Tuesday, November 3, 2026 Edition no. 7 · Morning
PTENES
Noticiário do Joãozinho Search

Artificial intelligence news from Brazil, in English

Noticiário do Joãozinho: Edition no. 7 · Nov. 3, 2026 · Morning

Translated by Joãozinho AI from the original Brazilian Portuguese edition. Facts, numbers and sources are the same as in the original; quotes, video titles and source names stay in their original language. Read the original in Portuguese

Top storyEdition no. 7 · Nov. 3, 2026 · Morning

Salesforce signs a deal to buy Cresta, maker of voice agents for customer service, in a transaction estimated at $2.7 billion

Salesforce announced on Monday (11/2) that it signed a definitive agreement to buy Cresta, described as the “voice-first operating system” for customer experience. Cresta's voice agents resolve complex calls end to end, coach human agents in real time and analyze every interaction; customers cited include Marriott, Comcast and Alaska Airlines. Salesforce did not disclose the price; The Information, cited by Techmeme, puts it at about $2.7 billion. Cresta CEO Ping Wu said the combination embeds conversational intelligence and agentic experiences into Salesforce's service platform. The deal is expected to close in the first quarter of fiscal year 2028, subject to regulatory approval. After closing, Cresta will continue to support the contact-center and telephony platforms customers already use.

Sources

PDF (pt-BR)

Models

1 story

High importanceNov. 2, 2026

Update: in the Pando blog post, OpenAI cites GPT-6 Astra as the model used to formalize the system

In the engineering write-up “Scaling Git at OpenAI,” OpenAI itself wrote that, with GPT-6 Astra, agents wrote Lean and TLA+ to formalize a “theory of Pando,” find bugs and prove system properties. Continuous sessions ran for hundreds of hours. Until now, in previous editions, GPT-6 Astra had appeared only as a market rumor; now the name shows up in an official company engineering blog.

Source OpenAI

Read the story

Research

0 stories

No relevant news in this edition.

Products

2 stories

Medium importanceNov. 2, 2026

Google opens Dreambeans to everyone in India: personalized daily stories with AI

Google Labs opened Dreambeans, an experimental app of daily collections of personalized stories, to people in India aged 18 or older. The app builds the stories with data from the Google apps the user chooses (Calendar, Gemini, Gmail, Photos, Search and YouTube), avoids infinite scroll and lets users dig deeper, save or share. It is on Android and iOS in 14 languages, including Hindi, Bengali, Tamil and other Indian languages.

Source Google Blog ÍndiaTechCrunch on Dreambeans in India

Read the story

Medium importanceNov. 2, 2026

Apple Watch starts Audio Intelligence beta: Live Rewind and conversation summaries via Siri

Apple Watch Audio Intelligence entered developer beta on iOS 27.3, after the announcement at September's Surprise and Shine event. There are four opt-in features on Series 12 and Ultra 4: important-sound recognition, music identification (Shazam), Live Rewind and Siri Recaps. With Live Rewind, a double tap on the Digital Crown shows the last 15 seconds as text; the snippet disappears in 30 seconds if not saved, and a sound and animation warn people nearby. For Siri Recaps, Apple says it does not identify speakers, does not keep raw audio and does not deliver word-for-word transcription — only a short summary with key points; if nothing is saved, recaps erase after seven days. Audio goes through a secure enclave and Private Cloud Compute; Apple says the recap is then deleted from PCC.

Source TechCrunch

Read the story

Business and Investment

4 stories

High importanceNov. 3, 2026

OpenAI hires Dan Driscoll, former U.S. Army secretary, as senior counselor for national security modernization

OpenAI announced Dan Driscoll in the new role of Senior Counselor for National Security Modernization. He advises leadership on AI uses in national security and cybersecurity and works with allies on responsible adoption of what the company calls “democratic AI.” Sam Altman said the U.S. has “an extraordinary opportunity” to spread that adoption with other countries. Driscoll, an Iraq War veteran, said national security requirements are changing faster than at any point in the last fifty years and that AI can cut paperwork and deliver answers faster. According to CNBC, he left the Army after reported tensions with Defense Secretary Pete Hegseth. On his updated LinkedIn he already lists the role; the work is also tied to the Daybreak cyber-resilience program.

Source OpenAICNBC: the hire and the political context

Read the story

Medium importanceNov. 2, 2026

Trajectory raises $45 million in Series A at a $300 million valuation to post-train models without being an engineer

Trajectory, a startup of about 18 people founded about a year ago by three Stanford classmates with stints at Google DeepMind and Apple (Ronak Malde, Michael Elabd and Arjun Karanam), announced a $45 million Series A at a $300 million valuation, led by Sequoia, with Nvidia, Bessemer Venture Partners and Bain Capital Ventures. The platform connects open-source models to the customer's applications, collects feedback and retrains (continual learning / post-training) for specialized tasks — which Malde compares to turning a “smart PhD on day one” into someone with “30 years on the job.” Customers cited include Harvey, Rogo, Decagon, Mercor and Nvidia itself. The post-training platform is becoming publicly available.

Source TechCrunch

Read the story

Medium importanceNov. 2, 2026

NetApp completes purchase of PEAK:AIO to accelerate AI storage at zettabyte scale

NetApp announced it has completed the purchase of PEAK:AIO, which specializes in metadata architecture and high-performance parallel file systems. The idea is to strengthen NetApp Novus and ONTAP with metadata services and a parallel namespace for AI clouds and GPU-as-a-service providers, while keeping an upgrade path for ONTAP customers. CEO George Kurian spoke of “zettabyte-scale architecture”; Mark Klarzynski, CTO and founder of PEAK:AIO, is quoted in the release about the combination.

Source NetApp

Read the story

Low importanceNov. 2, 2026

Squarespace lays off about 19% of staff (~300 people) as AI changes website creation

Squarespace confirmed to TechCrunch, via Jeff Seedman, a restructuring that cuts about 19% of the global workforce — around 300 jobs — to “simplify operations, sharpen priorities and move faster.” TechCrunch ties the moment to the shift toward prompt-based site builders (“vibe coding”), with rivals such as Wix's purchase of Base44 and Google's Stitch. Squarespace itself already has Blueprint AI and announced Forma, a native AI platform on a waitlist. The company was taken private in a $7 billion PE deal in 2024. The statement does not say the layoffs exist “to fund AI.”

Source TechCrunch

Read the story

Regulation

2 stories

High importanceNov. 2, 2026

The New York Times and other publishers sue Brave over scraping to train AI and over bypassing paywalls

According to The Verge (with Reuters), The New York Times and a group of publications sued Brave Software for illegal scraping to train AI and for letting users bypass paywalls. In a complaint from the New York Daily News, the Chicago Tribune and the Orlando Sentinel, Brave allegedly monetized the index of scraped sites by selling AI companies paid access to “real-time online content” from the Brave Index — to train models, answer queries, or both. The complaint cites Anthropic, Perplexity, Mistral, xAI and Cohere among those said to have received content. The browser also promotes a “Paywall blocker.” Times spokesman Graham James said Brave “ignored repeated requests” to stop; Perplexity pushed back with a critique of the Times's lawyers. Brave, Anthropic, Mistral, xAI and Cohere did not immediately respond to The Verge.

Source The Verge

Read the story

High importanceNov. 2, 2026

OpenAI announces next security sprint with Cathedral and more than quadruples bug-bounty ranges

OpenAI is preparing its next internal security sprint in partnership with Cathedral Systems: external researchers use OpenAI's models alongside its teams to find flaws. In the previous sprint, 25% of product engineers were dedicated to finding and fixing weaknesses with Codex. In a separate post, the company more than quadruples, on average, the midpoint of reward ranges for medium, high and critical findings; a critical flaw in ChatGPT, Codex or the API reaches $30,000, exceptional findings still go up to $100,000 and, with bonuses, up to about $172,000. There are multipliers for quality, fidelity and priority campaigns, and retroactive payment for pending reports and bounties from the last 30 days. The team will be at PROMPT2PWN in Oakland on November 20 and 21.

Source OpenAIRewarding defenders with bigger bug bounties

Read the story

Open Source

0 stories

No relevant news in this edition.

Agents and Code

1 story

High importanceNov. 2, 2026

OpenAI launches Pando, a Git backend for giant monorepos used by engineers and agents

OpenAI described Pando, a backend compatible with the Git protocol and client built so its monorepo can keep up with Codex's pace: many parallel pushes on different branches and nearly 1 petabyte of Git data per day. In production, median push latency fell about 5× (about 80%). The system coordinates writes per branch, reuses packfiles across fetches and keeps the Git workflow for engineers and agents. It was built in about 10 weeks with help from Codex and went into production in April. Greg Brockman said on X that the company is considering offering Pando to others and asked for interest at [email protected].

Source OpenAIGreg Brockman on X about Pando

Read the story

Local Models

0 stories

No relevant news in this edition.

Finance

0 stories

No relevant news in this edition.

Games

1 story

Medium importanceNov. 2, 2026

Wanaka Studio launches an AI 3D game studio: describe the idea and play in the browser, no code

On the site wanaka.app, Wanaka presents itself as an AI-assisted 3D game studio: the user describes the game, the agent helps create characters, environments, animations, audio and logic, and the result becomes a playable link in the browser, with no download and no need to know how to program. The site cites formats such as racing, shooting, platform, tower defense, exploration, puzzle, simulation and multiplayer arenas. The main public text on the site consulted does not highlight a “version 1.0” label; secondary coverage attributes a launch to 11/2.

Source Wanaka StudioWanaka Studio

Read the story

Brazil

0 stories

No relevant news in this edition.

Social buzz

5 stories

XMedium importanceNov. 2, 2026

Thariq: $200 plans are already enough for great work — “this is the most expensive intelligence that will ever exist”

Thariq (@trq212) wrote that, after a week, his conclusion is that he does not need unlimited tokens to do great work: the $200 plans “are incredible,” and “this is the most expensive intelligence that will ever exist.”

Engagement 376.8K views7.2K likes143 reposts88 quotes438 replies532 bookmarks (as of Nov. 3, 2026 08:20)

Post @trq212

Read the story

XHigh importanceNov. 3, 2026

Addy Osmani warns of “cognitive surrender”: when the agent's answer becomes yours without you checking

Addy Osmani (@addyosmani) published the article “Critical thinking when agents do the work,” describing a spectrum from offloading (delegating while keeping responsibility) to “cognitive surrender” — a term he attributes to Steven Shaw and Gideon Nave (Wharton): the AI's output silently becomes yours, with no urge to check. He says he is a fan of agentic workflows and proposes habits of hypothesis, verification and ownership of what ships.

Engagement 104.2K views1.7K likes181 reposts48 quotes93 replies3.1K bookmarks (as of Nov. 3, 2026 08:20)

Post @addyosmaniArticle by Addy Osmani

Read the story

XMedium importanceNov. 2, 2026

Viral post: Cartoon Saloon copyright-strikes an AI short accused of copying a Song of the Sea character

The account @CartoonsOTMoon wrote (in Spanish) that Cartoon Saloon removed by copyright the short of an AI creator who had allegedly used a character design without authorization, and that the author got angry saying he did not even know the studio's films. Related coverage and posts link the case to @antrofrog_ and a resemblance to Ben from Song of the Sea; the creator himself, in replies cited by third parties, treated the resemblance as minor and unintentional.

Engagement 970.7K views36.2K likes2.6K reposts257 quotes238 replies3.5K bookmarks (as of Nov. 3, 2026 08:20)

Post @CartoonsOTMoon

Read the story

XHigh importanceNov. 2, 2026

Trending post: Revolut allegedly put ~2,000 engineers on self-hosted GLM and cut its AI bill to ~1/10

The post by @m0xt_ claims that Revolut (valued in the text at $200 billion) switched the default for its ~2,000 engineers from Claude Code/Codex to an open-source GLM model running on GPUs (citing Together AI), with a bill about one-tenth of what it paid OpenAI and Anthropic, unlimited tokens, coding time −60% and lead time −70% — figures the author attributes to Revolut itself. Claude and GPT would remain the “exception.” Treat it as a claim from the post/network; no official corporate confirmation from Revolut was found in this reporting.

Engagement 293.2K views2.4K likes246 reposts82 quotes143 replies2.1K bookmarks (as of Nov. 3, 2026 08:20)

Post @m0xt_

Read the story

XMedium importanceNov. 2, 2026

Hume AI launches the Audio Decision API: one call for questions about audio (speakers, laughter, frustration)

The account @hume_ai introduced the Hume Audio Decision API: questions like “how many speakers?”, “is anyone laughing?”, “is the caller frustrated?” in a single call — which, according to the post, used to require 10+ models. It says it was trained with more than 10 million human judgments, in real time or across hours of recording, and asks for access requests.

Engagement 2.7M views1.7K likes5.1K reposts208 quotes181 replies1.3K bookmarks (as of Nov. 3, 2026 08:20)

Post @hume_ai

Read the story

About this edition

Edition
Edition no. 7 · Morning edition
Date
Tuesday, November 3, 2026
Coverage
Covers what came out since edition No. 6 (closed around 6:30 p.m. Brasília time on 11/2) through around 8 a.m. Brasília time on 11/3, including the overnight watch desk's leads. Closing: 11/3/2026, around 8 a.m. Brasília time.
Sources scanned
Official blogs and changelogs (OpenAI, Anthropic, Google DeepMind, Meta, xAI, Microsoft, NVIDIA, Apple, Mistral, DeepSeek, Qwen, Moonshot), arXiv and HF Papers, Hugging Face and GitHub trending, X, press from the US, Europe, China, Japan, India and Brazil, and YouTube.
Criteria
Only news read in a real source, preferably a primary one, gets in, with every link checked. Pages with access blocks (403) were kept; broken links (404) were discarded.

News by day

See the full archive

PTENES