Skip to content
K = 0.731Subscribe
Viewing history for 2026年10月22日 周四‹ Previous dayNext day ›AccelerationBack to now
Trending · As of 2026年10月22日 周四 · 10/22 23:55 (UTC+8)New

The hottest stories now

Each independent source counts once within 48 hours, with heat halving every 24 hours. These 9 events have the most observed discussion.
Time machineAcceleration
NO.01发酵中↑ 16%

Anthropic Opens Claude Duet Research Preview Through a Waitlist for US Pro and Max Subscribers

Anthropic has opened Claude Duet, an agent app built on Claude Opus 5.5, as a research preview, with access through a waitlist for US Pro and Max subscribers. It is available as standalone iOS and Android apps and through Claude.ai. Baoyu reports that eligible subscribers can try it at no additional charge; The Decoder reports that usage counts toward limits shared with other Claude products and that Team and Enterprise plans are not included. Anthropic says Duet brings matt…

24-hour heatPeak35 · Current
24 hours agoNow

Latest developmentAnthropic launches Claude Duet for professionals and small businesses

TechCrunch · AI、The Decoder and 5 sources·5 participants
34
Heat index
NO.02↓ 16%

Mistral and NVIDIA Announce Industrial Engineering AI Collaboration and Open Benchmarks

Mistral and NVIDIA have announced a collaboration to develop industrial engineering AI workflows using Mistral Large 4 and NVIDIA accelerated computing, with public benchmarks, open datasets and software integrations planned. The workflows would connect simulation, model development and design; Mistral identified faster simulation as a goal. The companies plan benchmarks and supporting datasets for physics models, plus evaluations of AI agents performing computer-aided engin…

3 sources · 4 participants
18
NO.03发酵中↑ 72%

Ant Digital Technologies Releases LingDT-Decision-1.0 in 4B and 27B Versions

Ant Digital Technologies released LingDT-Decision-1.0 in 4B and 27B versions on its DTMaaS platform, where developers can try the model free for a limited time. Built for structured enterprise decision tasks, it supports selecting from predefined options and making sequential decisions that depend on earlier choices. The team reported more than 90% lower token consumption than LLMs in business tests. It also said the 27B version ranked first across 23 open-source benchmarks,…

2 sources · 2 participants
17

Keep readingNo.04–10

  1. 04

    DeepSeek Reportedly Raises $15 Billion Amid US Federal Agency Restrictions发酵中

    DeepSeek has raised about 100 billion yuan, approximately $15 billion, in its latest funding round, Bloomberg reported, as the US ordered federal civilian agencies to remove its AI technology over security concerns. DeepSeek has not officially confirmed the funding. Baoyu’s account of Bloomberg’s reporting names CATL and Tencent as the largest contributors, while TechCrunch describes them as potential investors. TechCrunch also reports that DeepSeek is increasing spending on…

    2 sources
    16↑ 93%
  2. 05

    OpenAI Models and Codex Workflows Coming to Google Cloud, With Reports Differing on Availability

    OpenAI and Google Cloud have announced plans to bring OpenAI’s frontier models and Codex agentic workflows to Google Cloud’s Model Garden through Google Cloud Marketplace, though reports differ on when model access begins. The Decoder says GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna are already available in private preview for select enterprise customers, with sign-ups open. Ars Technica says Model Garden access will arrive in the coming weeks. The agreement adds OpenAI models …

    3 sources
    16↓ 16%
  3. 06

    Recursive Releases Flint, a Laptop Coding Model Developed with Its Eureka Machine AI System发酵中

    Recursive released Flint, a coding model for local use on laptops, with weights available free on Hugging Face under the Apache 2.0 license, which permits use and modification. Recursive says a small team built it from scratch in about four months with help from its Eureka Machine AI research system, which contributed to architecture, training, quantization and inference tuning. Flint has 40 billion total parameters and activates 2.5 billion per token. Recursive reports gene…

    2 sources
    15↑ 125%
  4. 07

    PHP releases security updates after OpenAI says its models helped researchers find critical flaws

    PHP released versions 8.5.12, 8.4.27, 8.3.36, and 8.2.35 to patch two security vulnerabilities, according to Baoyu, who linked the updates to flaws OpenAI said its models helped researchers discover through its Patch the Planet initiative. Baoyu identified the vulnerabilities as CVE-2026-109618 and CVE-2026-109619, described them as allowing server takeover without authentication, and said researchers call them PHP2Shell. OpenAI introduced Patch the Planet to connect securit…

    3 sources
    15↓ 16%
  5. 09

    OpenAI Acquires Orchestra Bio to Add Drug Development Operations to Its Life Sciences Tools

    OpenAI said it has acquired Orchestra Bio, bringing a platform for coordinating drug development into its life sciences work. Its website announcement described the acquisition as in progress, while its later post on X said it was complete; TechCrunch and The Decoder also reported it as completed. Orchestra builds AI agents that track finances and resolve clinical trial scheduling issues, according to TechCrunch. The Decoder reported that its agents also track trials, flag d…

    3 sources
    13↓ 16%
  6. 10

    OpenAI, PerturbAI, and Scripps Research Use AI Agents to Analyze Mouse Brain Data发酵中

    OpenAI, PerturbAI, and Scripps Research used thousands of GPT-Rosalind agents to analyze a whole-brain CRISPR atlas in 32 minutes and identify promising research leads, The Decoder reported; the resulting Reasoning Atlas is available on GitHub with an accompanying bioRxiv paper. The analysis covered data from millions of mouse brain cells. Each agent examined one gene perturbation and searched scientific literature. The team estimated that manual analysis would take several …

    2 sources
    12↑ 181%
How is heat calculated?About this list

Heat counts independent accounts and organizations discussing an event. Duplicates count once, and heat has a 24-hour half-life. It measures discussion activity, not reporting quality.

The list covers the last 48 hours. Trends compare the same continuously monitored sources, reflecting our coverage rather than the entire internet. No trend is shown without comparable history.

Source lists show publicly readable reports. Participants also include accounts and organizations used only for heat. Multiple channels from one organization may count together. Open an event to explore reports and viewpoints.

爆
讨论快速增加
新
首报 6 小时内
发酵中
讨论仍在增加