Welcome, humans.

So apparently, the Pope has thoughts on your AI art. Pope Leo XIV argued that human creativity differs fundamentally from software generating images using patterns learned from other people’s work. His objection goes beyond how the picture looks: algorithms lack “the spark of humanity.”

He also called for a renewed alliance between the Church and artists. Michelangelo set a fairly high bar for that collab. No pressure.

And now, for the counter point: please enjoy THIS AI-generated masterpiece… tell me this isn’t a spark of humanity, Mr. Pope! I dare you!  

Here’s what happened in AI today:

• 😺 a16z found few companies reporting tracked AI results.

• 📰 OpenAI tapped former White House cyber official Thomas Lind.

• 📰 Amazon pledged $1B+ to data center communities.

• 🍪 Suno Speech generates spoken audio and original music together.

• 🎓 Karpathy suggests visuals for confusing AI explanations.

… and a whole lot more that you can read about by clicking here.

😺 a16z maps AI’s shift from software to infrastructure

DEEP DIVE: Read our full breakdown on the a16z report

And today, in total market concentration overload: Tech accounted for roughly 76% of S&P 500 earnings growth in 2026 through late August, according to venture firm a16z’s latest markets report. It’s a great read, full of loads of spicy charts!

Our takeaway: AI is changing who gets paid. The biggest tech companies are pouring profits into chips, electricity, and networking equipment. And their spending becomes someone else’s sales.

Here’s how big the buildout has become:

• Alphabet, Amazon, Meta, Microsoft, and Oracle spent $416B on capital investment in 2025, per the report’s spending chart.

• July analyst estimates put 2026 spending at $777B and forecast $1.1T for 2027.

• Major cloud providers’ revenue backlogs more than doubled year over year: customer commitments that haven’t yet become recognized revenue.

Those totals cover capital investment, not AI spending alone. “We’re investing in AI” is papering over a lot of details in those budget meetings TBH.

Important fact: Expanding AI capacity takes lots of equipment and electricity. Big tech profits have funded much of the buildout; now borrowing is funding more. Plus, manufacturing investment, electrification, and defense spending add demand beyond AI.

Older equipment is staying useful, too. NVIDIA’s older (feel like eons old in AI years) A100 chip rental prices remained at or above their start-of-year levels, a16z says. Newer chips arrived, but customers still wanted the old ones’ computing power.

Cheaper AI models to run can make that appetite bigger too. Lower costs make more tasks easier to automate. a16z argues those efficiency gains expand usage rather than simply shrink bills. Jevon’s paradox, folks. Internalize it.

Why this matters: The buildout is easier to measure than the payoff. Nearly 30% of S&P 500 companies reported “quantifiable impact” from AI, but only about 2% disclosed a metric they track over time. Without that, outsiders can’t tell whether the reported gains persist or grow, even if companies measure them internally. Separately, only about 2% of U.S. households paid for AI services in April.

Software, meanwhile, faces a growth problem. About 75% of public software companies were profitable, but only around 30% were growing 20% or more. For my finance nerds in the chat, you know this one: With higher interest rates, investors are less willing to pay high-growth prices for slower-growing businesses.

Our take: a16z sees robotics, biotechnology, and healthcare driving the next wave of AI demand. If that bet holds, AI investment will need to flow into factories, biotech labs, and hospital systems in addition to data centers.

So we gotta ask ourselves: Can a nurse spend less time on paperwork? Can a lab test a promising treatment faster? Can a robot take over a dangerous job without creating a new one for whoever supervises it?

That’s the next flex: what all this spending actually gets people. Preferably something besides another subscription.

Personally, we’re struck by the recent observation that AI would actually make a lot of gains in automated material science vs automated biology science because you don’t need those pesky human trials in the middle. The bottleneck there is actually good material science data. Who’s building THAT?

FROM OUR PARTNERS

Their AI applied for a job while they went for a walk

That's a real request from a TinyFish user.

TinyFish connects to the AI you already use and lets it work on actual websites: it opens the page, signs in, fills in the form and hits submit.

People have had it build new Google Forms, run their Shopify stores and check on their Vercel projects.

More than 250,000 developers use it now, and the last 100,000 joined in 18 days.

Top up this month and you get 30% extra: add $50, get $65.

Claim 30% Bonus.

🎓 AI Skill of the Day: Get AI to explain it a different way

AI gave you an explanation. You read it twice. Still nothing. Andrej Karpathy shared four ways to make model output easier to understand, starting with a writing trick he’s found useful and moving beyond text. Another six paragraphs may not be the cure.

• Make the language simpler. Ask for an explanation using ASD-STE100, an English standard made for aerospace maintenance documents. Karpathy sometimes requests “80% of the way to ASD-STE100” to loosen its strict rules while keeping the writing readable.

• Ask for a diagram. Have the model draw what it’s explaining instead of describing everything in prose. This can make relationships easier to follow.

• Turn it into a webpage. In a coding-capable AI tool, request the explanation “in HTML,” the format browsers display. Karpathy suggests interactive pages and animations so you can explore the explanation.

• Try a narrated video. Ask for a visual explainer like 3Blue1Brown’s, with ElevenLabs narration. Connecting ElevenLabs requires an API key (an access credential); Karpathy also suggests asking for free alternatives that run on your computer.

Our suggestion: take one explanation that lost you and ask for a diagram first. Check it against the original. The real test is whether you can explain the idea on your own, not just whether the output looks impressive.

Have a specific skill you want to learn? Request it here.

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

1. *Discover the potential of artificial intelligence with our comprehensive cheat sheet. Learn more about the concepts, platforms, & applications of AI.

2. Suno Speech generates spoken audio and an original musical score together from text, with the beta open to everyone.

3. FLUX 3 Image gives you control over image composition by drawing boxes where you want specific elements to appear.

4. Shopify Canvas puts store pages in one desktop view for editing themes with Sidekick, in selected-store early access.

5. Axiem edits photos from text prompts entirely offline on compatible Apple devices.

6. pi-durable helps developers resume agent work after crashes by saving conversations and checkpoints; it remains experimental.

📰 Around the Horn

• OpenAI reportedly hired former White House cyber official Thomas Lind to lead cyber and strategic risk policy.

• Amazon pledged another $1B+ over five years for education, training, and other U.S. data center community projects.

• Trillium Labs, a new AI research nonprofit from Nathan Lambert, launched to share open methods for refining already-trained AI models.

• Researchers trained an AI “speech clock” to estimate age from voice; larger gaps from actual age were associated with dementia, a possible biomarker, not a diagnosis.

• Planet Labs joined a coalition launching a $100M, ten-year sensor-and-AI network to track Amazon wildlife.

🌟 Sunday Special

Top 5 Stories of the Week

• OpenAI launched Dots, an AI agent with its own cloud computer and browser, alongside a wider batch of DevDay announcements.

• Meta and Manus expanded their personal-agent offerings with Muse and Manus 2.0, respectively.

• Trump and tech CEOs backed a voluntary AI safety pledge; the administration adopted “super intelligence” terminology in federal communications.

• OpenAI parted ways with three safety researchers, saying they had mishandled sensitive information.

• Tavus said 26 of 54 people mistook Griffin for a human in its company-run, one-minute video-call study.

Top 5 Tools of the Week

• Imbue Studio builds personal software from a description of the workflow you want.

• ChatGPT Try On previews clothing on your own photo.

• Muse for Small Business works across connected business apps in the background, asking for approval before sending, publishing, or spending.

• Claude Sonnet 5.5 tackles coding tasks with faster responses and fewer tokens, according to Anthropic.

• VoiceCap turns meetings into transcripts, summaries, and action items in more than 100 languages.

New from The Neuron: AI Explained

Watch and/or Listen: Spotify | Apple Podcasts | YouTube (click the image above!)

A Cat’s Commentary

Btw: We just launched a robotics newsletter! Sign up for it here.

P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here.