Skip to content
November 2026 · FridayPublished every day at 08:00

AI Daily · November 2026 · Friday

AITOPNEWS

Issue 820November 2026Friday
12stories9sources4first-party postsAbout 19 min read
Lead

Anthropic begins trading on Nasdaq under ticker ANT at $111.50 per share

Anthropic began trading on the Nasdaq under the ticker ANT at $111.50 per share, up from its $83 IPO price, which had valued the company at around $1.5 trillion. Its S-1 filing reported an $80.7 billion net loss in the first half of this year, mostly from an accounting charge, alongside $16.3 billion in revenue, and outlined a governance structure keeping control with founders rather than shareholders. CEO Dario Amodei said Anthropic will pace training and deployment of AI systems as long as necessary to be confident they are safe and under human control, and the filing lists risks including models that may behave in unintended, misaligned, or autonomous ways.

The Verge · AI1 other sources reportedOriginal
01

Model releases

Latent Space

Google DeepMind releases Gemini 4 Neon, its fourth Gemini 4 model in seven weeks

Google DeepMind shipped Gemini 4 Neon, the fourth Gemini 4 model in about seven weeks, priced at $3 per million input tokens, $0.15 for cached input and $15 for output, with a 1M-token context window and four effort levels. Neon debuted second on Artificial Analysis's Agent Arena with 1,336 points and scored 63.7% on CursorBench 4.1 at default effort, up from Hydrogen's 57.3%. It is rolling out to the paid Gemini API in AI Studio and Android Studio, Gemini for business and Antigravity for Pro and Ultra subscribers, with day-one integrations in Cursor, GitHub Copilot and LangChain Managed Deep Agents.

02

Product releases

Baseten BlogFirst-hand

Baseten brings deferred tool loading to open models, cutting coding agent cost by 62.5%

Baseten added deferred tool loading to open models (DeepSeek, Kimi, GLM) on Baseten Model APIs, supporting both Chat Completions and Messages API formats. In Claude Code with 3 MCPs enabled, coding benchmarks ran 62.5% cheaper and 26.2% faster on average, while agentic benchmarks were 51.1% cheaper and 14.9% faster. Instead of injecting every tool schema into the system prompt, the harness passes tool names plus a tool_search tool (~1.3k tokens versus ~141k tokens for 3 MCP servers) and renders the full schema inline only when needed, keeping the KV cache prefix valid.

AWS Machine Learning BlogFirst-hand

Amazon Polly launches voice cloning for brand voices

Amazon Polly voice cloning is now generally available as a gated release, building a custom voice from a few minutes of a consenting speaker's recordings on the Polly generative engine. It launches in seven languages (English, Spanish, German, French, Italian, Portuguese and Hindi), with each cloned voice speaking only the language it was recorded in. Voices are encrypted, scoped to the creating AWS account and controlled through IAM, and creating a voice is free while usage is billed pay-as-you-go.

Databricks BlogFirst-hand

Databricks Genie ZeroOps Enters Public Beta: An Agent That Automatically Fixes Data Pipelines

Databricks released Genie ZeroOps in public beta, an autonomous agent that detects, investigates, verifies and remediates data pipeline issues, with workspace admins able to enable it from the Previews page. The company says private preview testing covered over 80,000 real customer issues across hundreds of companies, and that more than half required organization-specific knowledge to triage. New features include inbox sharing, notifications, one-click PR creation and custom business instructions; ZeroOps never changes code or data without human approval.

AWS Machine Learning BlogFirst-hand

Amazon Bedrock AgentCore harness upgrades to Strands defaults with built-in tools and lifecycle hooks

Amazon Bedrock AgentCore harness now runs on the open-source Strands harness, giving agents managed defaults such as a recommended model, a task-oriented system prompt, automatic prompt caching for Anthropic models on Amazon Bedrock, and proactive context management. The upgrade adds built-in tools for planning, web retrieval, granular filesystem access, context offloading and subagent delegation, plus four lifecycle hooks (before_invocation, before_tool_call, after_tool_call, after_invocation) that let Lambda functions allow or deny actions while SNS and EventBridge targets send notifications. AWS says no new resource type or integration is needed, existing explicit configurations still take precedence, and the rollout to new and existing harnesses began on November 11.

03

Industry

Big Technology5 other sources reportedFollow-ups

Eight Thoughts on Anthropic's IPO and What Comes Next

Anthropic raised $81 billion in its IPO on Thursday, selling 980,000,000 shares at $83 each, with shares set to open above $100 and a $2 trillion valuation in sight, making it the largest IPO in history. 7 billion in Q1 2026 to $17 billion in Q3.

TechCrunch · AI

Timeline: Jensen Huang's affair with an alleged Chinese spy and his exit from Nvidia

Nvidia co-founder Jensen Huang resigned as CEO after admitting an extramarital affair, with U.S. officials believing Megaspeed executive Alice Huang, one of his romantic interests, is a Chinese spy. Nvidia's board accepted the resignation and named CFO Colette Kress interim CEO, while the company reported $115.8 billion in quarterly revenue, up 103% year-over-year, and said it does not plan to reinstate Huang. Huang said he would be honored to return if the board decided it was in shareholders' interest, and he agreed to testify before Congress before Thanksgiving.

The Decoder

US export controls hit Chinese AI firms as TSMC and Samsung halt chip shipments and overseas cloud access closes

Chinese AI companies are facing immediate fallout after the US Commerce Department closed the cloud loophole for AI chips, with overseas data centers now requiring a BIS license and a 90-day transition period. Moonshot AI canceled a $150 million contract with an Indonesian cloud provider and paused further talks; it had planned to rent Nvidia Blackwell Ultra B300 chips there to train Kimi K4 and had agreed to prepay about 50 percent. TSMC and Samsung also stopped certain processor shipments to Chinese customers while applying for licenses, affecting customers including ByteDance and Baidu's Kunlun, and exemptions for non-AI chips were eliminated as well.

TechCrunch · AI

Report: Meta's AI agent Muse mixes up conversations and may expose private data

Meta employees say the company's AI agent Muse has mixed up conversations across different accounts and surfaced private data, according to a Reuters report based on internal company posts. Employees described receiving messages about childcare requests they never made, career advice they never asked for, product orders they never placed, and bank account balances that did not belong to them, with one chat adding Capital One and Charles Schwab balances unprompted.

04

Research

Hacker News · AI

Architect Labs unveils Frostbite, an AI-designed interaction-net CPU claiming 4.59x performance-per-watt

Architect Labs introduced Frostbite, a general-purpose CPU with a new instruction set built on interaction nets, which it says its AI systems designed autonomously across compiler, ISA, microarchitecture, RTL, verification and final GDS. On 25 workloads, including 15 held out until final evaluation, Frostbite reached a 4.59x geometric-mean performance-per-watt multiple versus a modern silicon-proven superscalar CPU, with 6.02x on the search set and 3.83x on the held-out set. The chip was synthesized on a 7 nm process at 800 MHz timing and 500 MHz in the final GDS, with each quad plus four L3 nodes measuring 0.792 mm² against 0.895 mm² for the comparison core.

Hacker News · AI

Autonomous Codex audit of Webmin yields 507 reproduced vulnerabilities

An experienced security researcher adapted pre-AI auditing techniques into a fully autonomous Codex campaign that produced 507 canonical, reproduced findings against a pinned Webmin 2.670 source revision. The run consumed 8,347,303,160 model tokens, which at published Standard API rates would cost about US$4,578, and the catalogue records 524 applicability paths with CVSS 4.0 Base assessments plus 567 confirmed historical issues. The author judges the audit captured over 90% of classic defects at that revision, possibly every XSS, and plans a pull request with fixes for findings still needing upstream changes.

05

Briefs

(End of this issue)

AiTopNews · Daily: compiled automatically from public reports, every entry linking its source · Daily bound volume