# Anthropic Related Articles

HTX News Center provides the latest articles and in-depth analysis on "Anthropic", covering market trends, project updates, tech developments, and regulatory policies in the crypto industry.

AI Billing Black Box Exposed: 1.7 Million Overcharged, Anthropic Refunds But Doesn’t Admit Fault

A startup named Vaudit, founded by former Oracle director Michael Hahn, audits AI bills for companies and claims to have identified approximately $1.7 million in overcharges across 60 businesses, totaling $34 million in reviewed bills. The alleged discrepancies primarily involve charges for Anthropic's Claude Code. Common issues cited include billing for newer, more expensive models when older, cheaper ones were used; charging for failed or errored requests; and "retry storms" where AI agents silently retry failed tasks, accumulating costs unnoticed. Major clients like Panasonic, HP, and Honda were among those audited. While Vaudit reports that around 80% of the disputed charges were refunded by providers like Amazon, Google, Microsoft, Anthropic, and OpenAI after申诉, the AI companies largely deny systemic problems. Anthropic stated overcharges do not appear widespread and it does not bill for uncompleted requests or errors, while OpenAI said it found no evidence of such issues affecting its customers. The situation highlights the inherent opacity and complexity of AI billing, which is based on token usage that is difficult to track and predict, especially with multi-agent, multi-model workflows. This complexity is creating a new market for third-party AI bill auditing services like Vaudit, which charges fees based on recovered amounts. Separately, Anthropic faces a proposed class-action lawsuit alleging its high-tier subscription plans deliver far less usage than advertised. The case underscores growing scrutiny over AI service pricing and transparency as major providers prepare for IPOs.

marsbit06/29 09:34

AI Billing Black Box Exposed: 1.7 Million Overcharged, Anthropic Refunds But Doesn’t Admit Fault

marsbit06/29 09:34

GPT-5.6 Suffers 'Brain Surgery', Fable 5's Return to Be a Weaker Version?

GPT-5.6 and Fable 5 Face "Lobotomization" and Restrictions OpenAI has launched its new GPT-5.6 model in three limited tiers: Sol (advanced), Terra (balanced), and Luna (fast/economical). However, access to the most powerful Sol version is severely restricted, reportedly due to U.S. government intervention over safety concerns linked to its capabilities. Similarly, Anthropic's Fable 5 was globally disabled for 72 hours after a demonstration revealed its potential to exploit security vulnerabilities. It is now set to return in a heavily restricted "neutered" version for select partners only. Industry experts warn that such stringent safety "guardrails" could lead to "lobotomized" models with diminished intelligence and creativity. Developers are concerned about restricted access, potential extra costs, and models that no longer match their original promised capabilities. This marks a shift towards controlled access, contrasting with the earlier era of widely available, powerful AI.

marsbit06/29 08:37

GPT-5.6 Suffers 'Brain Surgery', Fable 5's Return to Be a Weaker Version?

marsbit06/29 08:37

Anthropic's Latest Report Reveals Global Workers' Patterns: Seeking Sleep at 5 AM, Asking for Recipes at 6 PM

A new report from Anthropic analyzes millions of hourly user interactions with Claude AI, revealing detailed patterns in daily life and work. The data shows distinct rhythms: people most frequently ask about sleep help around 5 AM, seek news at 7 AM, and search for dinner recipes at 6 PM—the day's single largest query spike. Usage sharply diverges between weekdays and weekends. Workdays are dominated by professional tasks like business emails and coding (backend, APIs). Weekends see a surge in personal use—nearly 50% of conversations—focused on emotional support, creative writing (especially fan fiction), medical advice, and side projects like AI agent design or game development. Weekend "entrepreneurial" queries peak globally, while job-hunting activity drops. The report introduces "artifact" analysis, finding 93% of conversations produce a tangible output (explanation, document, code, etc.). Blog posts are 81% work-related, while creative writing is over 80% personal. High-wage professionals (e.g., marketing managers, programmers) use Claude more intensively outside work hours, with longer conversations, more tokens consumed, and greater use of deep thinking features compared to lower-wage roles. Interestingly, Claude's responses typically register at a higher reading level than user prompts (by about one educational year on average), except for audience-focused writing like emails or blogs where the gap nearly disappears. The data also captures specific cultural moments, like an 8x spike in tax-related queries on the U.S. filing deadline. Precise hourly data transforms fragmented queries into a collective diary of modern life—mapping not just economic activity, but also cycles of anxiety, creativity, and daily rhythm, with AI acting as both a productivity tool and an intimate, always-available confidant.

marsbit06/29 08:20

Anthropic's Latest Report Reveals Global Workers' Patterns: Seeking Sleep at 5 AM, Asking for Recipes at 6 PM

marsbit06/29 08:20

The Eight Fathers of Transformer: Where Are They Now?

The founding authors of the seminal 2017 paper "Attention Is All You Need," which introduced the Transformer architecture, have all departed from Google and are now pioneering diverse ventures across the AI landscape. Their current paths highlight the field's evolution and fragmentation. Ashish Vaswani co-founded Essential AI after a stint at Adept, but recent reports suggest he and his team are being recruited by Nvidia. Noam Shazeer left Google twice, first to found Character.AI—later acquired by Google—before recently joining OpenAI. Niki Parmar, after co-founding both Adept and Essential AI with Vaswani, is now at Anthropic. Jakob Uszkoreit is the CEO of Inceptive, applying AI to RNA and drug design. Llion Jones co-founded Sakana AI in Tokyo, exploring collective intelligence from smaller models. Aidan N. Gomez is the CEO of Cohere, focusing on enterprise AI solutions. Łukasz Kaiser remains in research at OpenAI, contributing to models like GPT-4 and o1. Illia Polosukhin co-founded the NEAR Protocol blockchain, positioning it for an AI-agent economy. Despite their divergent paths—spanning biotech, enterprise software, blockchain, and foundational research—a shared sentiment persists: the Transformer is not the final answer. Each founder continues to seek the next significant architectural breakthrough, driven by the same innovative spirit that created the modern AI era.

marsbit06/28 05:30

The Eight Fathers of Transformer: Where Are They Now?

marsbit06/28 05:30

Anthropic Employees Leave to Start a Business, Valued at $1 Billion, Also Focused on 'Recursive Self-Improvement'

Former Anthropic employees Behnam Neyshabur and Harsh Mehta have founded a startup named Mirendil, securing $200 million in seed funding at a $1 billion valuation from investors including Andreessen Horowitz, Kleiner Perkins, and NVIDIA. The company, staffed by researchers from Anthropic, xAI, Google DeepMind, and OpenAI, aims to accelerate scientific discovery by focusing on "recursive self-improvement" for AI. Their vision, described as "AI for AI for Science," is to build a platform enabling research labs in fields like medicine and materials science to develop and iteratively improve their own specialized AI models, rather than relying on general-purpose models from large tech firms. Neyshabur argues this approach, involving AI that can optimize its own code under human supervision, is the fastest path to AI-accelerated science. The founders believe a market opportunity exists because major AI companies, while increasingly using AI to advance their own research, restrict external developers from using their models to train competing products. Mirendil plans to release its first models soon to gather early user feedback, with the goal of empowering thousands of labs to tackle critical scientific challenges.

marsbit06/26 03:26

Anthropic Employees Leave to Start a Business, Valued at $1 Billion, Also Focused on 'Recursive Self-Improvement'

marsbit06/26 03:26

Fable 5 Begins Gradual Rollout? Countdown to the June 26th Deadline

Exciting news has emerged in the AI community: Fable 5, which was previously disabled, appears to be undergoing a partial, gradual rollout. Users are reporting sightings of the model in the Claude mobile app's model selector and through specific commands, accompanied by messages about weekly usage limits. Code strings in recent Claude updates hint that Fable 5 might be permanently integrated into existing subscriptions as a weekly quota, rather than a separate purchase. The rollout seems inconsistent—available only in specific chat threads for some users, with indications that previous geographic restrictions may be loosening. A brief appearance on AWS, requiring U.S. ID verification and use-case approval, was later clarified as a UI bug. Simultaneously, rumors are circulating about the imminent launch of Claude Sonnet 5 (codenamed Fennec), which is speculated to offer performance nearing that of Opus 4.8. This would represent a significant generational leap. Anthropic has also quietly removed usage limits for the current Claude Sonnet model, fueling speculation about an upcoming major release. This flurry of activity comes just before a critical June 26th deadline. U.S. lawmakers have demanded that the Department of Commerce clarify its standards and timeline for potentially modifying export controls on such AI models. The recent testing and leaks are seen as high-stakes maneuvers ahead of this regulatory decision. The AI world is now watching closely to see if a wider reinstatement of advanced models will follow.

marsbit06/26 02:36

Fable 5 Begins Gradual Rollout? Countdown to the June 26th Deadline

marsbit06/26 02:36

Fable 5 is about to make a comeback, code exposed? Anthropic CEO kicked out of the White House

Fable 5, a previously restricted AI model from Anthropic, appears poised for a comeback. Evidence from leaked code in the Claude Code v2.1.190 version suggests a shift in its business model from a separate purchase to a potentially limited weekly usage allowance within standard Claude subscriptions. Furthermore, the model has reportedly reappeared in Amazon Bedrock documentation. This potential revival coincides with significant internal changes at Anthropic. According to a report by The Wired, CEO Dario Amodei was reportedly sidelined from negotiations with the Trump administration over Fable 5's export restrictions. Government officials found him difficult to communicate with. Co-founder Tom Brown and policy head Sarah Heck took over discussions, leading to more productive technical talks aimed at addressing White House security concerns about the model being "jailbroken." External pressure is mounting as a bipartisan group of US lawmakers has demanded answers from the Commerce Department by a June 26 deadline regarding the criteria and timeline for potentially reinstating public access to Fable 5. The potential return of Fable 5 comes as competitors OpenAI and Google have reportedly delayed their own major model releases. If Anthropic successfully navigates the government's security review, Fable 5 could gain a significant "safety-certified" advantage in the enterprise market. The countdown to the June 26 deadline is now underway.

marsbit06/25 07:29

Fable 5 is about to make a comeback, code exposed? Anthropic CEO kicked out of the White House

marsbit06/25 07:29

Claude's Capabilities Extracted at Large Scale? Anthropic Accuses Ali-Related Parties of 'Distilling' Models

Anthropic alleges that entities associated with Alibaba and its AI lab Qwen used nearly 25,000 fraudulent accounts to extract capabilities from its Claude AI model in what it calls the "largest known" model distillation attack. The alleged incident occurred between April 22 and June 5, involving over 28.8 million interactions with Claude. Model distillation involves using a powerful model's outputs to train another model, potentially replicating abilities like software engineering and agent reasoning without stealing the underlying code. The accusations emerge amid heightened US AI export controls and the Pentagon's listing of Alibaba as a "Chinese military company." Anthropic detailed the claims in a June 10 letter to the US Senate Banking Committee, urging better threat intelligence sharing. While not conclusively proving direct Alibaba involvement or successful capability replication, the case highlights growing concerns over AI model outputs as contested assets. The incident may push for stricter controls on model access and user verification, increasing compliance costs for AI firms and potentially limiting Chinese companies' access to advanced foreign models.

marsbit06/25 06:29

Claude's Capabilities Extracted at Large Scale? Anthropic Accuses Ali-Related Parties of 'Distilling' Models

marsbit06/25 06:29

Giants Wage the Context War, Reconstructing AI Moats

The article "Giants Launch the Context War, Reconstructing AI's Moat" discusses how leading AI companies—OpenAI, Anthropic, and Google—are shifting their competitive focus from model size to acquiring, managing, and utilizing user context (Context). Initially, Context referred to the length of text a model could process, leading to a "arms race" for longer context windows. However, the competition has evolved through three key phases: expanding text capacity (long context windows), enabling memory across sessions, and finally, integrating AI into real user environments like browsers and desktops to capture dynamic task states. Each company is pursuing a distinct strategy. OpenAI is building Context around the ChatGPT account, turning it into a central hub that accumulates user understanding across various integrated applications and tools. Anthropic, lacking a major user base, focuses on high-value verticals like coding, empowering its Claude model to actively gather Context through GUI interaction (Computer Use) and system connections (MCP protocol). Google, with vast existing user data from products like Search and Gmail, faces the challenge of restructuring this data into actionable, AI-understandable Context for its Gemini model within its ecosystem. The core argument is that the nature of competitive advantage in AI is changing. The internet era prized network effects—connecting more users. The AI era values "individual depth": the ability to build deep, task-specific understanding of a user. This creates a new moat through 1) the compounding value of accumulated Context, 2) deep integration with user tools and permissions, and 3) the establishment of trust for complex tasks. Therefore, the battle for Context is fundamentally about capturing "task entry points" and converting existing digital ecosystems into environments where AI can effectively understand and act, rather than merely scaling user numbers.

marsbit06/23 23:13

Giants Wage the Context War, Reconstructing AI Moats

marsbit06/23 23:13

OpenAI Partners with PE Firms, Investing $4 Billion. Let's Talk About Silicon Valley's Hottest New Role: FDE.

The hottest new role in Silicon Valley is the Forward Deployment Engineer (FDE), a hybrid of engineer and business consultant whose core mission is to transform AI demos into native, practical workflows within client organizations. The recent surge in demand is driven by a strategic shift from leading AI companies. OpenAI, partnering with 19 private equity firms in a $4 billion investment, formed a Deployment Company and acquired Tomoro along with its 150 FDEs. Anthropic also announced a $1.5 billion joint venture with financial institutions like Blackstone. The article, based on interviews with industry experts Jove (FDE lead at Cresta) and Oliver (VP at Invisible Technologies, ex-McKinsey), explores the FDE role and the rise of deployment-focused companies. Key insights include: **The FDE Role:** Jove describes an FDE as a "Forward Deployed CTO"—a technically strong engineer who works intimately with clients to implement AI solutions, learn from the process, and feed those insights back to improve the core product. They require expertise in AI agents, client-facing experience, resilience, and the ability to handle complex, imperfect systems. While AI tools enhance their efficiency, the role's complexity makes full automation a distant prospect. **Industry Shift:** Model companies are moving beyond selling tools to ensuring real-world adoption. This blurs the line between model and application companies. Collaborations with private equity (PE) firms are key, providing access to large portfolios of traditional businesses needing AI transformation. For PE firms, these partnerships offer signal value to LPs, create tangible value in portfolio companies, and provide exposure to high-growth AI assets. **Consulting & Transformation:** AI deployment involves deep, customized workflow redesign, moving beyond simple tool augmentation. Companies like Invisible Technologies build modular platforms to create bespoke, AI-native workflows for clients. While traditional consulting will see growth in helping businesses rethink their models for AI, the real value is captured by firms that leave behind transformed, operational systems. Critical success factors include building robust data foundations and strategically deciding which workflow steps should be deterministic versus AI-driven. The ultimate goal shifts from pure cost-cutting to unlocking new revenue opportunities previously impossible without AI-scale capabilities.

marsbit06/23 07:28

OpenAI Partners with PE Firms, Investing $4 Billion. Let's Talk About Silicon Valley's Hottest New Role: FDE.

marsbit06/23 07:28

活动图片

Hot Categoriesright-arrow

Hot Tagsright-arrow