In partnership with

November 16, 2026   ·   READ ONLINE

Evolving AI Insights. Keeping you ahead in AI. The daily briefing, 4 min read.

Welcome back, AI enthusiasts.

Jensen Huang's departure from Nvidia may not be the end of the story. Just days after resigning, he's discussing a comeback, with some of Silicon Valley's biggest names pushing to bring him back while Nvidia's board searches for his replacement. Let’s Dive In!

IN TODAY'S ISSUE

01 Jensen Huang Discusses Comeback as Nvidia CEO

02 ChatGPT Agents Can Now Pay With Your Credit Card

03 Alibaba Unveils Its Most Powerful AI Yet

01   /   JENSEN COMEBACK

Source: Bloomberg

Evolving AI:Jensen Huang has discussed returning as Nvidia CEO just days after resigning, with powerful allies pushing to bring him back.

KEY POINTS

  • A Jensen Huang’s ally claims a Silicon Valley "war room" is forming to lobby Nvidia's board and seek White House support.

  • Elon Musk and Tim Sweeney backed a public comeback campaign, while investors urged Nvidia to reconsider.

  • Treasury Secretary Scott Bessent wants an outside CEO, while CFO Colette Kress is interim chief.

DETAILS

Nvidia's board has approached Arm CEO Rene Haas as a successor, while Huang has discussed returning with confidants. One ally claims Huang supports the comeback effort, with executives coordinating through calls and group chats. An online petition with over 150,000 signatures calls for his reinstatement, though Nvidia's board must approve his return.

❝

WHY IT MATTERS

Nvidia's influence over the AI industry was built under Huang, whose partnerships helped make its chips essential. His return could reassure investors and customers, but reinstating him during the investigation could raise questions about board oversight. Choosing a successor carries its own risks, as a new CEO inherits huge AI infrastructure commitments without Huang's relationships. For Microsoft, OpenAI, and other companies investing billions in AI, the concern is whether Nvidia can honor its roadmap and commitments.

AI is only as good as what you tell it

The fastest way to give AI more context is to talk. But that only works if the transcription is right. Wispr Flow runs on Canto, its own speech model, trained on real dictation instead of studio audio.

In Wispr's tests on 10 hours of real dictations from 2,300+ speakers, Canto had the lowest word error rate of every model tested, including Google and OpenAI. It also pulls from your personal dictionary, so tool names, project names and teammates come out spelled right.

Talk through the whole problem, hit enter, and get a better answer. 4x faster than typing, in every app.

02   /   AI WALLET

Evolving AI: OpenAI launched ChatGPT Wallet, letting its personal AI agents, called dots, make online purchases using your saved credit card.

KEY POINTS

  • ChatGPT dots can pay through existing website checkouts without retailers needing a Wallet integration.

  • Every purchase requires biometric approval, while replacement card numbers keep your actual card details hidden.

  • Built with American Express, Mastercard, and Visa, Wallet is rolling out to adult US ChatGPT Pro subscribers.

DETAILS

ChatGPT Wallet lets dots handle the shopping process, from finding products to preparing payments, so users can review and approve orders without entering card details. It builds on OpenAI's earlier Instant Checkout, which only worked with participating merchants. Users retain eligible card rewards and dispute protections, while retailers handle shipping and refunds. For now, users need a US-issued card and US billing address, while fraud checks may block some purchases.

❝

WHY IT MATTERS

ChatGPT could become the starting point for online shopping, putting pressure on Google’s search business and changing how retailers reach customers. If people increasingly rely on AI agents, OpenAI could influence which brands get recommended. Retailers may have to compete for visibility inside AI assistants, as on Google. For shoppers, this could mean fewer websites to visit and less time comparing products, while agents shape what they buy.

A Slack engineer turned her personal Claude Code setup into a shared AI hub for roughly 1,150 colleagues. In this 20-minute talk, she explains how teams can organize their AI tools, share what works, and track real usage so everyone doesn't have to build the same things repeatedly.

03   /   MODEL LAUNCH

Evolving AI: Alibaba launched Qwen4-Max, its most powerful AI model yet, built to challenge leading U.S. models on complex professional tasks while keeping costs low.

KEY POINTS

  • Alibaba's tests show Qwen4-Max beating GPT-6 Astra and Claude Opus 5.5 on an AI coding benchmark, though it trails on several others.

  • Its API costs $4.50 per million output tokens, compared with $20 for Claude Opus 5.5 and $50 for GPT-6 Astra.

  • The model supports a one-million-token context window and can understand text, images, and video.

DETAILS

Alibaba built Qwen4-Max using a mixture-of-experts architecture, which activates just 71 billion of 2.4 trillion parameters per token. On Work360, Alibaba's internal test across 721 fields, Qwen4-Max averaged $1.67 per task versus Qwen3.8-Max's $4.43. In demonstrations, the model helped plan factory-part machining and made 454 tool calls to assess potential Alzheimer's drugs. Qwen4-Max is available through QwenCloud and Alibaba's apps, with smaller open-weight versions planned.

❝

WHY IT MATTERS

OpenAI and Anthropic may still lead on demanding tasks, but agent calls add costs. If Qwen4-Max proves reliable beyond Alibaba's tests, developers could use it for simpler steps and reserve premium models for harder work. This could put pressure on premium AI providers to justify higher prices. For businesses, choosing the right model for each task could make automation cheaper.

04   /   HUMAN OR MACHINE?

One is real. Trust your eyes?

Two images. One photograph. Click the one you think is real.

05   /   QUICK HITS

📉 In a Federal Reserve survey, 87% of market contacts flagged AI as a financial risk, up from 50% this spring.

💻 The llama.cpp 1.0 release adds versioning and stability commitments for developers building apps that run AI on local hardware.

🧠 Thinking Machines released Inkling 2 Small’s weights for download, opening up its 594-billion-parameter model to developers.

🚨 The FBI warns that hackers use the ARTEX AI attack framework to breach networks, sometimes gaining privileged access in under three hours.

06   /   TRENDING AI TOOLS

🗣️ Wispr Flow - Voice-to-text AI that turns speech into clear, polished writing in every app*.

🤖 Rakazo - Runs persistent AI teammates on your own machine and models.

🎬 CaptureFlow - Lets Claude, ChatGPT or Cursor record demos for you.

🧩 Tulvo - Builds docs, boards and workspaces from plain-English instructions.

*PARTNER LINKS

BEFORE YOU GO

Worth your four minutes?

Login or Subscribe to participate

See you tomorrow,
The Evolving AI team

Evolving AI. The home of artificial intelligence.

INSTAGRAM   /   LINKEDIN   /   THREADS

SUBSCRIBE   ·   ADVERTISE

Reply

Avatar

or to participate