# Anthropic İlgili Makaleler

HTX Haber Merkezi, kripto endüstrisindeki piyasa trendleri, proje güncellemeleri, teknoloji gelişmeleri ve düzenleyici politikaları kapsayan "Anthropic" hakkında en son makaleleri ve derinlemesine analizleri sunmaktadır.

Anthropic Has Taught Models to Understand Morality and Opened a New Path for Distillation

Anthropic's research "Teaching Claude Why" reveals a new, data-efficient method for AI alignment. Instead of relying on massive reinforcement learning with punishment (RLHF), which only teaches models to mimic safe answers without true ethical understanding, they used a small dataset (3 million tokens) of "difficult advice." This data consisted of detailed moral deliberations, reasoning, and debates, teaching the model the *why* behind decisions. The key was "deliberation-enhanced" Supervised Fine-Tuning (SFT). The model was trained on responses that included a "chain of thought" (CoT) process based on a constitutional framework. This framework included top-level principles, practical heuristics (like the "1000-user test"), and an 8-factor utility calculator (evaluating harm probability, reversibility, consent, etc.) for weighing complex trade-offs. This approach dropped model misalignment rates from 22% to 3% and showed strong generalization to unseen scenarios. The success challenges the old belief that "SFT memorizes, RL generalizes." It shows that SFT can generalize powerfully if the training data has two features: 1) high prompt diversity (many different scenario types) and 2) CoT supervision (showing the reasoning steps, not just the final answer). The model learns the underlying *thinking framework*, not just surface-level behaviors. This method points to a new paradigm for training AI in "non-RLVR" domains—areas like ethics, creative writing, or strategy where there's no single verifiable answer. The formula is: Domain Constitution + Heuristics + Multi-Factor Deliberation Framework + Diverse Deliberative CoT Data = Generalized capability. It represents a new form of "distillation," moving competition from pure compute towards who can best structure expert knowledge into high-quality reasoning datasets.

marsbitDün 10:55

Anthropic Has Taught Models to Understand Morality and Opened a New Path for Distillation

marsbitDün 10:55

Claude's New Policy Abandons Its Most Loyal Agent Users

Anthropic, in a move signaling the end of the "all-you-can-eat" era for AI subscriptions, has separated programmatic usage from its Claude subscription plans. Starting June 15, 2024, usage of the Claude Agent SDK, `claude -p` command, and third-party tools like OpenClaw will no longer draw from subscription limits. Instead, users receive a fixed monthly credit based on retail API prices: $20 for Pro, $100 for Max 5x, and $200 for Max 20x. This change drastically reduces usable capacity for heavy users—previously, their shared subscription limit was worth an estimated $2,000-$5,000 in API value. While Anthropic simultaneously increased Claude Code interactive limits to appease users, the new policy primarily impacts developers running automated, high-frequency agents, pushing their effective costs nearly ten times higher. Seizing the opportunity, OpenAI promptly announced a free two-month migration plan for its Codex enterprise service, which does not differentiate between interactive and automated usage, directly targeting discontented Claude users. This marks an opening salvo in the broader ASI (Artificial Superintelligence) competition, where the final battle is shifting from pure model capability to ecosystem strength, developer loyalty, and infrastructure. The article frames this as a necessary correction of a pricing "loophole" by Anthropic ahead of its IPO, as programmatic calls lack training data value and can incur massive costs. The move underscores a wider industry trend towards consumption-based billing for AI, mirroring the evolution of cloud computing.

marsbitDün 00:22

Claude's New Policy Abandons Its Most Loyal Agent Users

marsbitDün 00:22

Bezos, Schmidt, Powell Jobs: The Three AI Investment Philosophies of Silicon Valley's Old Money

Jeff Bezos, Eric Schmidt, and Laurene Powell Jobs, three prominent figures from Silicon Valley's "old money," are deploying massive personal fortunes into AI, but with distinctly different investment philosophies reflecting their visions for the future. Eric Schmidt, the former Google CEO, approaches AI as a geopolitical and infrastructural arms race. Through his family office, Hillspire, he invests heavily in defense AI companies, energy infrastructure (like Bolt Data & Energy to power data centers), and space launch capabilities (Relativity Space). For Schmidt, the ultimate AI advantage lies in physical resources—energy, transport, and military application—framing it as a national competition requiring state-level strategy and endurance. Jeff Bezos is building a vertically integrated, full-stack AI empire. His bets span the model layer (via Amazon's massive investment in Anthropic), the application layer (through investments like Perplexity), and now, the physical execution layer. His new venture, Project Prometheus, with $6.2 billion, aims to inject AI into manufacturing, creating a closed loop from AI chips and cloud compute (AWS) to real-world production, potentially for Amazon's own ventures like the Kuiper satellite network. In contrast, Laurene Powell Jobs adopts a more subtle, human-centric approach through her Emerson Collective. Her AI investments focus on specific, positive-impact applications—such as AI for healthcare (Proximie, Atropos Health), education (Curipod), and European AI sovereignty (Mistral AI). A key, high-profile bet was her early backing of Jony Ive's design firm LoveFrom and its spin-off, io, an AI hardware device company later acquired by OpenAI. Her philosophy prioritizes improving human-machine interaction and addressing societal needs over sheer scale or control. These three strategies—Schmidt's focus on state-level infrastructure and security, Bezos's pursuit of end-to-end industrial integration, and Powell Jobs's emphasis on human-centered design and applied solutions—represent fundamentally different wagers on what will define the next decade of AI. While the eventual winner is unknown, the sheer scale of this capital migration from internet-era giants is already reshaping the industry's trajectory.

marsbit2 gün önce 08:11

Bezos, Schmidt, Powell Jobs: The Three AI Investment Philosophies of Silicon Valley's Old Money

marsbit2 gün önce 08:11

Altman is Anxious: OpenAI's Three-Year Reign Has Just Been Overtaken by Anthropic

In May 2026, Anthropic has historically overtaken OpenAI in workplace adoption, according to Ramp's AI Index based on real business spending data. Anthropic's enterprise adoption rate reached 34.4%, surpassing OpenAI's 32.3%. Over the past year, Anthropic's rate grew nearly fourfold, while OpenAI's grew only 0.3%. Key to Anthropic's surge is its shift to a usage-based, per-token billing model and the integration of AI Agents into core business workflows, driving massive token consumption and revenue growth—reportedly reaching an annualized ~$44 billion. Despite concerns over rising costs and service stability, enterprises exhibit heavy reliance on Claude for productivity gains. OpenAI has responded with defensive measures, including offering free enterprise trials. The competition has shifted from model benchmarks to deep workflow integration, marking a new phase in the enterprise AI battle.

marsbit2 gün önce 07:23

Altman is Anxious: OpenAI's Three-Year Reign Has Just Been Overtaken by Anthropic

marsbit2 gün önce 07:23

Suzerain State: Anthropic

Anthropic, a five-year-old AI lab dubbed a "suzerain," has rapidly gained unprecedented influence by securing massive financial and computational commitments from tech giants, positioning itself at the center of AI infrastructure power dynamics. In May 2026, it announced securing over 300 MW of computing power from SpaceX's Colossus 1 data center, on top of earlier multi-billion dollar deals with Amazon and Google, effectively locking in over 20 GW of future compute. These investments are tied to reciprocal spending commitments on the investors' cloud platforms, resembling infrastructure pre-sales. This "suzerain" status is fueled by explosive growth. By May 2026, Anthropic's annualized revenue reportedly surged to over $44 billion, with Claude surpassing OpenAI in LLM market share. Its high-revenue-per-user efficiency and flagship product Claude Code have secured a strong enterprise foothold. However, its pre-IPO status faces scrutiny. OpenAI challenged Anthropic's accounting, alleging its reported revenue includes gross payments shared with cloud partners, unlike OpenAI's net revenue reporting. The resolution of this debate is critical as both companies approach public listings. Currently, Anthropic holds unique leverage as the only top-tier model available across AWS, Google Cloud, and Microsoft Azure, inverting traditional vendor-customer dynamics. Yet, its suzerainty is considered a time-limited game, dependent on converting its current advantages into sustainable, audited profitability and navigating the complex web of strategic dependencies with its powerful patrons.

marsbit2 gün önce 00:41

Suzerain State: Anthropic

marsbit2 gün önce 00:41

AI Bull Market Reprices Everything, Including the 'Male Valuation System' in the Marriage Market

The AI boom is redefining value across markets, including the male "valuation system" in the dating scene. A new hierarchy is emerging, based on company valuation, employee income, and industry status within the AI sector. At the top are NVIDIA and SK Hynix employees, dubbed the "T0 version." NVIDIA is the AI world's cash machine, while SK Hynix employees are seeing astronomical bonuses due to HBM demand, making them highly sought-after "AI concept stocks" in Korea's dating market. Next are OpenAI and Anthropic staff, representing the "new elite." Unlike the paper wealth of the past internet boom, these employees are actively realizing significant wealth through stock sales, though their status is considered more volatile. DeepSeek and ByteDance AI team members are rated as "top-tier." Their companies are engaged in fierce talent wars with massive investments, making these employees scarce, high-value players. Samsung and Tencent employees are seen as "NPCs" still searching for their AI "ticket." Samsung has been outpaced by SK Hynix in the memory race, while Tencent's more cautious AI investment contrasts with ByteDance's aggressive strategy, raising questions about their future position. Finally, traditional finance and crypto men are rated at the bottom ("pulled"). Their once-dominant wealth and status are being eclipsed by the new AI-driven economic order and its redistribution of value and opportunity.

Odaily星球日报05/13 12:53

AI Bull Market Reprices Everything, Including the 'Male Valuation System' in the Marriage Market

Odaily星球日报05/13 12:53

Anthropic Refuses Access to China's Think Tank for Most Powerful AI Model Mythos, Intensifying US-China AI Competition

Anthropic rejects Chinese think tank request for access to its most advanced AI model, Claude Mythos, escalating US-China AI competition. The request was made informally at a Carnegie Endowment meeting in Singapore, prompting concern within the U.S. National Security Council. Mythos, released in April 2026 under the "Project Glasswing" initiative, is described as "digital weapon-grade" technology for its unprecedented ability to discover zero-day vulnerabilities. Access is limited to about 40 US and UK entities in cybersecurity, finance, and tech, with China explicitly excluded as an "adversarial nation." While the official Chinese response has been muted, the incident highlights China's exclusion from cutting-edge AI defense tools despite its critical infrastructure running on software Mythos can analyze. The Chinese cybersecurity market anticipates massive growth and domestic model development, but a significant capability gap remains. The event coincides with internal Trump administration debates on pre-release AI model security assessments and an upcoming presidential visit to China, where AI dialogue is expected. Analysts caution that past US-China AI safety talks have seen China more focused on information gathering than substantive collaboration.

marsbit05/13 06:29

Anthropic Refuses Access to China's Think Tank for Most Powerful AI Model Mythos, Intensifying US-China AI Competition

marsbit05/13 06:29

How the $900 Billion Anthropic Was Built?

Anthropic, the AI startup behind Claude, is reportedly in early talks to raise at least $30 billion in new funding, targeting a valuation exceeding $900 billion. This would propel it past OpenAI's recent $852 billion valuation. The funding round is expected to close by late May 2026. The company's valuation surge is driven by extraordinary revenue growth, reportedly reaching an annualized $30 billion by March 2026 from $1 billion in December 2024. However, OpenAI questions this figure, suggesting a net revenue closer to $22 billion after cloud platform fees. Despite high revenue, Anthropic's gross margin is reportedly around 40%, and it is not yet profitable, with breakeven projected for 2028. A significant portion of the new capital would fund massive, pre-committed computing infrastructure with partners like Amazon, Google, and Microsoft. This highlights a new AI financing model where high valuations fuel compute spending, which in turn requires even higher future valuations to sustain. Notably, many early-stage investors are reportedly sitting out this round. Bankers privately estimate a potential IPO valuation between $400-500 billion, creating a rare scenario where the final private funding round valuation ($900B+) could far exceed the expected public market debut. Anthropic is targeting an IPO between October 2026 and the first half of 2027. Its public listing is poised to be a critical test for the entire AI sector's valuation logic, potentially validating or challenging the high-stakes "valuation-compute-valuation" cycle that has defined private market investments.

链捕手05/13 02:42

How the $900 Billion Anthropic Was Built?

链捕手05/13 02:42

600 People, $66 Billion: The First Major Cash-Out in the Era of Large Models

The first systematic "big cash-out" of the AI era occurred in October 2025, when over 600 current and former OpenAI employees sold a total of $6.6 billion in shares via a secondary market. Approximately 75 individuals maxed out a $30 million per-person sale limit, while around 525 others cashed out an average of $8.3 million each. This event, exceeding the scale of any 2024 US IPO, functioned as a "shadow IPO." It marked a radical departure from the traditional Silicon Valley path of waiting for a public listing, instead allowing employees to convert equity to cash after just two years of tenure—a direct retention tool in a fiercely competitive talent market where rivals like Meta have offered packages worth hundreds of millions. This massive liquidity event presents a dual-edged sword for OpenAI. While it helps retain talent, it also risks triggering a brain drain as newly wealthy employees may depart. Furthermore, it creates a dilemma for those who sold: they forfeited potential future gains as the company's valuation soared from $400 billion to $852 billion within months. In stark contrast, employees at rival Anthropic demonstrated greater reluctance to sell during their own secondary offering. The financial narratives of the two labs also diverge sharply. OpenAI, while achieving over $20 billion in annualized revenue by 2025, faces massive projected losses (up to $14 billion in 2026), a long path to cash flow positivity, and significant revenue-sharing payments to Microsoft. Anthropic reports rapid revenue growth, improving gross margins, and a faster path to profitability. OpenAI's trajectory is thus balanced precariously between skyrocketing valuation based on funding narratives and the pressures of sustained financial losses post-cash-out. The event underscores that the AI race has evolved into a capital and human experiment, where immense wealth crystallizes the complex calculations of greed, fear, and ambition within the industry.

marsbit05/12 07:46

600 People, $66 Billion: The First Major Cash-Out in the Era of Large Models

marsbit05/12 07:46

Anthropic and OpenAI Have Single-Handedly Severed the Logic of Pre-IPO Stock Tokenization

The pre-IPO stock token market is experiencing significant turmoil following strong statements from AI giants Anthropic and OpenAI. Both companies have updated their official policies, declaring that any transfer of their company shares—including sales, transfers, or assignments of share interests—without prior board approval is "invalid" and will not be recognized in their corporate records. This means buyers in such unauthorized transactions would not be recognized as shareholders and would have no shareholder rights. A major point of contention is the use of Special Purpose Vehicles (SPVs), which are legal entities commonly used by pre-IPO token platforms to pool investor funds and indirectly acquire shares from employees or early investors. The companies explicitly state they do not permit SPVs to acquire their shares, and any such transfer violates their restrictions. They warn that third parties selling shares through SPVs, direct sales, forward contracts, or stock tokens are likely engaged in fraud or are offering worthless investments due to these transfer limits. This stance directly threatens the core model of many pre-IPO token platforms, which rely on SPV structures. The announcement revealed additional risks within this model, such as complex "SPV-within-SPV" layering that obscures legal transparency, increases management fees, and creates a chain reaction risk of invalidation. Following the news, tokens like ANTHROPIC and OPENAI on platforms like PreStocks fell sharply (over 20%). The market reaction highlights a divergence: while asset-backed pre-IPO tokens plummeted, purely speculative pre-IPO futures contracts, which are bilateral bets on future IPO prices with no claim to actual shares, remained relatively stable as they are unaffected by the transfer restrictions. The industry is split on the implications. Some believe the fundamental logic of pre-IPO token trading is broken if leading companies reject SPV-held shares, potentially causing a domino effect. Others, like Rivet founder Nick Abouzeid, argue that buyers of such unofficial tokens always knowingly accepted the risk of non-recognition by the company. The statements serve as a stark risk warning and a corrective measure for a market where valuations for some AI-related pre-IPO tokens had soared to irrational levels, far exceeding recent funding round valuations.

marsbit05/12 05:04

Anthropic and OpenAI Have Single-Handedly Severed the Logic of Pre-IPO Stock Tokenization

marsbit05/12 05:04

活动图片

Market Analysis

Bitcoin