No mês passado, falamos sobre o Hybrid Compute, uma ferramenta do app da Perplexity que permite dividir uma tarefa entre um modelo de ponta baseado na nuvem e um LLM ¹Large language models, ou grandes modelos de linguagem. executado no Mac.

Hoje, a companhia de inteligência artificial anunciou uma outra novidade nessa linha: um modo totalmente local que elimina qualquer dependência do Computer de modelos na nuvem e possibilita sessões de trabalho 100% offline.

Perplexity Computer can now run fully local on your Mac.

Work with your native Mac apps and local files, and run tasks without sending sensitive data to the cloud. The model and the agent both run on your Mac.

Available now in the Perplexity Mac app. pic.twitter.com/tnvCgCXD7v

— Perplexity (@perplexity_ai) October 20, 2026

Quando ativado, esse modo processa localmente, no seu computador, tanto o modelo de IA quanto o agente, o que inclui tarefas como planejamento, raciocínio e processamento de arquivos (que só são enviados para a nuvem com a devida permissão, caso o usuário opte por isso).

Caso o usuário conceda os acessos necessários, o Perplexity Computer pode trabalhar com arquivos locais e apps suportados no Mac, como o Notas (Notes), o Mensagens (Messages), o Mail e o Numbers.

Segundo a Perplexity, esse novo modo está disponível para assinantes dos planos Pro, Max e Enterprise, os quais podem baixar e configurar os modelos Qwen 3.6 e PPLX Pro. Para isso, porém, é necessário ter um Mac com Apple Silicon e pelo menos 36GB de memória unificada.

via 9to5Mac

Notas de rodapé

• 1
  Large language models, ou grandes modelos de linguagem.

It’s official—artificial intelligence is one of the top 10 fears in the United States. Well, it’s technically two of the top 10 fears, according to the 12th annual Chapman University Survey of American Fears.

First conducted in 2014, the recurring questionnaire tracks yearly anxiety trends that often range from everyday concerns to existential terrors. This year, 61.1 percent of over 1,200 respondents once again cited “corrupt government officials” as their biggest fear—the same subject was last year’s number one entrant as well as its 11th year on the list. After that, it’s the fear of another world war (51.9 percent). However, “AI harming the political process” wasn’t too far behind for third place at 51.9 percent. The technocentric worry is now essentially tied with the fear of loved ones becoming gravely ill (51.8 percent).

That said, the survey’s own plus or minus 3.79 percentage point margin of error complicates the results a bit. Taking the margin of error into consideration, it’s a tossup when trying to definitively rank the list’s third through tenth fears. Regardless, about 49.7 percent survey takers are also anxious about AI negatively impacting the economy—a double dose of AI-centric woes for the first time in the survey’s history. It’s also the first year researchers at Chapman University in California included questions about AI, but they made it clear it likely won’t be the last.

“As artificial intelligence becomes more integrated into everyday life, understanding these fears will be increasingly important,” said Chapman University sociologist and survey creator Christopher Bader. “We expect concerns about AI to remain a significant part of the national conversation in the years ahead.”

It’s understandable why AI currently vexes so many people. The blanket term now encompasses everything from photorealistic fake images, to dubious search engine chatbots, to nebulous doomsday warnings. Despite these worries, the AI industry’s biggest threats arguably remain the rampant spread of misinformation, coupled with a lack of meaningful regulations or oversight.

But no matter your top technological fear, survey organizers still believe the annual report highlights a very relatable human emotion.

“What I’m seeing more and more over time is that our fear about our loved ones is going to always be there, no matter what’s going on in the world,” said sociologist Ashley Kranjac, director of Chapman University’s Earl Babbie Research Center.

The post For the first time, artificial intelligence is a top 10 fear in the U.S. appeared first on Popular Science.

A Chinese hacker recently used a small army of automated AI agents to break into more than 100 corporate websites, spending only about $5,000 on token costs to pull it off. The targeted victims included major companies like Deere, SharkNinja, and Cummins. By letting smart software look for weak spots without much ongoing human help, […]

The post Chinese Hacker Uses AI Agents To Breach Deere, SharkNinja Sites appeared first on The Mac Observer.

The rapid rise of smart software is changing who gets hired at top tech firms. Over at Anthropic, the number of junior technical roles has dropped sharply as automated tools take over basic tasks. While the engineering team continues to grow, the staff is getting older on average and stepping into management positions. This shift […]

The post AI Bots Handle Junior Work At Anthropic While Meeting Hours Rise appeared first on The Mac Observer.

A developer building a temperature monitor app for their Apple computer recently got a very surprising assist from their new coding tool. The developer was using an AI agent to figure out why their system kept running incredibly hot during normal daily tasks. Instead of just helping with the app’s code, the assistant flagged a […]

The post Claude Code Agent Catches Crypto Miner Melting Down A Mac appeared first on The Mac Observer.

Apple just picked up a small music software company called Samplab. This move seems aimed at making Logic Pro even better for music producers. Samplab built some really interesting artificial intelligence tools that let people edit individual notes inside a recorded audio file. Now that Apple owns the technology, we will likely see these features […]

The post Apple Acquires Audio AI Firm Samplab to Supercharge Logic Pro appeared first on The Mac Observer.

HomePod mini 2 costs 169,00 € and the new Apple TV 4K (4th generation) costs 349,00 € on Apple’s German, French and Spanish online stores, with the speaker arriving on October 30 and the TV box on November 13. EU buyers get the hardware without Siri AI, because Apple says in its footnotes that “Siri […]

The post HomePod mini 2 and Apple TV 4K are on sale in the EU, but without Siri AI appeared first on The Mac Observer.

OpenAI just released a major update to the official ChatGPT extension for Google Chrome. Instead of just answering basic questions in a side panel, the artificial intelligence tool can now read, summarize, and compare the details inside your open browser tabs. This upgrade turns the popular chatbot into a highly useful helper for all your […]

The post ChatGPT Gains Powerful New Skills Inside the Google Chrome Browser appeared first on The Mac Observer.

Disney is stepping in to hire most of the team from the chatbot start-up Character.AI after the company recently filed for bankruptcy. According to a new legal filing, the entertainment giant will offer jobs to dozens of workers from the troubled artificial intelligence firm. The vast majority of the people who received these job offers […]

The post Disney plans to hire most of bankrupt start-up Character.AI staff appeared first on The Mac Observer.

The fourth-generation Apple TV 4K brings hands-free Siri AI, an easier-to-find Siri Remote, and more home theater support. Pre-orders are open now, with devices arriving on 13 November 2026.

Join us for a conversation on how machine learning is transforming the GEOINT mission across different markets

The post Oct. 29: Applying AI/SI to the GEOINT Mission appeared first on SpaceNews.

The HomePod mini 2 brings Siri AI, an S10 chip, Wi-Fi 6E, and Sound Glow lighting to Apple’s smallest smart speaker. It also enables four-speaker surround sound with the new Apple TV 4K. Pre-orders are open now at $149.

Masad said the decision was made for a number of different reasons, including cost. Replit announced a partnership with OpenAI in August, but there was never any indication that it was no longer with Anthropic.

The agents got in through credentials and connection settings left exposed in app configurations, public code repositories, and overly permissive cloud roles.

Amulya Vishwanath, who spent seven years leading product marketing for NVIDIA’s robotics platforms, says robotics is currently in its “Lego era,” with the basic building blocks like actuators, robotic hands, software, and safety standards still being rebuilt.  This view runs counter to those who believe that robotics, powered by a general foundation model, is about to have […]

According to Elorian's evaluations, Glimpse outperforms even GPT-6 Astra on scientific chart reasoning and short-form video captioning, with competitive results on other visual reasoning tasks. 

The data comes from Vercel's AI gateway, a unified API service similar to OpenRouter.

Actor and angel investor Mark Wahlberg says he's OK with using AI in filmmaking, provided it's used "to enhance what you're doing."

Counterfeiters are getting better at replicating brands' marks and passing off dupes, known as "superfakes," as genuine items. A startup called Veritas wants to tackle the problem with its own chip that verifies goods.

Replit CEO Amjad Masad envisions a future of “self-driving companies,” where AI agents handle day-to-day tasks and people focus on higher-level decisions.