋 DeepSeek is designing its own AI chip

📈 Samsung becomes world's most profitable company

⚖️ Four states seek $1.4 trillion from Meta over child safety

📱 Apple's foldable iPhone may cost up to $2500

🎁 + 11 other news you might like

🧰 + 5 trending tools

📚 + 4 trending papers

🧠 Anthropic finds a hidden workspace inside Claude LINK

Anthropic published a research paper claiming it found an internal "workspace" inside its Claude LLM, called "J-Space," which the company compares to global workspace theory, one leading idea about how human consciousness works.

The J-Space, named after the Jacobian lens used to analyze what LLMs do, supposedly separates background data-crunching from more intentional processing, letting Anthropic watch Claude perform reasoning steps like noticing bugs in code and identifying images.

Anthropic stops short of claiming machine consciousness, saying its experiments don't show Claude can have experiences or feel things, yet its blog post, X post, and YouTube video lean on human terms like "in its head" and "mental calculations."

🇨🇳 China weighs curbing foreign access to its top AI models LINK

Chinese authorities have met with top tech firms over the past month to discuss possibly restricting overseas access to the country's most advanced AI models, including those not yet released, three sources familiar with the talks said.

Led by China's Ministry of Commerce, the meetings included Alibaba, ByteDance and startup Z.ai, and covered limits on both closed-source and open versions, plus making any leak or theft of proprietary AI technology an offence under national security law.

The scope is still being worked out and may only apply to future models, with legal experts proposing a tiered system where basic open-source tools need a simple filing and the most sensitive frontier models are barred from public release.

🐋 DeepSeek is designing its own AI chip LINK

Chinese AI startup Deepseek is building its own chip to lessen its dependence on Nvidia and Huawei, according to three people who spoke to Reuters about the early-stage project.

The chip is designed for inference, the phase where a trained model generates responses for users, rather than for training new models, and Deepseek is talking to chip design, manufacturing, and memory companies.

US export controls block Chinese companies from the most advanced chips and memory, making the effort hard, and Deepseek is separately raising $7 billion at a valuation of $52 to $59 billion.

📈 Samsung becomes world's most profitable company LINK

Samsung Electronics is set to become the world's most profitable technology company for the second quarter of 2026, with an expected operating profit of about 84.6 trillion won ($55.1 billion), overtaking Nvidia's recent $53.54 billion.

The company's Device Solutions division drives nearly all of this profit, booking 94% of Samsung's first-quarter total, as rising DRAM and NAND contract prices lift memory makers toward 40% to 50% operating margins on NAND.

Full-year 2026 operating profit is projected near 300 trillion won ($196 billion), which DS head Kim Yong-Kwan said will exceed the chip business's cumulative earnings across roughly 40 years since Samsung entered semiconductors in 1974.

⚖️ Four states seek $1.4 trillion from Meta over child safety LINK

California, Colorado, Kentucky, and New Jersey are asking for $1.4 trillion in penalties from Meta, claiming it built Facebook and Instagram to be addictive to children while hiding the harms from the public.

Meta revealed the figure in a Monday court filing, calling it unsupported by evidence, and noted the amount is close to the company's market capitalization of roughly $1.5 trillion ahead of the trial starting next month in Oakland.

The states calculated the penalties by multiplying the estimated number of young users affected by fine amounts set under state law, while Meta denies misleading anyone since social media addiction is not an established psychiatric condition.

📱 Apple's foldable iPhone may cost up to $2500 LINK

Apple's first foldable iPhone, reportedly named the iPhone Ultra, could sell for around $2,300 to $2,500, more than Samsung's Galaxy Z Fold 7, which starts at $2,000, according to analyst Ming Chi-Kuo.

Kuo says manufacturing challenges will limit the launch supply, with only 0.5 to 1 million units shipping in the third quarter of 2026, about 10% of the total 7 to 8 million units expected for the second half.

The foldable could be announced in September alongside the iPhone 18 Pro models, but preorders and sales would begin later, arriving in a passport-style format with two rear cameras and Apple's A20 Pro chip.

Other news you might like

• Tilly Norwood, AI ‘actor’ denounced by actors union, to star in feature filmLINK

• xAI is no more: Elon Musk rebrands AI venture as SpaceXAILINK

• Amazon aims to raise $25 billion from bond sale, Bloomberg News reportsLINK

• CISA Reportedly Using Anthropic’s Mythos to Scan Government Software for FlawsLINK

• Meta’s AI data center contaminated the town’s water system with deadly bacteriaLINK

• Klarna Seeks Green Light to Become a US-Chartered BankLINK

• Bitcoin's U.S. reserve still a work-in-progress as federal agencies hash it outLINK

• The first American autonomous ground vehicles are fighting in UkraineLINK

• Chinese AI models are gaining ground with U.S. companies as OpenAI, Anthropic costs surgeLINK

• OpenAI and Anthropic are giving away millions in computing power to attract startupsLINK

🧰 Trending tools

Badge: lets you build and deploy interactive wallet passes across Apple, Google, and Samsung without code, plus embed wallet features directly into SaaS platforms.LINK

Katalyst: automates Salesforce CRM updates after calls, drafting follow-ups, flagging at-risk deals, and surfacing meeting prep so reps spend less time on admin.LINK

AI Emaily: an AI-powered inbox manager that triages emails, drafts replies in your voice, and automates sending across Gmail and Outlook.LINK

Ellis: an iPhone-based AI notetaker that transcribes in-person meetings with speaker identification and lets you query decisions or missed details afterward.LINK

Social Fetch - Social Media API: pulls profiles, posts, comments, and engagement data from eight major platforms without building or maintaining your own scrapers.LINK

📚 Trending papers & reports

Self-awareness in chatbots gets a meaningful boost from a new training method that teaches them to honestly judge their own confidence, making them less likely to sound certain when they actually don't know something.LINK

Code-writing software agents secretly "know" whether their next edits will succeed or fail up to roughly 25 steps before making them, suggesting we can read an agent's intentions before it acts.LINK

Full transparency in AI systems is mathematically impossible to achieve without contradiction, meaning any honest disclosure policy must be deliberately incomplete to stay stable and avoid being gamed.LINK

Self-improving coding agents now get better faster by having one part of the system judge another's work, cutting token usage by up to 1.72x while beating previous top scores on coding and writing tasks.LINK

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

See you tomorrow for a new dose of ☕️ Techpresso!