Histórias consolidadas

Anthropic debuts updated Claude Fable and Mythos models, priced 20% below OpenAI's GPT-6 Astra Fortune

IPO da Anthropic: criadora do Claude começará a ser negociada na quinta‑feira, 19 de nov. Yahoo Finance

Anthropic assegura acordo de energia a gás de US$ 3,4 bilhões para abastecer data centers de IA, avaliação de IPO cai para US$ 1,2 trilhão Brave New Coin

Anthropic mira avaliação em torno de US$ 1 trilhão em IPO de IA blockbuster, abaixo da expectativa inicial Fox Business

"Nova" Sequoia aposta todas as fichas no Vale do Silício e na IA Startups

Presidente da OpenAI diz que empresa pode estar pronta para IPO em 2027 iMasters

Saída de CEO da Nvidia ameaça financiamento de empresas de IA, diz gestor CNN Brasil

IPO da Anthropic: O Momento é Agora para Segurança de AI, Competição e Regra de Equidade Darden Report

Caso Nvidia: renúncia de Jensen Huang pode afetar investimentos em inteligência artificial no Brasil Times Brasil | CNBC

Este é o maior fundo da empresa de investimentos até hoje.

Métricas de Marca do ChatGPT, IPO de $64B da Anthropic: Resumo de IA da DesignRush DesignRush

“Há muitas pessoas seniores na indústria que sabem que isso está acontecendo e, efetivamente, o toleram,” disse um consultor de inteligência de investimentos.

IPO da Anthropic mostra como as Big Tech estão ajudando a financiar o maior boom de gastos da indústria de IA Business Insider

OpenAI ainda não disse quais são seus planos para o futuro das viagens. Mas o fundador da Sonder, Francis Davidson, é uma figura bem conhecida na indústria de viagens.

Cada um dos sete cofundadores planeja vender exatamente o mesmo número de ações na IPO, o que lhes renderia cerca de US$ 250 milhões em dinheiro no ponto médio da faixa de preço, mantendo o grupo no controle da empresa.

Anthropic launches new suits of AI office productivity tools in run up to IPO Fortune

We present TorchCurves, a PyTorch library that treats vector-valued parametric curves as learnable primitives. Its central operation maps batches of scalar inputs to points on curves represented in B-spline or Legendre polynomial bases, with differentiation through both inputs and coefficients. Curve coefficients can be optimized jointly with the surrounding model or produced by another network. The library provides trainable modules, reusable basis modules, and a functional interface, supporting applications ranging from continuous feature embeddings to Kolmogorov-Arnold networks and functional regression. Its implementation combines classical evaluation algorithms with vectorized PyTorch operations, emphasizing readability and maintainability without custom hardware kernels. Timing experiments compare curve evaluation against NumPy and SciPy and KAN training against EfficientKAN acros…

Deploying LLM-based analytics agents in enterprise settings requires evaluation frameworks that can reliably detect failures across complex, multi-tool workflows. We present a three-phase comparative study of three evaluation frameworks, each representing a distinct evaluation paradigm (trace-based LLM judging, text-only LLM judging with red-teaming, and deterministic heuristics), applied to two analytics agents in a controlled research setting using frozen execution traces. Phase 1 quantifies evaluation harness stability through repeated measurements, finding that 97% of configurations meet the stability threshold and that stability patterns are agent-independent. Phase 2 measures known risk detection on 72 ground-truth test cases, where the trace-based framework achieves the best precision-recall balance (F1=0.90) while text-only evaluation matches on recall but generates 5x more fals…

Research ideation is iterative: researchers refine their ideas step by step based on feedback. Yet existing evaluations obscure this search process, making it difficult to understand how LLMs explore the space of research ideas and what limits their progress. We introduce Re-Search, which adapts the format of Twenty Questions to research ideation: given only the research problem of a published paper, a model must rediscover the paper's core idea by asking yes-or-no questions. Built from 200 ICML 2026 papers, Re-Search tracks how a model's idea evolves as it asks questions and receives binary oracle feedback, isolating ideation from execution. Across six recent LLMs, nearly all progress happens within the first 30 of 100 questions, which largely decide how the game ends; after that, most models plateau. Through controlled interventions on the candidate questions and the hypothesis space,…

In this paper, we present 1020 Group's submission to the WMT26 shared task on automated quality evaluation. Our system, Pterodactyl, was engineered specifically for subtasks 2 and 3, that is, segment-level quality score prediction and the detection of error-free segments. Pterodactyl is a highly customizable, ultra-low cost system optimized for binary classification in translation triage and data curation. It is based on an LLM-as-a-judge approach, with prompt engineering and model selection guided by ROC analysis. We also developed several variants of Pterodactyl that were specifically adapted for subtasks 3.

Abeta Summit: crise climática, inteligência artificial e novos investimentos estão na pauta do maior encontro de turismo de natureza do Brasil VOENEWS

B Venture Capital abre seleção para startups www.guairanews.com

Time-Series Foundation Models (TSFMs) are increasingly used in sensor applications, where sampling frequency (SF) varies across sensors, devices, and operating modes (e.g., power saving). Ideally, TSFMs must remain effective during such SF changes, provided the SF retains task-critical bandwidth. In this paper, we stress-test 4 prominent TSFMs: MOMENT, Chronos-2, Moirai-MoE, and TimesFM, to quantify their SF-invariance and investigate whether their behavior tracks the resulting loss of information bandwidth rather than the SF value itself. Our stress-test probes 3 dimensions: (i) the shift in embeddings, (ii) the degradation in forecasting performance, and (iii) the degradation in downstream performance, quantified by three novel metrics: FSI, FDI, and CDI, respectively. To our knowledge, this is the first study to place SF-invariance at the core of TSFM evaluations. Our results charact…

TSMC aprova investimento de até US$ 1 bilhão em Anthropic e OpenAI CNN Brasil

Anthropic lança Ferramentas de Escritório antes do IPO, desafiando a Microsoft Bloomberg.com

Anthropic mira avaliação de US$ 1,3 trilhão, abaixo do esperado, em estreia na bolsa BBC

Anthropic confirmou na terça-feira que seu IPO pode ser um dos maiores de todos os tempos, mesmo ficando aquém das expectativas anteriores. O preço do IPO deve ficar entre US$ 70,00 e US$ 78,00 por ação, o que elevaria sua avaliação a até US$ 1,4 trilhão, segundo o New York Times. O Wall Street Journal relata […]

VER A proposta de IPO de $1,4 trilhão da Anthropic cria um teste de alto risco para o mercado de IA Reuters

Unidade de chips de IA da Alibaba, T-Head, solicita IPO em Hong Kong Reuters

Anthropic AI define faixa de preço das ações kcentv.com

Few-step sampling reduces the cost of streaming video generation. Distribution Matching Distillation (DMD) supervises few-step generators with a separate fake-score network, which adds training memory and computation and must track the evolving generator. We extend APEX to streaming video by using condition shifting to estimate the velocity field of the generated distribution within the generator. The original text condition selects generation, while a shifted condition selects fake estimation using the same parameters. This removes the independent fake network while retaining S2-Avatar's frozen reference teacher, Self-Replay Forcing, and perceptual regularization. We evaluate APEX on 180-second videos using three backbone evaluations per block and a one-step spatial Refiner, with a 70-case human study and a 320-case StreamAV-Bench comparison. Compared with S2-Avatar, APEX achieves high…

Long-horizon personalized agents maintain an evolving memory by continuously deciding what to add, modify, or discard. However, learning an optimal update policy remains challenging. Conventional training relies solely on terminal question-answering accuracy, yielding a coarse feedback signal that treats all intermediate updates identically. Furthermore, update utility is inherently user-specific, as identical memory actions may yield opposite effects across different user personas. Existing methods fail to address this heterogeneity, relying on globally shared evaluation standards with at most a single scalar intermediate reward. To address these challenges, we propose \textbf{Personalized Memory Rubrics} (\textbf{PMR}), a framework that translates terminal feedback into fine-grained, per-step, and per-user credit assignment. PMR decomposes evaluation into a universal layer capturing c…

In reinforcement learning with verifiable rewards, a language model is rewarded for a correct final answer, and GRPO passes that reward to every token of the solution in equal measure. We instead place the credit where the model hesitated, in a method we call MCPO (Margin-Credit Policy Optimization). We measure hesitation by the top-2 margin $M=p_1-p_2$ of the policy's next-token distribution, map it to a score with two fixed thresholds, and spread each solution's advantage over its tokens so that the mean stays unchanged. The margin is read from a forward pass the trainer already performs, so the method needs no extra model evaluation and no batch statistics. We train MiniCPM5-2B-Base with LoRA for 200 steps on 2,048 DAPO-Math problems and evaluate on six competition-math benchmarks. MCPO beats an unmodified GRPO baseline for all five data seeds, by 6.55 points of macro avg@32 on avera…

Users of model APIs and third-party platforms often rely on the declared model name to identify the model responding to their requests. However, the actual model may differ from the declared one, motivating identity verification based on observable outputs. Previous studies have shown that preferences in simple numerical-choice tasks can serve as model fingerprints. We investigate whether a single fixed-length integer sequence can combine multiple numerical observations to identify its source model. For this task, we introduce NumPrint, a benchmark of 5,036 answers from 51 models, collected using Chinese, Japanese, and English prompts that request 294 integers between 1 and 355 in one response. We propose the Shallow MLP and Causal Transformer (SMCT) to capture aggregate numerical preferences and behavioral patterns in generated sequences. In 51-class closed-set evaluation, SMCT achieve…

Investimento em startups cresce 56% no Brasil, na contramão da queda mundial Valor Econômico

Com até $53 million da ARPA-H, os pesquisadores usarão AI para gerar herbicidas mais seguros que reduzam a dependência de importações e melhorem a segurança do abastecimento alimentar dos U.S.

Compositional generalization is an important form of out-of-distribution generalization, defined as the ability to train on some combinations of modules (e.g., functions such as sort, filter) and then generalize to unseen combinations. Despite substantial empirical work evaluating compositional generalization in transformers, findings have been inconsistent because evaluations were often conducted on a limited set of train-test splits, defined in an ad hoc manner. Through a systematic evaluation, we identify two key aspects of the train-test distribution that strongly affect performance: composition equivalence and module coverage. We show that the apparent performance of direct models (trained only on final outputs) results primarily from exploiting composition equivalences, meaning different sequences of modules that reduce to identical end-to-end functions. When benchmarks eliminate …

Simulation optimization typically returns a single estimated optimal point. When decisions also depend on criteria evaluated independently of simulation, such as resource cost, the decision maker instead needs the set of designs whose expected performance lies within a tolerance of the optimum, often choosing the tolerance only after seeing the tradeoffs. We study how to identify these near-optimal regions on a continuous design space, simultaneously for all tolerances up to a prescribed maximum $T$, with region estimation error at most $η$. We show that region identification is not a by-product of optimization. Specifically, when $α>ν$, any procedure, even one given the optimal design in advance, needs at least $Ω\bigl(η^{-(2+d/ν)}\min\{1,KT^{d/α}\}\bigr)$ simulation evaluations, where $d$ is the dimension, $ν$ the smoothness, and $(K,α)$ the covering parameters that describe the geome…

A empresa de IA deve definir sua faixa de avaliação esta semana, informou o Wall Street Journal.

Anthropic escolhe pior momento do mercado para fazer maior IPO da história CNN Brasil

Do Vale do Silício ao Brasil: o que falta para a inovação em saúde ganhar escala? Futuro da Saúde

Fintechs, aplicações de IA e negócios B2B são destaques do venture capital latino-americano no terceiro trimestre Funds Society

ACE Summit aposta em conexões para destravar negócios e investimentos em 2027 Startups

Y Combinator RFS 2027: aposta muda para IA física iMasters

O que muda na Foozi com o investimento da Quartzo para expandir IA que faz compras de restaurantes Business Moment

Endeavor retoma programa Scale-Up com foco em empresas de IA e projeta R$ 200 milhões em receita para turma selecionada Startupi

Três empresas de IA captam mais que Ásia e Europa juntas em venture capital Acionista.com.br

HeadOffice.ai capta R$ 10 milhões para agentes autônomos de IA TV Sim Brasil

Ex-Warren captam R$ 3 milhões com Domo.VC para colocar IA na operação de vendas de empresas Exame

Avanço da inteligência artificial redefine dinâmica de capital e estrutura de equipes no mercado de venture capital Brasil Inovador

Startups de ESG, inteligência artificial e crédito vencem o Procurement Startup Challenge 2026 ABC Agora

Eles têm 24 anos e convenceram investidores de que podem mudar a distribuição de insumos. Vão conseguir? AgFeed

SuperFrete entra em nova fase e reforça aposta em IA com investimento de R$ 150 milhões Pequenas Empresas & Grandes Negócios

Eletronet anuncia novo data center em São Paulo, com investimento de R$ 25 milhões DPL News

O uso de IA continua crescendo, mas os gastos estagnaram. Reduções de preços e controles de custos mais inteligentes entre os maiores gastadores podem desafiar as suposições dos investidores sobre o hiper crescimento da IA.

Nominal, startup que oferece uma plataforma de IA para engenharia de hardware e afirma ter a Anduril como cliente importante, levantou outra grande rodada de financiamento que triplica sua avaliação.

Anthropic lança os primeiros modelos de ponta desde o pedido de desaceleração de Amodei, enquanto o burburinho da IPO aumenta CNBC

Anthropic lança Claude For Advisors para ampliar, não substituir, a pilha tecnológica de consultores existente (pelo menos por enquanto) InvestmentNews

Dilema da IPO da Anthropic após a queda nas ações de IA: 'Condenado se fizer, condenado se não fizer' Yahoo Finance

IPO da Anthropic visto como o verdadeiro julgamento da IA Bloomberg.com