Anthropic used a July 12 video to outline how Claude Platform has changed over the past six months, with engineering lead Katelyn Lesse and product lead Angela Jiang describing a move beyond a basic reasoning-and-token API into a broader agent stack. According to the discussion, the platform now centers on capabilities such as memory, outcomes, dreaming, and multi-agent coordination, with Managed Agents serving as a key product line for those ideas. Jiang said the strongest developer feedback around Managed Agents has focused on three concepts: memory, outcomes, and dreaming. She said users valued the clarity of abstraction, meaning agents can keep a clear identity and boundary even as workflows change. Lesse added that agents should eventually hold identities separate from end users, request only the permissions they need, and operate through their own service accounts so actions can be audited and controlled. The pair also described a change in how developers build harnesses around models. As model intelligence, tool calling, and reasoning improve, they said those external frameworks are getting thinner, while higher-level strategies such as competing agents, adversarial agents, and advisor agents are becoming more common. They pointed to two examples: Urrea using agent systems to preserve factory expertise that could disappear when specialists retire, and Shopify’s internal River system, which Lesse described as an end-to-end software development agent covering PRs, requirements, environment setup, coding, and QA.

Taiwan Semiconductor Manufacturing Co. (TSMC) reported June 2026 revenue of NT$442.68 billion on July 13, up 6.16% from May and 67.87% from a year earlier, setting a new monthly high. First-half consolidated revenue reached about NT$2.404 trillion, up 35.61% year over year, also a record. According to Economic Daily News, the chipmaker’s second-quarter performance landed within its earlier guidance of $39 billion to $40.2 billion, based on an exchange-rate assumption of NT$31.7 to the U.S. dollar. TSMC shares closed up 1.04% on Monday after the update. The report also pointed to ongoing AI-driven demand for advanced nodes and CoWoS packaging. In parallel, National Science and Technology Council Minister Wu Cheng-wen said TSMC plans to add two advanced packaging plants in Chiayi Science Park, with the first already in mass production and the second set to follow. Investors are now watching TSMC’s July 16 earnings call for updates on AI demand, CoWoS expansion, N2P and A16 production timing, and whether it will raise its full-year U.S. dollar revenue growth and capital spending outlook.

Blockchain infrastructure company Startale has launched the Startale Card, a new payment card that lets users spend on-chain assets at merchants that accept Visa. The product is backed by Soneium and is designed to let users keep earning yield while using their assets for payments. Startale also said each purchase made with the card will receive USDSC cashback. The company has opened a waitlist for users interested in the card. The announcement focuses on combining everyday card payments with on-chain asset utility, while keeping rewards tied to spending activity.

OpenAI, Meta and SpaceXAI are racing to build more efficient AI models as leading labs push to reduce training and inference costs, according to Techub, citing Cointelegraph. The move is putting pressure on rivals including Anthropic. The report says the AI sector is entering an efficiency race, with top labs working on architecture optimization and engineering improvements to lower compute spending while keeping model performance intact. That shift points to a new competitive focus inside the industry: efficient inference. Rather than centering only on larger models or higher raw capability, the latest push is aimed at improving how much performance companies can extract from the same or lower computing budget. In the report’s framing, cost control across training and inference is becoming a key battleground among major AI developers.

South Korean President Lee Jae-myung said the government will be guided to support three priority areas: semiconductors, artificial intelligence data centers, and physical AI. The update was published by Odaily and attributed to Jin10. No further policy details, funding figures, or timeline were disclosed in the brief. The item was categorized under technology trends.

South Korean President Lee Jae-myung said on July 13 that the country will set up a "Future Response Fund" and direct surplus tax revenue into future-focused investment projects, according to BlockBeats. He said government support will be channeled into three large-scale areas: semiconductors, AI data centers, and physical AI. The statement outlines where public backing is expected to be concentrated, with the fund positioned as a vehicle for allocating extra tax income to projects the government sees as effective future investments. No additional details on the fund’s size, timeline, or implementation were disclosed in the brief.

A field report by Korea Economic Daily on Taiwan argues that Taiwan Semiconductor Manufacturing Co. (TSMC) has grown beyond a chipmaker into a force reshaping income distribution, housing prices, family decisions and even university preferences. The report says TSMC’s market value stands at NT$63.53 trillion, or about 47% of Taiwan’s total stock market capitalization. It also cites average annual compensation for TSMC employees at NT$4.072 million last year, compared with NT$760,000 for Taiwan’s overall salaried workforce. Entry-level semiconductor engineers were described as earning NT$1 million to NT$2 million, above the NT$500,000 to NT$700,000 national average for university graduates. Korea Economic Daily links that pay gap to rising property prices around Hsinchu Science Park, where housing prices climbed 88.1% over the past five years. The story also describes the rise of new social labels such as “science-and-engineering men” and “Hsinchu tech moms,” as higher household income feeds spending on tutoring, premium consumption and overseas education. The shift is also reaching classrooms. According to the report, top students are increasingly asking about electrical engineering, electronics and semiconductor majors rather than medicine, as jobs at TSMC or MediaTek appear to offer faster income growth and asset accumulation.

NOXA, a launchpad platform on Robinhood Chain, recorded $1.94 million in protocol fees yesterday, according to data cited by BlockBeats from DefiLlama on July 13. That put it ahead of Pump.fun, which posted $1.61 million in protocol fees over the same period. BlockBeats also reported that NOXA serves as the CASHCAT launch platform. Separately, the platform’s cumulative number of active user addresses reached 267,642. The figures point to strong recent activity on NOXA relative to Pump.fun based on the fee data referenced in the report.