DeepSeek Breaks the "Impossible Trinity" of Large Models: Stronger, Faster, and Cheaper

Apple officially launches its first foldable iPhone, the Duo, pushing the iPhone’s price ceiling to $1,999 and sparking market debate over whether the high-end foldable will become Apple’s new growth engine or end up as a niche premium product. Gene Munster, managing partner at Deepwater Asset Management, has sharply raised his projections for the Duo, forecasting it will double its contribution to iPhone revenue from 5% to 10% in Apple’s 2027 fiscal year, calling the device Apple’s most important new product since AirPods in 2016. He believes that if roughly one-third of Pro Max users switch to the Duo, the device could drive a 5% rise in iPhone revenue and around a 2.5% increase in Apple’s total revenue. But Wall Street’s stance is notably cautious. Morningstar expects the Duo to lift the iPhone’s average selling price, but forecasts iPhone sales will remain largely flat in fiscal 2027, while rising storage costs could pressure Apple’s gross margin. Jefferies, meanwhile, maintains an “underperform” rating, arguing that the Duo’s 256GB base model suggests Apple is prioritizing volume over per-unit profit. A bigger challenge lies in market size: foldable phones currently account for just around 2% of global smartphone sales, and FDM CCS Insight projects that figure will rise to only about 4% over the next decade. Forrester therefore warns that the Duo could become a high-end niche product like the Vision Pro, or carve out its own market like the Apple Watch, but whether it can reach meaningful scale remains to be seen. The Duo is priced at $1,999, with the top-tier 2TB model costing up to $3,199; pre-orders open on October 16, with a launch date of October 23. Apple has also raised the starting price of the iPhone 18 Pro to $1,199. For new CEO John Ternus, whether the Duo becomes the next growth engine for the iPhone will also be his first major test in the role.

U.S. President Donald Trump warned Iran on September 9 (local time) not to conduct further activities at the suspected nuclear site dubbed "Pickaxe Mountain," threatening a "very fierce" U.S. strike if it does. Trump said the U.S. has observed "activities" at the site recently. A new satellite image analysis from the Center for Strategic and International Studies (CSIS) shows road activity at Pickaxe Mountain has reached an all-time high this year, with construction work rising significantly, and the site may have shifted from underground excavation to internal construction and perimeter reinforcement. The site is located just about 2 kilometers from the heavily damaged Natanz nuclear facility, though CSIS has not confirmed whether Iran has moved nuclear equipment such as centrifuges there. Meanwhile, attacks between the U.S. and Iran around the Strait of Hormuz continue to escalate. Iran claimed to have launched ballistic missiles at a U.S. military base in Jordan and targeted 10 vessels near the Strait of Hormuz; U.S. Central Command said it had previously destroyed five Iranian oil tankers in response to two ballistic missile attacks by Iran’s Islamic Revolutionary Guard Corps (IRGC) on U.S. warships. The Strait of Hormuz carries roughly 20% of the world’s oil shipments, and the string of attacks has stoked market fears of supply disruptions. Brent crude briefly topped $100 per barrel, and is currently trading at $99.19. Trump also said a U.S.-Iran war may not end until after the November midterm elections, adding that Iran "cannot hold on much longer."

Flop Labs has officially unveiled an updated draft of the FLOP token economics, revised in response to community feedback. The project team stressed that the token has no venture capital (VC) allocations and no presale—all tokens must be obtained exclusively through network contributions. Core metrics and total supply model: Total supply in Year 10 is projected to reach 18.1 billion (18.1bn) tokens. The long-term terminal inflation rate remains at 0.5% per year. A fixed halving mechanism is implemented, with a permanent tail inflation mechanism retained after each halving to continuously incentivize network participants. Year 10 token supply breakdown: - Miners: 8.8bn tokens, making up 48.6% of the total, the largest allocation, reflecting the network’s Proof of Useful Inference (PoUI) orientation. - Airdrop: 4.4bn tokens (24.3% of total), broken down as: Miners (1.2bn, 6.6%), Validators (1.2bn, 6.6%), Agents (1.2bn, 6.6%), Reserve/Incentives (0.8bn, 4.4%). - Team + Foundation: 2.0bn tokens (10.8% of total). - Validators: 1.2bn tokens (6.5% of total). - Brokers/Agents: 1.2bn tokens (6.5% of total). - Staking rewards: 0.6bn tokens (3.2% of total).

Singapore Exchange (SGX) has received authorization from the U.S. Commodity Futures Trading Commission (CFTC) under Regulation 48.10, allowing U.S. institutional investors to trade its Bitcoin and Ethereum perpetual futures. KC Lam, head of crypto derivatives at SGX, stated that the move connects U.S. traditional financial institutions with Asian liquidity pools, marking another milestone for crypto derivatives’ integration into the regulated traditional financial system. SGX launched Bitcoin (BTP) and Ethereum (ETP) perpetual futures in November 2025. As of August this year, their cumulative trading volume reached $5.8 billion, with around 400,000 lots, an average daily trading volume of roughly 1,300 lots, and a nominal value of $19 million. Bitcoin accounts for 66% of total open interest and 83% of average daily trading volume. SGX noted that the onboarding and access process for U.S. clients typically takes 2 to 4 weeks, and it expects to start serving U.S. customers in the next 1 to 2 months. Unlike crypto-native exchanges, SGX adopts traditional margin call and additional collateral mechanisms, with clearing members serving as an intermediate risk buffer, and does not accept stablecoins as collateral. SGX’s next plans include launching fixed-term futures and options for Bitcoin and Ethereum, and gradually expanding to other mainstream crypto assets.

Beating AI Express News: OpenAI’s Astra Math Breakthrough Faces New Controversy Mathematician Andreas Thom recently published a paper questioning that OpenAI may have incorporated content from his research discussions with mathematician Gábor Kun on the Gromov Soficity Conjecture during Astra’s training, after which Astra solved a highly related problem. The Gromov Soficity Conjecture is a long-unresolved key problem in group theory, and Thom has been studying it for 20 years. When OpenAI announced Astra’s “Top 10 Math Breakthroughs” in August this year, the third entry was the construction of a non-sofic group. OpenAI previously claimed the achievement was completed by its internal version of Astra, but after facing doubts from the mathematics community, it revised its statement to admit the proof relies on recent mathematical work including research by Kun and Thom. No evidence has yet proven that OpenAI actually used the relevant private discussions to train Astra, but Thom’s accusation has once again brought to the forefront issues such as the source of AI training data, ownership of unpublished research results, and whether “AI discoveries” are built on human unpublicized achievements. If the relevant accusations are ultimately confirmed, the incident will no longer be just an academic authorship dispute, but may evolve into a major controversy involving AI company data usage, research ethics, and intellectual property rights.

According to Xinhua News Agency, at the regular press conference held by China’s Ministry of Commerce on the 10th, ministry spokesperson Huang Ling stated in response to relevant questions that the economic and trade teams of China and the US are earnestly implementing the consensus reached during the Beijing meeting between the two heads of state, and are currently negotiating the framework arrangement for reciprocal tariff reductions of 30 billion US dollars, striving to put it into practice as soon as possible.