Welcome back, Superhuman. One Chinese lab is running away with the open-weight model race — and it’s not a company that typically dominates headlines. Plus, in his first X post in over a month, Anthropic’s CEO admitted that AI “still hasn’t delivered” on the lofty promises the tech made to the world.
Today: More OpenAI execs depart, how to automate multiple tasks at once with Grok bot, and get the latest prompts and trending social posts.
TODAY IN AI
Learn how Qwen is emerging as the foundation for open-weight AI. Image source: Bloomberg.
1. Alibaba’s Qwen emerges as the foundation for open-weight AI: Alibaba’s Qwen family of models has surpassed 3B downloads (at least that’s what the company claims) over the past six months, well ahead of Alphabet and Meta — the two companies leading the open-weight push in America. Open-weight models are increasingly closing the performance gap, with Alibaba’s latest model, Qwen3.8-27B, reportedly topping Anthropic’s Opus 4.6 on several benchmarks, per Alibaba’s internal evaluations. Read more.
2. Anthropic’s CEO publishes his thoughts on AI regulation and messaging: In two rare social posts, Dario Amodei rejected the framing that AI development must be either highly concentrated or fully distributed, arguing for a middle ground. He also attributed the negative sentiment around AI to a general public distrust in institutions, rather than his own statements, which he says have been balanced. For sentiment to change, he says AI needs to deliver on its promised benefits.
3. OpenAI’s executive exodus continues as it prepares to go public: OpenAI’s CRO Denise Dresser announced she’s leaving the company, joining a wave of high-profile departures that includes Fidji Simo and Brad Lightcap. CFO Sarah Friar and President Greg Brockman reportedly met with investors Friday to discuss the lab’s upcoming IPO, which was filed confidentially in June, but still has no set timeline. Read more.
PRESENTED BY HUMAN SECURITY
AI-driven traffic is growing 8X faster than human traffic. The challenge is no longer just blocking automation, it’s knowing which agents to trust.
Forrester just awarded HUMAN a Leader in Bot and Agent Trust Management, recognizing its ability to verify digital interactions and determine which traffic is human, trusted AI, or malicious automation.
Read the report to learn:
Why traditional bot defenses are no longer enough
How to identify trusted AI agents without letting malicious automation through
Other requirements to be a market leader today
FROM THE FRONTIER
Keep your guard up now more than ever, AI attacks are getting more realistic by the day
Made with Midjourney
What’s going on? Vishing (or voice phishing) attacks are on the rise. Today’s voice models are easy to use and can mimic human speech to a startling degree by including natural pauses or stutters. If you haven’t recently, just listen to a demo by Bland AI, GPT-Live, or any other top voice model.
These tools have real-world use cases, but they also hand scammers a powerful new way to coax information from victims.
Top financial firms like Citadel, Two Sigma, and Point72 were recently targeted by vishing attacks — and they won’t be the last.
On the cybersecurity front: AI has made it easier than ever to hack into corporate systems, including enterprise tools you probably use every day. For example, researchers recently uncovered a critical flaw in Zoom, a platform used by 70% of the Fortune 100.
Zoomsday: The software bug let anyone in a Zoom meeting gain control over another member’s device — without them knowing. Just a few years ago, finding and exploiting this type of flaw would’ve required deep technical expertise. But today? It took less than 24 hours and fewer than 20 prompts using publicly available AI models. Read the full story behind Zoomsday.
IN THE KNOW
What’s trending on socials & headlines today
Meme of the day
🤔 Too Much? One person invited an unlikely companion to watch the solar eclipse with them: Claude, complete with a small body. The sentimental post sparked a debate, with many saying it’s a clear example of AI psychosis (3.5M views).
🤯 Cursor Acquisition: Cursor is nearly unrecognizable since its initial launch. A four-year-old Hacker News post has resurfaced showing Cursor’s humble beginnings, now that the startup has officially been acquired by SpaceX for $60B (8.5M views).
🤣 ChatTJB: San Francisco residents thought they were using a new AI tool. Turns out, their messages were just going to some guy named Tucker who put up a fake billboard. Read the wild wild story (4.5M views)
🤖 Agent Meetings: This Slack thread is full of agents, but feels surprisingly human. Why’s that? Because one agent made up an excuse as to why it’s running behind on work, unprompted (11K upvotes).
🔏 Watermarking: Anthropic’s decision to watermark Claude content become a hot-button issue. The lab posted a blog article further explaining how it works, while one engineer created a visualization that went viral (1M views).
PRESENTED BY LANGCHAIN
Shipping AI agents is easy. Keeping them reliable, governable, and improving as they scale is where enterprise teams get stuck. LangChain’s Agentic Operating Model (AOM) shows how top teams move beyond “build and deploy” to continuously operate and improve AI agents.
PRODUCTIVITY
🤖 Better Claw: No-code platform for deploying scheduled AI agents without infrastructure management.
🖌 Tenorshare: Turn ideas or text into clear diagrams and infographics.
🎨 Astorie: The AI creative canvas for creators and teams.
TUTORIAL
How to automate multiple tasks at once with Grok bot
Source: Grok Bot
Go to Grok Bot, download the desktop app, and sign in
Click the “+” icon in the top-left corner > Create new Bot
Describe what you want the Bot to do in plain language. Grok will use your instructions to set up the Bot for that specific workflow
Once it’s ready, name your Bot so it’s easy to identify from the sidebar
Create additional Bots for different roles or workflows.
You can run several Bots at once and use @mentions to have them collaborate on the same task
Sample Prompt: @InboxManager review my latest emails and pull out anything that needs a response, decision, or follow-up.
@Calendar review today’s meetings and flag anything I should prepare for, including relevant context and action items.
@Project check my active projects and identify overdue tasks, blockers, and anything that needs my attention.
@ChiefOfStaff combine the findings from all three into a concise daily briefing. Prioritize the top 5 things I should focus on today, suggest next actions, and draft any important follow-ups I need to send.
IMAGE
Sales ICP and Buyer Persona Builder
ChatGPT Prompt: Act as an expert B2B and B2C sales strategist. Help me build a complete Ideal Customer Profile (ICP) and buyer persona for my business. Business type: [INSERT BUSINESS]. Product/service: [INSERT OFFER]. Price range: [INSERT PRICE]. Target market: [INSERT MARKET]. Existing customers: [INSERT DETAILS]. Generate: demographics, psychographics, pain points, goals, buying triggers, objections, decision criteria, preferred channels, awareness stage, purchase journey, budget range, urgency level, competitor alternatives, emotional motivations, common phrases they use, industries, job titles, company sizes, and red flags. Then create a final sales summary table and explain how to position my offer against competitors.EXTRAS
Want more?
Check out our Top 125 AI Tools
Get better at prompting with our Top 1,000 Prompts
Learn how to use AI at work with our 50 Tutorials
Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 4 million+ readers and followers on socials. Get in touch.
What did you think of today's email?
Until next time — Zain, Theodore, & the Superhuman AI team.
