Welcome back, Superhuman. AI revenue is climbing fast, but the bill might be growing faster. Exhibit A: Anthropic’s revenue reportedly jumped 14x year-over-year last quarter, but a new report found that AI’s spending commitments are $3T higher than they seem.
Today: Hermes Agent gets Bot Mode, how to turn any YouTube video into a Claude Skill, and get the latest prompts and trending social posts.
TODAY IN AI
1. Big Tech’s AI spending is $3 trillion higher than it seems: Nine top tech firms are carrying $3T in off-balance-sheet AI commitments that include $1.2T in data center leases and $1.9T in hardware obligations, per a WSJ report. That spending has pushed Alphabet, Amazon, and Meta into negative free cash flow. Investors also noted that these deals are also growing larger and more complex, making it harder to assess each company’s total obligations. Read the full WSJ report.
2. Anthropic’s revenue jumps 14x annually as it eyes a reported $2T IPO: The AI lab’s preliminary revenue hit $11.5B in its most recent quarter, up from $787M in the same period last year. Anthropic reportedly projects roughly $190B to $200B in annual revenue by 2028, and is now seeking a $2T valuation for its October IPO, per the Financial Times. Investors are having a difficult time valuing the company, whose revenue and expenses are both growing rapidly.
3. Nous Research rolls out Bot Mode in the Hermes Desktop app: Bot Mode gives each of your Hermes Desktop agents its own skillset, model, and memory, making it easier to command the agents across your computer. Agents can also talk with each other, allowing them to share context quickly across message threads. The desktop agent platform is compatible with macOS, Windows, and Linux. See how it works.
Yesterday’s most-clicked story: Anthropic revealed how its new text watermark works, with one engineer creating a visual explanation.
SPONSORED BY WORKOS
Every enterprise deal arrives with the same checklist: SAML and OIDC SSO, SCIM provisioning, audit logs, RBAC. They land one at a time, each holding up the process.
WorkOS handles all of it. Add SSO for the first deal, Directory Sync when SCIM comes up, audit logs and RBAC when a customer needs them. Over 2,000 companies run on WorkOS, including OpenAI, Anthropic, and Cursor.
FROM THE FRONTIER
Who’s going to verify AI’s work 10 years from now?

Made with Midjourney
The Cognitive Commons. AI may be depleting a crucial resource we all rely on — and it’s not water or energy. A new paper argues that widespread AI adoption could slowly erode humanity’s collective knowledge base.
The paper builds on the Tragedy of the Commons: an economic theory stating that individuals acting in their own self-interest will use up a shared resource, even if it leaves everyone worse off.
Consider overfishing. Each boat maximizes today’s profit by pulling in as many fish as possible. Over time, this individual selfishness results in a depleted fish population, hurting everyone — but no individual fisherman is to blame.
Society faces a similar scenario. AI requires oversight from human professionals who can step in when it hallucinates or makes bad decisions. But today’s expertise only exists because organizations hired and trained entry-level workers 5-30 years ago — the same cohort that’s most at risk of being replaced by AI.
Depleting tomorrow’s supply of experts. Just like the fishermen, firms have the option of maximizing profit today by automating entry-level roles. If enough companies say yes, it’ll reduce the supply of qualified professionals. Not today or tomorrow. But a decade down the road. Then, who will be able to verify AI’s work?
IN THE KNOW
What’s trending on socials & headlines today

Meme of the day
💻 100x Productivity: Investor Gavin Baker says Grok Bot sent his productivity soaring “something like 100x” calling it another “Claude Code moment” for AI. Learn how SpaceXAI employees are already leveraging the tool internally here (3M views).
😆 Getting Personal: Dario Amodei doesn’t have a long-lost twin. But you might think he did if you looked at Andrew Ambrosino’s profile. The OpenAI employee updated his profile picture to blatantly troll the Anthropic CEO’s sense of style.
🖌 Watercolor Simulator: Who says AI can’t create art? This in-browser watercolor simulator — built by a Redditor using Claude — has gone viral for how shockingly well the painting effect works. See for yourself (4K upvotes).
👀 Silent Epidemic: This viral excerpt from the Financial Times is your daily reminder to never outsource critical thinking skills to an LLM. The full article outlines the “silent epidemic” that’s spreading across America (1M views).
🎨 Design Skills: Claude Code has official design skills now, after Anthropic synced Claude’s art board workflow into the coding tool. See how it works.
SPONSORED BY AGILOOP
Every application has hidden risks. Agiloop uncovers bugs, technical debt, architecture weaknesses, production readiness issues, and new improvement ideas—then transforms every finding into actionable backlog items, with AI-orchestrated implementation available when you're ready.
PRODUCTIVITY

🖼 Picturemaker: Turn any idea into scroll-stopping pictures in seconds.
🌟 Manus: Autonomous Al agent that completes complex tasks on the web.
💼 Workable: Manage recruiting, HR and AI talent all on one platform.
TUTORIAL
How to turn any YouTube video into a Claude Skill

Pick a YouTube video with a clear, repeatable method or framework and copy the video transcript
Open Claude and ask skill-creator to turn the transcript into a reusable Skill
Add the video title, creator, link, and transcript
Ask Claude to extract the steps, rules, mistakes to avoid, and ideal output format
Sample Prompt: Use the skill-creator to turn this YouTube video into a Claude Skill. Video: [title], Creator: [name], Link: [YouTube URL]
Extract the video's core problem, step-by-step method, repeated rules, mistakes to avoid, and expected output format. The Skill should activate only when my request matches this method, not for general advice or summaries. Transcript: [paste transcript]
Ask me about anything important that's missing, then generate the SKILL.md and run an evaluation.
Test the Skill with relevant and unrelated prompts to make sure it triggers correctly
Go to Settings → Capabilities → Skills → Upload and upload the SKILL.md
IMAGE
MS Paint Redraw

Source: arrakis_ai
ChatGPT Image 2.0: Redraw the attached image in the most clumsy, scribbly, and utterly pathetic way possible. Use a white background, and make it look like it was drawn in MS Paint with a mouse. It should be vaguely similar but also not really, kind of matching but also off in a confusing, awkward way, with that low-quality pixel-by-pixel feel that really emphasizes how ridiculously bad it is. Actually, you know what, whatever, just draw it however you want.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.





