ARK Invest Weekly Report: AI Agents Will Continue to Drive Anthropic and OpenAI Development, Grok 4.6 May Lower Frontier AI Cost Curve BlockBeats News: On August 23rd, ARK Invest, a subsidiary of "Wood Sister" Cathie Wood, released its weekly market report covering three main themes: Artificial Intelligence Agents Will Continue to Drive the Development of Anthropic and OpenAI. Anthropic has submitted an S-1 draft to the SEC on June 1st to prepare for a potential IPO and is testing market sentiment with investors. As of the end of May, Anthropic's annualized recurring revenue (ARR) has reached $47 billion, starting from approximately $9 billion at the beginning of 2026 and growing more than fivefold in just five months; Data provider TickerTrends estimates that its current ARR may exceed $74 billion. OpenAI ARR is approximately $41 billion, which is about twice the $20 billion at the beginning of the year. The combined ARR of the two companies has exceeded $115 billion, surpassing the 12-month revenue from the merger of SAP, Salesforce, and Adobe, and approaching the annualized operating rate of approximately $150 billion for Microsoft's productivity and business processes division. Both companies plan to support large-scale computing power construction through public market financing. Grok 4.6 may further drive down the cost curve of cutting-edge AI: The Grok 4.6 released by SpaceXAI is designed specifically for programming, agent tasks, and knowledge work, with a key parameter of a 500000 token context window, an input of $2/million tokens, and an output of $6/million tokens, significantly lower than OpenAI GPT-5.6 Sol ($5/30) and Anthropic Claude Fable 5 ($10/50). Artificial Analysis evaluates its Intelligence Index score as 61, which is on par with GPT-5.6 Sol, only about 1 to 2 points behind Claude Opus 5 and Fable 5, but at a significantly lower price. In terms of task cost, Grok 4.6 is approximately $0.84 per task, placing it at the forefront of intelligent cost Pareto; AA Briefcase's Elo score for long-term intelligent agent knowledge work is 1577, which is basically on par with Claude Fable's score of 1574. SpaceXAI has also launched Grok Bot, which extends competition from model performance to the Agent software layer that can run tasks for long periods of time. After the improvement of model capabilities, cost measurement should shift from "price per token" to "price per completed task". If training and inference costs continue to decrease by 85% and 99.9% respectively each year, enterprises will deploy intelligent agents in more workflows, greatly accelerating AI adoption. The detection of minimal residual disease (MRD) proved its effectiveness and continued to scale up: the key evidence came from the bladder cancer trials IMvigor010 and IMvigor011. The median total survival period of ctDNA positive patients improved by 50% after using atezolizumab. Last quarter, Natera's Signatera sales reached 283000 times (an increase of 34000 times compared to the previous quarter), while all other solid tumor MRD tests totaled only 41000 times. Natera accounts for approximately 87% of the solid tumor MRD market share. Signatera is expected to achieve approximately $1.5 billion in revenue in its fifth year, which is about twice that of Cologuard during the same period. The market consensus opportunity is about $20 billion, and cancer survivor monitoring, late stage treatment monitoring, CAR-T, and global expansion may further expand the scale.