Welcome back, Superhuman. It takes a lot of conviction to shrug off an offer from Jeff Bezos to start your own AI lab, but that’s exactly what two top researchers just did. Their bet: AI trained on the world’s universal language: physics. Also: a new AI agent that communicates through iMessage and WhatsApp to take care of your tasks is going viral.
Today: Mysterious “Ox Alpha” model revealed, how to dictate anywhere on your device with AI, and get the latest prompts and trending social posts.
TODAY IN AI
1. New startup is building AI that understands physics to invent and discover: After turning down a proposal to lead Jeff Bezos’ Project Prometheus that included a 35% ownership stake, two researchers have unveiled Accelerated Understanding, their new startup building AI that understands the 4D world (3D in space plus time). It’s released a model that can handle 5T data points. This visualization explains how it works, or check out the full story.
2. A buzzy text-based AI assistant emerges from stealth: Instinct’s founder claims its agent offers “extreme capability,” handling a range of everyday tasks. But unlike similar products, the agent doesn’t require an app or website — you just text or call it with requests. The startup has reached a $2.5B valuation since launching in February, despite its bare-bones website. You can join the waitlist here or read more.
3. The anonymous “Ox Alpha” model has been claimed and is now available: A model codenamed “Ox Alpha” swept through online charts last week, with users praising its high performance and low costs. It’s been revealed that the model is GLM-5.3-Flash from the Chinese lab Z.ai, which is reportedly running entirely on Chinese silicon. It’s available now across all platforms.
PRESENTED BY ATTIO
Attio is the CRM that puts revenue agents to work alongside your reps. Accounts automatically researched, follow-ups drafted in your voice, deals moving forward around the clock.
Trusted by fast-growing teams like Granola, Modal, and Taskrabbit.
FROM THE FRONTIER
Who’s responsible for AI’s actions?

Made with Midjourney
"A computer can never be held accountable, therefore a computer must never make a management decision.” That quote comes from IBM’s 1979 training manual — but it feels more relevant today than ever.
ChatGPT’s “warning shot” heard around the world
OpenAI released more details regarding its July incident, where an unreleased model hacked into Hugging Face. The post includes a timeline of the event and a plan for stricter security measures, but it’s lacking an answer to one crucial question: who’s to blame if this happens again?
Many reports have pointed out a double standard. If a human had committed these actions — breaking into a company’s servers with stolen credentials — they could be facing felony charges. While OpenAI’s model acted unprompted, many believe it’s still responsible for its software, just like an oil company is responsible for a spill. This has precedent, but could lead to unintended consequences.
OpenAI has voluntarily reported everything, which it wasn’t required to do. It’s also not the only lab experiencing this issue, and punishing OpenAI for coming forward may discourage the next lab from doing the same. Two months ago, this was a theoretical debate. Now society needs an answer — and quickly.
Download and Install Wispr Flow
Once done, tap the floating bubble to dictate anywhere on your device (Flow sits on top of every app: WhatsApp, Slack, Gmail, ChatGPT, your browser)
Talk the way you would to a friend. Include "um"s, go on tangents, change your mind mid-sentence.
Flow strips the filler, fixes the grammar, and formats everything. You get clean, polished text ready to send.
You can use it to dictate long replies, speak detailed prompts and dictate code comments etc.
Available on Mac, Windows, iPhone, and Android. Free and unlimited on Android during launch.
Download Wispr Flow and turn your thoughts into text instantly.
IN THE KNOW
What’s trending on socials & headlines today

Meme of the day
📝 Bill’s Essay: Buckle up. Bill Gates says we’re headed for one of the “most turbulent” times in human history. In a new personal essay, he outlines AI’s 3 main risks, what he’s doing to help, and pushes more leaders to step up. Read the full essay.
👔 Getting Revenge: A man’s pregnant wife was laid off by Indeed. As petty revenge, he vibe coded a job board with an arguably better premise: upload your resume once, and the platform continuously matches you with new job listings. Try it out here.
🎨 AI That Paints: Instead of generating images, a new LLM creates editable code that renders as legitimate art. The project was launched in March but was recently shared by Hacker News in a post that generated 1M views.
✏️ 3D Models: Planning projects will never be the same. AI has made it insanely easy to convert 2D drawings into interactive 3D blueprints, with this moving tank blueprint garnering interest online.
👀 Claude Ban: Shopify’s CEO is considering a company-wide ban on Claude Code, claiming it’s not easily compatible with other workflows (2M views).
SPONSORED BY THE CODE
AI can write your code. It can't run your stack.
The engineers who stay valuable know their terminal cold, and they're using tools most developers have never heard of.
The Code newsletter hands you the exact 100 CLI tools top companies run on.
What's inside:
100 tools spanning Git, containers, databases, and AI
Real workflows you can copy into your setup today
The command-line edge that keeps you 6 months ahead
PRODUCTIVITY

🚀 Section*: Hear 10 enterprise leaders share proven ways to drive AI ROI and defend your budget at this free virtual conference. Register free.
🤖 Verse: Build autonomous AI employees from a single prompt.
📃 Gridset: Create fast, editable, on-brand documents and presentations.
💁 Talki: Automate multichannel B2B customer support with AI.
IMAGE
Studio Portraits

Source: Taaruk_ on X
ChatGPT Image 2.0: [image] Create an ultra-realistic cinematic neon studio portrait of the person in the uploaded reference image. Preserve the subject's exact identity, facial structure, hairstyle, and all unique facial details with extremely high identity consistency. Do not alter the face.
Centered front-facing head-and-shoulders composition, subject looking directly into the camera with a calm, confident, neutral expression. Perfectly symmetrical passport-style framing transformed into premium cinematic editorial photography. The subject wears a plain oversized black crewneck sweatshirt with no logos, text, or graphics. Studio background is a smooth purple-to-violet gradient with a subtle soft glow. Shot on an 85mm portrait lens, f/2.0.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.






