Welcome back, Superhuman. Pricing has been the most apparent gap between Chinese and US labs, but many have pointed out a second, less-discussed difference: tone. Alibaba just launched a top-performing model with a release video that takes a notably gentler approach, making AI feel far less intimidating.
Today: Congress has a favorite chatbot, how to build automations using Claude Code, and get the latest prompts and trending social posts.
TODAY IN AI
Watch Alibaba’s new most powerful model in action. Source: Alibaba.
1. Alibaba launches a new model it calls an “always-on workmate”: The Chinese lab dropped Qwen-3.8Max, an open weight model rivaling the performance of US frontier models that excels at knowledge work, coding, and long-horizon planning. The lab claims Qwen-3.8Max improved on an existing research paper in 125 hours, coded autonomously for 10+ days, and beat hundreds of human teams in an online competition. Try it on QwenCloud or explore the lab’s guide to designing effective prompts.
2. New disclosure reveals US lawmakers’ preferred LLM: OpenAI’s ChatGPT accounted for nearly 90% of all AI spending for members of the US House of Representatives, with Anthropic’s Claude in second. The report lands as AI companies — including OpenAI, Anthropic, Meta, and Google — prepare to meet with the White House today. The agenda: reviewing a framework for testing and launching new models. Read more.
3. DeepSeek makes its cheapest model more powerful: DeepSeek-V4-Flash now rivals Anthropic’s Claude Opus 4.8 in performance, but charges 28 cents per million output tokens compared to $25 for Opus 4.8. US labs have responded to the consistent pricing pressure from Chinese labs in recent weeks: OpenAI cut prices on GPT-5.6 Luna and Terra last Thursday, while Google shipped an efficient family of models earlier this month.
Yesterday’s most-clicked story: Sam Altman accidentally drew the internet’s ire when he suggested using ChatGPT to create custom podcasts for family car rides.
SPONSORED BY VANTA
Join a technical walkthrough of how startups can use Vanta’s MCP server, agentic workflows, and native plugins to achieve fast, credible compliance without leaving their preferred environment.
You’ll learn how to:
Surface failing tests, open risks, and audit gaps with a prompt
Assign and track remediation without leaving your AI tool
Keep Vanta as your single source of truth, wherever your team builds
FROM THE FRONTIER
AI pricing is on a race towards zero — but only if one condition is met
Made with Midjourney
Intelligence is going the same way as oil or electricity
AI is turning into a commodity. Any of today’s top models are capable of drafting an email or creating a presentation. So unless you’re tackling a genuinely complex problem, the need to pay premium prices for frontier performance is quickly fading. Zack Kass, OpenAI’s former head of go-to-market, calls this “diminishing model returns” — and some companies have already stopped paying the premium.
Some companies are opting for cheaper models over frontier ones
Coinbase CEO Brian Armstrong revealed in June that his company kept AI spend flat while increasing usage, partly as a result of model routing. This tactic directs everyday tasks to cheaper models, and Armstrong specifically named GLM 5.2 and Kimi 2.7 as cost-savers. Microsoft has noticed this trend as well, and is considering adding DeepSeek V4 to Copilot Cowork as a cheaper option.
But commoditization assumes performance stays comparable
The current price pressure holds only as long as cheaper models keep pace with frontier performance. If a frontier lab pulls meaningfully ahead in performance, companies would have a reason to pay a premium prices again. And with how quickly labs are releasing models, this shift could happen on any given week.
IN THE KNOW
What’s trending on socials & headlines today
Meme of the day
😅 Alarmingly Real: Flux 3 is a little too good at generating historical video footage. Its results are so convincing that commenters are raising alarms about the tool’s potential to rewrite history. Watch the clip that sparked the debate (3.5M views).
🤣 Hilariously Wrong: Redditors keep asking AI to create “completely serious” images of common objects being used in the wrong context. The results are equal parts alarming and hilarious. View some of the top images (2K upvotes).
🧠 Thought Provoking: In one word, what’s the biggest lie in human history? Someone asked that to ChatGPT, Claude, DeepSeek, and Gemini — and got four very different answers (2M views).
📚 Interactive Education: It might take some time, but AI will almost definitely reshape education. Exhibit A: one developer vibe coded an interactive website that teaches human anatomy way better than any textbook ever could (10K bookmarks).
🎨 Logo Prompt: This viral logo prompt gives any brand logo a unique new twist by turning it into a holo-vinyl sticker using ChatGPT Images, great for sprucing up a website or social pages (5K bookmarks).
SPONSORED BY OUTREACH
The Outreach AI Maturity Model assessment identifies where your org stands today – giving you a clear roadmap to advance toward AI-Efficient GTM, where AI agents amplify each teammate and work on their behalf.
The result? The clarity leaders need to move from fragmented activities to measurable business impact.
PRODUCTIVITY
🚀 Memoket*: Your AI misses off-screen conversations. Memoket Gem is the AI wristband that captures them and turns them into context your AI agents can use.
👀 Pikes: Create consistent consumer brand visuals with AI.
📊 FigEditor: Create editable scientific research figures using AI.
🤝 Chatzy: Automate customer conversations, support, sales, and operations with AI.
TUTORIAL
How to build automations using Claude Code
Open the Zapier SDK page, copy the setup prompt and paste it in Claude code
Prompt: “Set up the Zapier SDK for me [paste Zapier SDK prompt]”
Sign in to your Zapier account and connect your work apps (Gmail, Slack, Notion, Google Calendar, etc.)
Describe the automation you want to build in plain English
Prompt: “Build an automation that [describe your workflow]. Use my connected apps through Zapier.”
Ask Claude to test the automation once and verify everything works correctly
Once you're happy with the result, ask Claude to deploy it and schedule it to run automatically (e.g. every morning)
IMAGE
Editorial Portrait
Source: Kashberg_0
ChatGPT Image 2.0: Bold surreal editorial portrait of [CHARACTER], shown in a centered frontal chest-up composition, wearing [CLOTHING]. Keep the portrait highly realistic and symmetrical, but interrupt one full vertical section of the subject with translucent overlay [ELEMENT], treated as a dominant surreal replacement strip that occupies roughly one-third to nearly one-half of the face, cutting from the top of head through the eye, cheek, mouth, neck, and into the clothing. The replacement strip must feel structural and graphic, not subtle or decorative, with crisp hard edges and strong visual contrast. Build background as a flat graphic poster field using [BACKGROUND], composed of large pastel rectangles, photo of a [SCENERY] circles, line clusters, cropped panels, and abstract inserts, with a few retro-print accents integrated across the composition. Use a [COLORS] palette, soft paper grain, slight print texture, clean edges, strong geometric balance, and a polished art-poster finish. Make the surreal intervention large, unmistakable, and visually commanding.Vertical 4:5.
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.
