Skip to content
Mingworks
EN · Language: English

About 2 min read

ChatGPT adds slide tool, Google releases new flagship model

Today's AI news is packed. OpenAI finally added a slide-making tool to ChatGPT that supports real-time editing by multiple people, while Google and Meta each rolled out new large models to go head-to-head. Elsewhere, an OpenAI engineer was arrested in a leaking scandal involving hidden recording gear, and companion app Character.ai entered bankruptcy proceedings after burning through too much cash.

Picked from 290 AIHOT stories, 3 posts on X, 1 podcast episode, 1 blog post.

  • Products3
  • Tips1
  • Models2
  • Industry2

Today's highlights

  1. ChatGPT launches team real-time collaboration slide tool Slides
  2. Google globally rolls out Gemini 4 Argon flagship model
  3. FBI arrests OpenAI engineer suspected of stealing secrets in confidential meeting room

01Products3 items

  1. OpenAI launches slide tool Slides in ChatGPT

    OpenAI has rolled out the Slides feature in ChatGPT. Users can directly generate presentations through natural language conversations, uploaded files, or imported outlines, and apply built-in templates. The feature supports real-time collaborative editing and comments among team members, and can automatically adjust page layouts based on revision feedback.

    Why it matters When preparing reports or speeches, you can directly send an outline to generate formatted slides, with support for multi-user editing.

    Read original· ithome.com(opens in a new tab)
  2. Grok Bot supports importing ChatGPT and Claude memories

    Grok Bot has introduced a Memory Transfer feature, which supports directly migrating user usage preferences and context built up over time in ChatGPT or Claude. Users can directly import historical data. All content requires manual line-by-line confirmation before being written, along with source timestamps, and will not overwrite original settings.

    Why it matters Users who want to switch to Grok without losing their habits no longer need to retrain the AI from scratch.

    Read original· x.com(opens in a new tab)
  3. Apple releases HomeView, its first smart display

    Apple has released HomeView, a smart hardware device equipped with a 7-inch Liquid Retina screen. The device features built-in Siri AI, combining video and audio entertainment, FaceTime calls, home monitoring, and voice-activated automation control. Both the standard and wall-mount versions are priced at 299 yuan [Note: keeping original text currency value as 299 dollars], with sales set to begin on November 13.

    Why it matters Apple ecosystem users get a desktop hub with a screen, making home voice control and video calls smoother.

    Read original· apple.com(opens in a new tab)

02Tips1 item

  1. Open-source traffic light prompt control for agents to avoid wasting credits

    Developer Karl's AI Watts has open-sourced a traffic light prompt mechanism of under 300 words. It has agents output green, yellow, and red statuses at the end of each response: green means proceed, while yellow and red mean tasks are finished and the agent should actively stop and wait for instructions. This effectively solves the issue where tools like Claude Code run out of control and blow past usage limits.

    Why it matters Adding this rule when running automated tasks prevents the AI from talking endlessly and exhausting your call quota.

    Read original· mp.weixin.qq.com(opens in a new tab)

03Models2 items

  1. Google globally launches Gemini 4 Argon flagship model

    Google has started rolling out Gemini 4 Argon globally, calling it its strongest frontier model. Users can experience it through AI Studio, the Gemini app, and Google Antigravity. However, regular free users and Plus subscribers can currently only use the less capable Flash series.

    Why it matters Experiencing this model requires a higher-tier subscription; regular users are currently limited to Flash.

    Read original· the-decoder.com(opens in a new tab)
  2. Meta releases Muse Wonder flagship large models

    Meta has released two flagship AI models: Muse Wonder and Muse Wonder Full. Official data shows that the performance of the two new models in logical reasoning, text writing, and code testing approaches Anthropic's top-tier levels.

    Why it matters Meta has strengthened its core models once again, providing new alternatives for daily writing and coding.

    Read original· ithome.com(opens in a new tab)

04Industry2 items

  1. FBI arrests engineer suspected of stealing secrets in OpenAI office

    The FBI announced the arrest of 35-year-old OpenAI engineer Abdulrahman Alfozan. The charges allege that he hid an encrypted recording device wrapped in toilet paper behind a television in a company confidential meeting room, suspected of stealing trade secrets and threatening national security. OpenAI captured the employee attempting to retrieve the device after installing security cameras.

    Read original· arstechnica.com(opens in a new tab)
  2. Character.ai files for bankruptcy and discloses heavy losses

    Star companion app Character.ai has filed bankruptcy documents. The documents show that the company long burned through high amounts of cash with limited revenue, compounded by massive legal expenses from user suicide lawsuits and minor protection regulations, making it ultimately unsustainable. The company acknowledged that even excluding legal dispute expenses, its business model is difficult to sustain long-term.

    Why it matters Readers who enjoy using chat companion products should note the risks of data retention and sudden shutdowns.

    Read original· 404media.co(opens in a new tab)