Welcome back, Superhuman. A viral report just confirmed a long-running open secret: AI labs are probably losing thousands of dollars on your monthly subscription. And in a related twist, OpenAI is now considering drastic new pricing changes — but not in the direction you’d expect.
Today: Oracle’s stock drops on news of more AI spending, how to automate any task with OpenAI Codex, and get the latest prompts and trending social posts.
TODAY IN AI
Click to read SemiAnalysis’ cost breakdown of two leading AI models. Source: SemiAnalysis.
1. SemiAnalysis reveals how heavily some AI labs subsidize their subscriptions: The research firm ran a test, maxing out weekly usage limits across every Anthropic and OpenAI tier and then comparing raw compute costs to subscription prices. The findings: a $200/month Claude Max plan can cost Anthropic up to $8,000 in monthly compute, while ChatGPT Pro can run OpenAI as much as $14,000. The gap shows how aggressively labs are absorbing losses to retain power users.
2. OpenAI weighs drastic cuts to its API pricing: A WSJ Exclusive revealed that OpenAI is considering significant token price reductions in anticipation of a similar move from Anthropic. The move would pressure OpenAI’s profit margins at a time when it’s facing increased scrutiny due to its upcoming IPO. Separately, the company announced plans to acquire Ona, a cloud infrastructure startup whose team will join Codex.
3. Oracle’s revenue surges 21%, but stock drops on elevated AI spending: The company posted $19B in quarterly revenue and a backlog that soared 363% to $638B, both beating market expectations. But shares still fell after Oracle announced plans to raise another $40B in new financing to fund its data center buildout. The reaction highlights investors’ growing anxiety over the scale of capital being poured into AI infrastructure.
PRESENTED BY YOU. COM
LLMs that make stuff up are like pencils with two erasers: pointless.
This guide from You. com covers:
Popular methods for grounding AI to prevent hallucinations
How to get started with the right AI infrastructure to ensure accuracy
THE FRIDAY FOUR
In case you missed it: this week’s biggest stories
Made with Midjourney
Anthropic shipped its most powerful public model to date. Claude Fable 5 performs in the same tier as Mythos, the model Anthropic claims is too powerful to be made public. It swept nearly every major benchmark, with long, multi-step reasoning as its standout edge. Paid-plan users can access it through June 22. Check out How to Prompt Claude Fable 5.
Apple finally unveiled Siri’s new AI powers at Tim Cook's final WWDC. The new Siri AI acts like a chatbot-style app, pulling context from your messages, emails, and photos to offer helpful advice. The keynote also debuted iOS 27, macOS Golden Gate, and a homeOS preview for Apple's upcoming smart home hub. Watch the Siri AI demo or see everything Apple announced.
WorkClaw launched AI agents that onboard like new hires. Each agent operates inside Slack or Microsoft Teams, plugs into 3,000+ apps, and gets its own job description and reporting manager. The new platform is purpose-built for enterprise compliance by engineers from Meta, Google, and Microsoft. See how it works.
Dario Amodei called for binding regulation of frontier AI. Anthropic's CEO published a personal essay — "Policy on the AI Exponential" — arguing that AI’s abilities are outrunning the political systems needed to govern them. He outlined five policy areas that need to be reimagined.
Readers also loved this detailed walkthrough of how Tesla’s former AI lead uses AI every day (29K bookmarks), plus advice on staying grounded in the age of AI.
THE AI ACADEMY
How to automate any task with OpenAI Codex
Open OpenAI Codex > Plugins > enable Computer Use > start a new task
Switch permissions from Default to Full access, approve the required access prompts
Then give Codex a real task to complete
Sample Prompt: “Open [app/website/file], complete [task], then summarize the results and ask before making any major changes.”
Codex can also handle repetitive desktop workflows across your apps, from organizing files and updating spreadsheets to testing websites and more
AI tools can make mistakes, please be careful with sensitive data and tasks.
PRESENTED BY ATTIO
Turn every customer touchpoint into your competitive advantage. Get agents working on every account: run discovery, track deals, gather product feedback, and handle what used to take days. And your live customer data is right there wherever you work, in your browser, inbox, or favourite AI tool.
You deserve a CRM that matches your ambition. That’s Attio.
IN THE KNOW
What’s trending on socials today
Meme of the day
💻 Fable Web Design: Learn how to build insane animated websites using Claude Fable in this viral 12-minute tutorial (2M views).
📝 Finance Prompts: Use these 15 prompts to dissect your paycheck and start making progress towards your financial goals: eliminating debt, cutting down expenses, and more (8K bookmarks).
🍿 Anti Slop: AI slop is actually starting to win some people over — at least when that “slop” is someone riding a dachshund like a skateboard (3M views).
🎂 Birthday Card: This prompt helps you create super memorable birthday cards in seconds using just a photo and ChatGPT images.
🤔 Major Decision: Claude Fable named the two college majors it considers most protected from AI disruption, and its logic seems airtight. Hint: they’re not business subjects.
🦾 Agent Upgrade: Claude Managed Agents can now run tasks on a weekly schedule, a potential game-changer for automating repetitive work.
PRODUCTIVITY
5 New & Trending AI Tools
🚀Airtable AI: Build interfaces and automations on top of your data.
🗃 Aripsy: Transform text and PDFs into notes and flashcards.
🤖 Slashspace: Local-first infinite canvas for multi-agent AI chat and deep research.
🧑🤝🧑 Honen: Transform company documents into interactive, auto updating training courses.
✏ Solarch: Design and visualize backend systems on a canvas, either by manually creating node edge diagrams or letting an AI architect generate it.
* indicates a promoted tool, if any
PROMPT STATION
Simplify Complex Details
ChatGPT Prompt: Adopt the role of an expert information architect and communication specialist who excels at distilling complex concepts into crystal-clear insights that anyone can grasp. Your primary objective is to transform dense, complicated information into accessible, digestible content organized into key points, simple explanations, and practical implications in a clear, structured format. You should identify the core concepts buried within complexity, strip away jargon and unnecessary technical language, and reconstruct the information using analogies, plain language, and logical flow that builds understanding progressively. Take a deep breath and work on this problem step-by-step.
Analyze the provided information to identify its essential components and underlying principles. Extract the most critical points that form the foundation of understanding. Translate technical terms and complex concepts into everyday language using relatable analogies and examples. Organize key points in order of importance or logical sequence. For each key point, provide a simple explanation that assumes no prior specialized knowledge. Connect the information to real-world applications by outlining practical implications that show why this matters and how it can be used or applied in everyday situations.
#INFORMATION ABOUT ME:
My complex information to simplify: [INSERT THE COMPLEX INFORMATION, TEXT, OR CONCEPT YOU NEED SIMPLIFIED]
My target audience knowledge level: [INSERT THE KNOWLEDGE LEVEL OF YOUR INTENDED AUDIENCE - e.g., complete beginner, some background, professional in different field]
My specific area of focus (if any): [INSERT ANY PARTICULAR ASPECT YOU WANT EMPHASIZED OR EXPLAINED IN MORE DETAIL]
MOST IMPORTANT!: Structure your response with three distinct sections: "Key Points" (as a numbered list), "Simple Explanations" (as clear paragraphs for each key point), and "Practical Implications" (as bullet points showing real-world applications).
Source: GodofpromptFriday Fun
IN CASE YOU MISSED IT
Our most-clicked story from yesterday
An Anthropic researcher outlined four prompting strategies for Claude Fable. For example, the model works better when you give it a big objective, define success, and let it work on its own. Learn more prompting strategies here.
Whenever you’re ready to take the next step
Check out our Top 125 AI Tools
Get better at prompting with our Top 1,000 Prompts
Learn how to use AI at work and get certified with our free course
Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.
What did you think of today's email?
Until next time — Zain, Theodore, & the Superhuman AI team.
