Edition no. 2Artificial intelligence · English editionOct. 31, 2026

Noticiário do Joãozinho

joaozinho.news
Saturday, October 31, 2026Evening edition11 stories · 1 video

Noticiário do Joãozinho — Edition no. 2 · Oct. 31, 2026 · Evening

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 story

OpenAI cuts the target for its new round to up to US$20 billion, at a US$1.25 trillion valuation

CNBC confirmed this Saturday (10/31) that OpenAI is seeking to raise up to US$20 billion in the ongoing round, below the previous target of about US$30 billion. According to the Financial Times, which broke the news, the round values the company at US$1.25 trillion post-money; OpenAI gave investors until Friday to commit, and some reduced their stakes. Investors from the United Arab Emirates are leading the round alongside Coatue. The round has not closed yet and the numbers may change.

Why it matters It is the first concrete sign that the market's distrust of AI spending, which dragged tech stocks down on Thursday and Friday, has also reached the labs' private capital. Less money and a lower valuation affect OpenAI's ability to pay for data centers and a price war, and they come in the same week as Anthropic's IPO filing.
Download the edition PDF (in Portuguese) Permalink to this edition

Models 1 story

Medium Oct. 31, 2026 · Kimi API Platform

Moonshot closes the Kimi K2 series to new API users and sets its retirement for December

Since 10/31, the K2 series models (such as kimi-k2.6 and kimi-k2.7-code) are no longer available to people who sign up now on the Kimi platform; existing users are not affected. Moonshot plans to retire the entire K2 series by the end of December 2026 and recommends migrating to kimi-k3.1-pro.

Why it matters Anyone with a product running on K2 now has a two-month deadline to migrate and retest. It is a reminder that Chinese API models also turn over generations quickly.

Source: Kimi API Platform

Story page · WhatsApp ·

Research 1 story

High Oct. 31, 2026 · WIRED

Agent powered by Claude Sonnet 5 picks a gene therapy and performs an injection into a live mouse's brain

The startup Exin Therapeutics connected an agent to motorized surgical equipment and only asked it to reduce the seizures of a mouse with epilepsy. The agent chose one of 19 gene therapies, the brain region and the coordinates, and controlled the needle and the injection. A veterinarian opened the skull, approved every irreversible step and held an emergency button. According to WIRED, Claude Opus 5.5 refused to perform the surgery and the team switched to Sonnet 5; midway through the procedure, a calibration error stalled the agent, which refused to change its own code, and the fix was made by another agent. The effect on the seizures has not been measured yet and nothing has gone through peer review.

Why it matters It shows agents leaving the screen and operating lab hardware in decisions with real consequences. It also exposes the dilemma of safety guardrails: the rule that stops the agent from modifying itself was the same one that delayed the surgery.

Source: WIRED

Story page · WhatsApp ·

Products 1 story

High Oct. 30, 2026 · Google

Google receives the first AI response generated by a TPU in orbit, in Project Suncatcher

The Project Suncatcher prototype satellite, launched on October 1, ran the open model Gemma 4 on board and sent back a haiku about being in space, received during a pass over Svalbard, Norway. Google calls it the first AI response from a data-center-class TPU in orbit. The next tests evaluate the TPUs under thermal stress and radiation, and the next milestone is two linked satellites in 2027.

Why it matters It is the first practical step of the idea of taking AI computing to space, where solar energy is abundant. It is still an experiment, but it shows where big tech is looking for a way out of the data center energy bottleneck.

Source: Google

Story page · WhatsApp ·

Business and Investment 2 stories

High Oct. 30, 2026 · CNBC

Intel plunges 29% on Friday, the stock's worst day since the 1970s, amid the memory crunch caused by AI

Intel forecast fourth-quarter revenue of US$15 billion to US$16 billion and adjusted earnings of 25 cents per share, versus US$17.1 billion and 43 cents expected by the market. The company blamed weak PCs, squeezed by the price of memory that AI data centers are buying in bulk, and the closure of a plant in Israel. CFO David Zinsner said the PC slowdown should last at least through the first half of 2027.

Why it matters It is the side effect of the AI boom: the race for memory makes PCs and phones more expensive and hurts those who depend on consumers. The drop added to Alphabet's and deepened the week's tech stock sell-off.

Source: CNBC

Story page · WhatsApp ·

Medium Oct. 30, 2026 · TechCrunch

Nvidia in talks to guarantee the debt of Nscale's data center for Anthropic in West Virginia

According to TechCrunch, citing The Information and Bloomberg, Nvidia is discussing a residual value guarantee for billions in construction financing for the data center. The first building, of 460 MW, has a signed US$44.6 billion contract with Anthropic and is expected to use Vera Rubin chips. The guarantee has not been signed yet.

Why it matters If it happens, the chip supplier starts sharing its customer's credit risk, a sign of how the industry is financing the data center build-out at a time of more cautious investors.

Source: TechCrunch

Story page · WhatsApp ·

Regulation 1 story

Medium Oct. 30, 2026 · OpenAI

OpenAI tightens rules against violence in ChatGPT, expands limits for teens and promises a safety report

For users under 18, ChatGPT will restrict practical information about weapons and security systems, even without signs of violent intent. For everyone, OpenAI is limiting detailed reconstructions of mass attacks and says ChatGPT may end conversations that show a fixation on violence and temporarily suspend accounts. The company also promises its first Community Safety Report in early 2027, covering the second half of 2026, with updates every six months.

Why it matters Child safety has become a central topic of regulatory pressure on chatbots. A periodic report with data creates a public yardstick to hold OpenAI accountable.

Source: OpenAI

Story page · WhatsApp ·

Open Source 1 story

Medium Oct. 31, 2026 · GitHub (DeepSeek)

DeepSeek open-sources the dsh-superagent prototype, a proactive coding agent with a Router and Worker architecture

The repository, published on 10/31 under the MIT license and with about 2,000 stars on the same day, contains an agent that listens to a GitHub repository (and, optionally, Feishu), opens a work item for each request and replies in the original conversation, for example with a pull request. The design described by DeepSeek has two layers, Router and Worker, and each task runs in its own session and directory. DeepSeek warns that it is research code, not an official product.

Why it matters It shows how a frontier lab organizes agents that work continuously as a team, with a focus on spending little human attention. It serves as an open reference for anyone building coding agents.

Source: GitHub (DeepSeek)

Story page · WhatsApp ·

Agents and Code 1 story

Medium Oct. 30, 2026 · GitHub (Vercel Labs)

Vercel Labs takes agent-desktop to version 1.0: agents operate native apps through the accessibility tree

The open source tool (Apache-2.0), written in Rust, reached version 1.0 on 10/28 and 1.0.2 on 10/30. It reads the accessibility tree of macOS, Windows and Linux and returns a JSON with a reference for each element the agent can act on; the agent acts on that reference and takes a new snapshot of the screen to check, without comparing pixels. The decision loop stays with the agent that calls it.

Why it matters Computer use by agents still fails a lot with screenshots. Reading the actual structure of the interface tends to be more reliable and cheaper, and it works for any agent, including those running locally.

Source: GitHub (Vercel Labs)

Story page · WhatsApp ·

Local Models 0 stories

No relevant news in this edition.

Finance 1 story

Medium Oct. 31, 2026 · Sicoob

Sicoob launches Sebah, an AI agent for IT operations

The agent supports the IT Operations area, which monitors thousands of events, indicators and alerts from the digital systems 24 hours a day: it consolidates information from several sources, generates reports and analyses, identifies patterns and issues alerts. One of its uses is checking the clearing of boletos (Brazilian payment slips), which must be validated at seven-minute intervals. Sebah is part of Sicoob's AI strategy, built on its own Sisbr IA platform.

Why it matters It shows Brazilian banks and credit cooperatives bringing agents into IT operations, not just customer service, a direct case for anyone working with AI in banking.

Source: Sicoob

Story page · WhatsApp ·

Games 0 stories

No relevant news in this edition.

Brazil 2 stories

Medium Oct. 31, 2026 · Estadão (TecMundo)

UFG launches Lumi, an open 4-billion-parameter language model built for Portuguese

The Center of Excellence in Artificial Intelligence (CEIA) of the Federal University of Goiás launched Lumi on Wednesday (10/28). The open source model, with 4 billion parameters, was adapted to the Brazilian context and, according to its developers, can run on more affordable computing infrastructure. According to CEIA, it will be made available on Hugging Face for free use.

Why it matters A small, open model in Portuguese lets Brazilian companies fine-tune AI with their own data and run it locally, without depending on services billed in dollars.

Source: Estadão (TecMundo)

Story page · WhatsApp ·

Low Oct. 30, 2026 · IBTimes Brasil

Finep approves R$300 million for a clinical intelligence platform with Einstein

Brazil's Financier of Studies and Projects (Finep) approved R$300 million for a platform, in partnership with Hospital Israelita Albert Einstein, that will integrate health information, AI models and medical decision support tools, to expand the use of data in the hospital setting and support health professionals.

Why it matters It is large public funding for AI applied to health in Brazil, in an area where organized data makes a direct difference in patient care.

Source: IBTimes Brasil

Story page · WhatsApp ·

Videos 1 video

Our first response from the Project Suncatcher prototype satellite.

Google · Oct. 30, 2026

Google shows the first response generated by Gemma 4 on a TPU aboard the Project Suncatcher satellite: a haiku about being in space, received during a pass over a ground station.

Watch on YouTube

About this edition

Edition
Edition no. 2 · Evening edition
Date
Oct. 31, 2026
Coverage
Covers what came out since edition no. 1 (closed around 5 p.m. Brasília time on 10/31), the hourly watch leads, and news from 10/28 to 10/30 that had not yet been published. Closing: 10/31/2026, around 6:30 p.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.