7 topics covered

Listen to today's briefing
0:00--:--

OpenAI GPT-6.2 Astra and Sol Launch

What happened: OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, with early testers and benchmark trackers weighing in on cost and quality.

Key details:

  • Available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, and via the API as gpt-6.2-astra and gpt-6.2-sol.
  • API pricing per million tokens: Astra is $8 input, $0.40 cached input and $40 output; Sol keeps its predecessor's pricing at $2, $0.10 and $10. Both have 1M context and a 128K output limit.
  • Artificial Analysis Intelligence Index: new Astra scores 62, just behind Fable 5.2 at 65. Sol ties September's Astra at about 15% of the cost per task.
  • OpenAI rated Astra High for AI Self-Improvement, a first for one of its externally released models.
  • Epoch verified seven FrontierMath Erdős solutions in Lean out of 68 problems, versus a previous best of three.
  • On Mercor's OpenWorldEngineer, Astra scored 67.2%, Fable 5.2 54.6%, and GPT-6 Astra 44.2%.
  • Enterprise configuration adds a legal search index; Artificial Analysis measured 20.8% on Harvey LAB-AA versus 15.0% for base Astra.
  • Matt Shumer called the models "essentially unusable" after weeks of testing.

Why it matters: Benchmark results show Astra trailing Anthropic's Fable 5.2 on the overall index while leading on several agentic tests, and early reviews are sharply split. Latent Space notes OpenAI leads on intelligence per dollar.

Practical takeaway: If you use Astra for agent or coding work, test it against your own tasks and compare cost per task, since the reported results vary widely by harness and workload.

OpenAI Embedded Independent Assessments

What happened: OpenAI began the first phase of embedded assessments, signing agreements with four independent assessors.

Key details:

  • Embedded assessments give independent experts deep access to a lab's internal models and systems over an extended period.
  • The stated goal is to assess model alignment and the effectiveness of safeguards, and to uncover unanticipated failures.

Why it matters: Giving outside evaluators sustained access to internal models is a step toward publicly verifiable safety reporting, according to OpenAI's framing as summarized by TLDR.

Practical takeaway: Watch for public reports from the four assessors, which OpenAI says can give the public a clearer view of how labs manage risk.

Nvidia Q3 Earnings and Huang Return Statements

What happened: Nvidia reported Q3 earnings under interim CEO Colette Kress, while its board said it does not intend to reinstate Jensen Huang and Huang said he would return if the board wanted.

Key details:

  • Quarterly revenue of $115.8B, up 103% year over year and above the company's expectation of $108B; data center revenue was $107.6B, up 110%.
  • Edge computing revenue was $8.2B, up 42%; profit was $65.3B, a new record.
  • Nvidia expects fourth-quarter revenue of $130B. Kress said "Vera Rubin is off to an explosive start."
  • Board chair Harvey Jones said the board "does not intend to reinstate Huang as CEO" and that the company would not take questions on the circumstances of his resignation or the ongoing investigation.
  • In a statement, Huang said: "If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO."
  • Huang also said: "I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life."
  • More than 193,000 people, including Elon Musk and Tim Sweeney, signed a letter to "bring Jensen back." CNN reported the board is "not considering reinstating Huang."

Why it matters: The earnings show strong AI demand even as leadership remains unsettled, and the board has publicly closed off Huang's return despite his offer and outside support.

Practical takeaway: Watch for a permanent CEO announcement and the outcome of the board's independent investigation, which the company said it will not discuss on the earnings call.

Meta Exec's AI Deepfake of Sam Altman on Instagram

What happened: Meta's chief AI officer Alexandr Wang posted an AI-generated video of OpenAI CEO Sam Altman to promote Meta's Muse agent, drawing attention to deepfake content on Instagram.

Key details:

  • The clip shows Wang, Mark Zuckerberg, Sam Altman, and Jolly, Muse's mascot, on a four-way video call, with Altman dressed like Jolly.
  • The caption reads "When the homies meet about openai product strategy."
  • Meta requires AI-generated videos depicting people to be labeled; static images have no such requirement.
  • A Meta spokesperson, Tracy Clayton, said Wang's post is labeled as AI content and pointed out that Altman has made his likeness available for AI content generation.
  • OpenAI did not respond to requests for comment.
  • Separately, an AI-generated image of actress Mia Goth spread on Instagram in October without disclosure, and was shared by a fan account with a blue verification badge.

Why it matters: A major executive posting an AI video of a competitor is notable on a platform already criticized over AI-generated content, and it shows the gap between labeling rules for video and for images.

Practical takeaway: Expect AI-generated likenesses of public figures to remain common on Instagram, and check labels before trusting clips of executives.

Google Spark Becomes Free Gemini Tasks

What happened: Google rebranded its Spark AI agent as "Tasks" in Gemini and made it free, powered by Gemini 4 Hydrogen.

Key details:

  • Tasks can create meal plans, manage your calendar, and search emails for recurring expenses.
  • US users can have the agent call businesses to compare product pricing and check inventory, purchase items with Google Pay, and sign into accounts supported by Google Password Manager.
  • US users can assign tasks by emailing gemini@google.com or texting it.
  • Spark was first introduced in May and was previously limited to Google AI Pro and Ultra subscribers.
  • Tasks is available to all users in the US, Canada, Japan, Korea, Australia, and India.
  • To use it, users switch the homepage toggle from "Chat" to "Tasks."
  • Google says Gemini runs tasks in a "hardware-isolated virtual machine that's dedicated to your account," and it asks permission before sensitive actions such as purchases.

Why it matters: Agentic features that were paywalled are now available to a wide free user base across six countries.

Practical takeaway: If you are in one of the listed countries, you can try Tasks by toggling from Chat in the Gemini app; expect the agent to ask before making purchases.

Perplexity Frontier Research Initiative

What happened: Perplexity launched Perplexity Frontier Research, an initiative to build new algorithms, models, and evaluations in the open.

Key details:

  • The stated mission is building an automated scientist.
  • The research effort will be supported by 10,000 GPUs.
  • Applications are open until December 5.
  • Results will ship as papers, blog posts, model weights, and open repositories.

Why it matters: It is a large, open-research push from a company better known for search, aimed at AI research automation.

Practical takeaway: If you do research or want to contribute, applications close December 5.

DOJ and Commerce Investigation of Megaspeed

What happened: The Justice and Commerce Departments confirmed they are investigating Singapore-based data center company Megaspeed, which is tied to the resignation of Nvidia CEO Jensen Huang.

Key details:

  • The statement was issued Monday evening, after reporting suggested an investigation had been halted last year.
  • Commerce initiated an investigation into Megaspeed in 2024 for potential export control violations; in May 2025 it requested Nvidia's records on sales to Megaspeed.
  • In June 2025, Nvidia placed Megaspeed on its "no sell" list, and Bureau of Industry and Security agents with Singaporean law enforcement conducted an on-site end-use check and interviewed CEO Alice Huang.
  • After that enforcement action, the DOJ took the lead role in the criminal investigation.
  • The departments said they "will bring the full extent of the law against Megaspeed for any criminal conduct that may have occurred."
  • Megaspeed says Alice Huang left the company in August 2025.
  • The New York Times earlier reported a Commerce investigation into Megaspeed "came to a sudden halt" late last year.

Why it matters: The government's public statement contradicts reporting that the probe had stalled, and it places the case in a national security and chip-export context.

Practical takeaway: Track whether the DOJ files charges against Megaspeed, since the departments have only said they will pursue any criminal conduct that may have occurred.