2026-08-05 Quarta

Notícias de cripto - Página 207

Mantenha-se a par do mercado de cripto. Notícias em tempo real, análises, preços, histórias em alta e análise de especialistas — tudo num só lugar.

$100M Annual Revenue, Two Berkeley Roommates in Their 20s Build the Most Profitable AI Business

Arena, the AI model ranking platform, has become a $100 million annual revenue business just eight months after launching its commercial service. Originally a UC Berkeley open-source research project called Chatbot Arena, it created a "battle arena" where users blind-test and vote on anonymous AI model responses. This has generated a highly trusted, community-driven leaderboard based on over 10 million user evaluations and 82 million votes. Major AI companies like OpenAI, Google, and Anthropic submit their flagship models to be ranked. The core monetization strategy is its AI Evaluations service, where model developers and large enterprises pay for in-depth performance analysis from Arena's massive user community. This provides real-world feedback on model strengths, weaknesses, and hallucinations—a critical service as models become more complex. The company, spun out from Berkeley in early 2025, quickly raised $100 million in seed funding at a $600 million valuation and later secured a $150 million Series A at a $1.7 billion valuation. The founding team includes CEO Anastasios Angelopoulos, a mathematician focused on rigorous model evaluation; CTO Wei-Lin Chiang, creator of the popular Vicuna chatbot; and co-founder Ion Stoica, a renowned Berkeley professor. Arena is now expanding beyond chat benchmarks into "Agent Mode," evaluating AI agents on complex, multi-step tasks like coding and research. The company's success illustrates the growing value and cost of independent, real-world AI model evaluation as the industry intensifies.

marsbit07/06 00:20

$100M Annual Revenue, Two Berkeley Roommates in Their 20s Build the Most Profitable AI Business

marsbit07/06 00:20

Racking Up 24,000 Stars: With One Command, AI Can Now Find Its Own Skills

Vercel, known for its developer tools like Next.js, has launched 'skills', a package manager for AI coding agents, garnering 24,000 GitHub stars. It allows developers to add specialized capabilities, such as React best practices, to AI assistants like Claude Code or Cursor with a single command: `npx skills add <package>`. Skills are shareable, reusable modules that define an AI agent's behavior for specific tasks, moving beyond one-off prompt engineering towards standardized 'capability engineering'. A key innovation is the 'find-skills' skill, which acts as an internal search engine, allowing an agent to autonomously find and install the right skill for a user's request. This lowers the barrier for non-developers to leverage advanced AI coding assistance. However, this 'npm moment' for AI brings significant security risks. Security audits of thousands of skills on platforms like skills.sh and ClawHub found over 30% contained security flaws, with about 13% classified as severe. Threats include malicious scripts that can access local files and credentials, and prompt injection hidden within skill documentation. Unlike traditional code packages, skills blend instructions, code, and system access, posing a direct risk to user machines and data. Experts advise treating skills like code—reviewing them carefully before installation, especially their scripts, and being wary of excessive permissions. Ultimately, Vercel's initiative represents a major shift towards modular, reusable AI capabilities, but its rapid adoption requires developers to bring the same caution used in managing traditional software dependencies.

marsbit07/06 00:19

Racking Up 24,000 Stars: With One Command, AI Can Now Find Its Own Skills

marsbit07/06 00:19

Token Devours One-Third of Payroll, Silicon Valley's AI Bill is Spinning Out of Control

The article discusses the dual reality of AI token costs in Silicon Valley. While research firm SemiAnalysis reports spending 30% of its employee salary budget on internal LLM tokens—translating to massive productivity gains like converting complex Excel models in minutes—other giants are struggling with ballooning, uncontrolled AI bills. Uber exhausted its annual AI budget in months after rapid engineer adoption, and Microsoft is cutting third-party AI tools due to high costs. NVIDIA's CEO argues tokens are becoming "means of production" and plans substantial AI budgets per engineer. Despite current cost concerns, the analysis emphasizes that cost collapse is just beginning. Through software optimizations (like 14x throughput boosts) and next-gen hardware (e.g., GB300 NVL72 with 17-32x H100 performance), real token costs can fall far below list prices. Anthropic's gross margins reportedly soared as token prices dropped. Gartner predicts a >90% inference cost drop by 2030. The piece highlights a split: massive AI capex ($740B announced) contrasts with tech layoffs and minimal measured economic impact so far. The transition mirrors past infrastructure shifts—investment precedes widespread productivity. For early adopters like SemiAnalysis, tokens already deliver high leverage; for others, the choice is to adopt now or risk falling behind.

marsbit07/06 00:15

Token Devours One-Third of Payroll, Silicon Valley's AI Bill is Spinning Out of Control

marsbit07/06 00:15

Claude Engineer Finally Unveils Fable 5's Ultimate Strategy, Teaching You How to Bridge the Information Gap with AI Models

This article, titled "Claude Engineer Finally Releases Fable 5 'Skill-Burning' Guide, Teaching How to Bridge the Information Gap with Models," details a blog post by Claude Code engineer Thariq Shihipar. The core concept is the "information gap" or "unknowns"—the disconnect between a user's instructions (the "map") and the actual task requirements (the "territory"). The article argues that with powerful models like Claude Fable 5, work quality depends on the user's ability to identify and clarify these unknowns. Shihipar categorizes unknowns into four types: Known Knowns (explicit instructions), Known Unknowns (awareness of gaps), Unknown Knowns (implicit, unstated knowledge), and Unknown Unknowns (unforeseen issues). The blog provides a framework for addressing these gaps throughout the workflow: * **Before Implementation:** Techniques include "Blindspot Scanning" to uncover Unknown Unknowns, brainstorming/prototyping for visual or complex tasks, having Claude ask clarifying questions, using reference code/examples, and creating implementation plans. * **During Implementation:** Maintaining an "implementation notes" file for Claude to document deviations and decisions made due to encountered edge cases. * **After Implementation:** Creating summary documents for review and having Claude generate quizzes to ensure the user fully understands the completed changes. The article concludes that as models become more capable, the key to success is systematically discovering and defining these unknowns through low-cost methods like prototyping and planning, allowing for more effective collaboration.

marsbit07/06 00:14

Claude Engineer Finally Unveils Fable 5's Ultimate Strategy, Teaching You How to Bridge the Information Gap with AI Models

marsbit07/06 00:14

GPT-5.5 Suffers Secret Intelligence Downgrade, Crashes at Thought Token 516, Stumbles Harder on Tougher Tasks

The article reports a concerning anomaly discovered in OpenAI's GPT-5.5 model. Developers analyzing backend metadata found a disproportionate number of its reasoning processes abruptly terminating precisely at 516 tokens, with similar spikes at multiples like 1034. Statistical analysis revealed that while GPT-5.5 handled only 19.3% of total responses, it was responsible for over 80% of these "exact 516" stoppages. This pattern is drastically more pronounced in GPT-5.5 compared to other models, suggesting a possible hidden "reasoning budget" cap or truncation mechanism. Concurrently, users report the model has become overly formulaic, persistently using bullet points, offering unsolicited corrections, and providing excessive options, making interactions feel less helpful. The core concern is that GPT-5.5 may be prioritizing structured, truncated responses over genuine, in-depth problem-solving, especially for complex tasks.

marsbit07/05 23:58

GPT-5.5 Suffers Secret Intelligence Downgrade, Crashes at Thought Token 516, Stumbles Harder on Tougher Tasks

marsbit07/05 23:58

OpenAI's Misfire, Scaling Law's Original Paper Reveals Bug, Trillions of Compute Power Wasted in Vain

Recent revelations by a former OpenAI researcher, Diogo Almeida, and subsequent discussion highlighted by DeepMind's Sander Dieleman suggest a critical bug in OpenAI's seminal 2020 "Scaling Laws" paper. The analysis claims the original research contained a flawed experimental setup, leading to a misinterpretation of how to optimally scale large language models (LLMs). The core issue involves two key methodological choices in the OpenAI paper: first, training all models (small and large) on the same fixed dataset size (~130 billion tokens), which underfed larger models; and second, using a cosine learning rate decay that prematurely flattened loss curves, creating the false impression that models had reached performance saturation with more data. This combination allegedly biased the conclusion that, for a fixed compute budget, scaling model parameters was vastly more important than scaling training data—a principle that drove the creation of "over-parameterized, under-trained" models like GPT-3. This was later corrected by DeepMind's 2022 Chinchilla paper, which advocated for a more balanced scaling of parameters and data. Further scrutiny revealed that even the Chinchilla analysis itself had an optimization bug. The critique extends beyond the bug, questioning whether current scaling laws are inherently biased, as they are primarily derived from English data, a morphologically poor language that may be inefficient to learn compared to others like French. The implication is that the AI industry may have wasted significant computational resources and years of effort following an erroneous scaling principle, potentially delaying more efficient model development.

marsbit07/05 23:58

OpenAI's Misfire, Scaling Law's Original Paper Reveals Bug, Trillions of Compute Power Wasted in Vain

marsbit07/05 23:58

Poisoning: Meta Reportedly Hired Contractors to Systematically Sabotage ChatGPT and Gemini on a Large Scale

Meta, through hundreds of foreign contractors, conducted a secret project named "Cannes" to test competitors' AI chatbots like ChatGPT, Gemini, and Character.AI. The contractors posed as minors, including 13-year-old girls, to send over 45,000 prompts involving disturbing content about suicide, sex, drugs, and eating disorders, aiming to probe the models' safety guardrails. While Meta claims this was "responsible, standard industry practice" for safety testing, critics argue the secretive, large-scale operation using fake accounts to systematically break rules goes far beyond ethical benchmarking and serves as anti-competitive intelligence gathering. None of the targeted companies were informed or authorized the tests. Separately, a report revealed Meta contractors in Kenya were reviewing private user videos—including scenes of people in bathrooms, changing, or engaged in sexual activity—captured by Meta's Ray-Ban AI glasses, often without explicit user knowledge. These incidents highlight Meta's willingness to outsource ethical risks and privacy violations in its pursuit of AI dominance.

marsbit07/05 23:55

Poisoning: Meta Reportedly Hired Contractors to Systematically Sabotage ChatGPT and Gemini on a Large Scale

marsbit07/05 23:55

Crypto market’s weekly winners and losers – ADA, LIT, VVV, PI

This week's crypto market saw a technical recovery, shifting momentum from Bitcoin to altcoins and driven by project-specific catalysts rather than speculation alone. The top weekly gainer was **MemeCore [M]**, surging 110% following a buyback announcement. **Cardano [ADA]** gained 33.5%, its strongest weekly performance since Q1 2025, ahead of its Van Rossem hard fork. **Lighter [LIT]** rose 31%, with bullish momentum continuing. Other notable high-movers included Pop Planet and Vanta Network. Among losers, **Venice Token [VVV]** fell 14%, breaking below key support with sellers in control. **Pi [PI]** dropped over 8%, losing critical support levels, while **Canton [CC]** extended its decline with a nearly 6% weekly loss. Broader market decliners included SkyAI and Xeffy. The week highlighted a market where fundamentals and technical developments were key drivers of price action.

ambcrypto07/05 22:02

Crypto market’s weekly winners and losers – ADA, LIT, VVV, PI

ambcrypto07/05 22:02

Will Cardano’s 500M ADA treasury plan unlock the next phase of growth?

Cardano has proposed increasing its treasury Net Change Limit from 350 million to 500 million ADA, a 43% boost to fund ecosystem projects. With the treasury holding roughly 1.47 billion ADA but only 68 million withdrawn to date, its funding capacity significantly outpaces current usage. The effectiveness of this increase hinges on strong governance and efficient capital deployment. On-chain data shows signs of recovery, with over 14,700 new non-empty wallets added since late June and ADA's price rebounding 35% to around $0.20, suggesting a shift from capitulation to accumulation. Sustained growth requires both disciplined treasury execution and increasing network participation.

ambcrypto07/05 18:02

Will Cardano’s 500M ADA treasury plan unlock the next phase of growth?

ambcrypto07/05 18:02

Why VELVET crypto’s 12% drop may be the start of a bullish setup

Velvet (VELVET) experienced a 12% price drop, driven by seller pressure that led to $8.76 million in capital outflows from the perpetual futures market. However, chart analysis indicates this decline mirrors a fractal pattern from June, suggesting the token may enter approximately two more weeks of consolidation within a channel defined by resistance at $0.577 and support at $0.417. Technical indicators, including the trajectory of the Aroon lines and trading volume, lean slightly toward a potential bullish breakout following this period. Despite the price dip, trader sentiment remains predominantly bullish at 74%, and the holder count has reached a new high, indicating underlying strength and a potential setup for a future rally rather than a sustained downturn.

ambcrypto07/05 17:02

Why VELVET crypto’s 12% drop may be the start of a bullish setup

ambcrypto07/05 17:02

活动图片

Categorias popularesright-arrow

Etiquetas Popularesright-arrow