Polymarket generated more than $11 million in revenue this week, marking a new all-time high, according to data from DefiLlama cited by BlockBeats on July 12. The protocol’s cumulative revenue has also climbed past $97 million. The update points to a new peak in weekly earnings for the prediction market platform, based on the figures referenced in the report. No additional operating details were provided in the source brief.

OpenAI researcher Ethan Knight said on X on July 12 that GPT-5.6 Sol Ultra generated a proof paper and prompting setup for the cycle double cover conjecture, a well-known unsolved problem in graph theory. According to Knight, the model completed the result in about one hour, using up to 64 parallel sub-agents and without internet search. OpenAI’s public paper is three pages long. It first reduces the problem to cubic regular graphs, then uses the 8-flow theorem and GF(3) labels, and applies linear algebra to build a structure in which each edge is contained in exactly two cycles. The claim has not been published in an academic journal, has not gone through peer review, and has not been checked by formal verification systems such as Lean or Coq. Thomas Bloom, a mathematician at the University of Manchester in the UK, called the proof “very impressive” while also saying the references were not sufficient. The report was cited by ZDNet Korea Semiconductor and carried by Odaily.

Shenzhen-based Yunbao Smart has had its application for a ChiNext IPO accepted by the Shenzhen Stock Exchange, putting the chip startup on track to pursue the label of China’s first listed DPU company. Founded in August 2020 by Stanford PhD Xiao Qiyang, the company focuses on data processing units, a segment that gained attention after Nvidia introduced the DPU concept and Jensen Huang outlined the “CPU+DPU+GPU” architecture. According to the prospectus cited in the report, Yunbao Smart developed what it describes as China’s first high-performance, general-purpose programmable DPU SoC chip, with network bandwidth of 400 Gbps, four times the performance of traditional solutions, and power consumption cut by more than 50%. The company said it moved from FPGA verification to 6nm SoC mass production in four years. Financially, revenue rose from RMB 170,000 in 2023 to RMB 36.35 million in 2024 and RMB 370 million in 2025, while net losses were RMB 667 million, RMB 600 million, and RMB 1.19 billion over the same period. Tencent, which first invested in the angel round and kept backing later financings, held 19.7792% before the IPO through affiliated entities, making it Yunbao Smart’s largest shareholder. The report said the company’s valuation exceeded RMB 14 billion after a funding round in late November 2025.

Citrini analyst Jukan said on X on July 12 that Nvidia has decided, at least for now, not to adopt Taiwan Semiconductor Manufacturing Co.'s COUPE, or Compact Universal Photonic Engine, technology route. He attributed the move to slow progress in TSMC's silicon nitride, or SiN, development and weaker-than-expected work on a 2D grating coupler. According to Jukan, Nvidia's earlier "Plan A" had been built around TSMC's co-packaged optics platform based on COUPE. The latest "Plan B," he said, shifts to a silicon photonics platform from Tower Semiconductor. Jukan had also previously said that the release timing for Nvidia's next-generation Rubin architecture products could be pushed back again. The post did not include any additional timeline or product details beyond those points.

The Kobeissi Letter said on July 12 that the U.S. economy is becoming increasingly dependent on AI investment, with AI-related spending now making up more than 25% of U.S. GDP growth, a record high. The figures cited cover software, IT equipment, research and development, and data center construction. In practical terms, the note said that for every $4 of economic growth in the United States, more than $1 is now being driven by AI investment. It also said AI-related spending has risen to about 8% of U.S. GDP, also a historic high. For comparison, during the 2000 dot-com bubble, spending on IT equipment, software, and research and development peaked at roughly 6.5% of GDP. The Kobeissi Letter said current U.S. economic growth is increasingly centered on AI.

Iran said future transit management arrangements for the Strait of Hormuz should be worked out jointly by Iran and Oman, following talks between the two countries’ foreign ministers in Muscat on July 11. Iranian Foreign Ministry spokesperson Baghaei said on July 12 that the meeting focused on navigation in the strait and a coordination mechanism for passage management, with legal and technical delegations from both sides taking part. According to Baghaei, the discussions covered maritime safety in the Strait of Hormuz, respect for the sovereign rights of both countries, and the application of international law. The two sides also reviewed matters related to Article 5 of the Islamabad Memorandum of Understanding. Iran said any future arrangement should take into account developments over recent months, especially U.S. and Israeli military actions against Iran and their impact on shipping security in the waterway. Baghaei added that Iran and Oman agreed to continue consultations at the political, legal, and technical levels in an effort to reach a common understanding on shipping security in the Strait of Hormuz. A Qatari government delegation joined part of the discussions.

Anthropic has published a detailed explanation of a recurring source of confusion in Claude Code: many users treated model selection as the main fix when coding performance dropped, while the company says the bigger factor was often the Effort setting. The post revisits the backlash from March, when developers said Claude Code had become less reliable, skipped files, avoided tests, and stopped tasks midway. According to Anthropic, the issue followed a March 4 change that lowered the default Effort level from high to medium to reduce latency. The default was restored on April 7, and subscribers received a usage reset. In Anthropic’s framing, Model determines what Claude can know and do because it swaps in a different set of frozen weights, while Effort controls how much work Claude puts into a task, such as reading files, validating outputs, running tests, and pushing through multi-step jobs. The company also says users should first check prompts, tools, and CLAUDE.md context before changing any settings, then decide whether a failure reflects a capability gap or insufficient effort. Anthropic argues that this distinction matters in practice: a smaller model on high Effort can outperform a larger model on low Effort in many coding tasks. The company also linked this idea to its newer ultracode option, which can invoke xhigh effort and, for substantial tasks, let Claude decide whether to use a multi-agent workflow.

Iranian Parliament Speaker Ghalibaf said today that the era of unilateral agreements is over, according to Odaily. He said Iran had previously asked relevant parties to fulfill their commitments and obligations, warning that otherwise they would pay a price, and added that it is now time to face reality. Ghalibaf also attached part of Article 5 from an earlier set of 14 memorandums of understanding. The clause stressed the restoration of navigation through the Strait of Hormuz, provided that arrangements set by the Iranian side are observed. The statement centered on Iran’s position toward prior commitments and the conditions tied to reopening passage through the waterway.