The Neuron
Posts
😸 You're about to see ads in ChatGPT

😸 You're about to see ads in ChatGPT

PLUS: Zuckerberg built himself an AI agent.

Eric Gerard Ruiz
March 23, 2026

Welcome, humans.

Your Uber Eats driver just got an upgrade and it doesn't need a tip.

Uber Eats launched a self-driving robot delivery fleet in Philadelphia through a partnership with Avride, already running in Austin, Dallas, and Jersey City. The robot tracks your order in real time, waits up to 10 minutes at your door, and operates through rain, shine, and snow. Oh, and no awkward small talk required.

Here’s what happened in AI today:

😺 ChatGPT rolled out ads to all free and Go users in the US.
📰 LLMs consistently recommend trendy buzzword strategies, regardless of industry or prompt.
📰 SoftBank broke ground on a $500B AI data center in Ohio.
🍪 Mark Zuckerberg is building a personal AI agent to help him run Meta.

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

P.S: Want to reach 675,000 AI-hungry readers? Click here to advertise with us.

😺 ChatGPT's Free Ride Is Over. Here's the Bill.

Every time you ask ChatGPT a question, it costs about a third of a cent to answer. Multiply that by a trillion queries a year, and you're looking at $3 billion in electricity costs alone before salaries, data centers, or research.

That math is why ads are now showing up in ChatGPT.

After testing ads with select US users, ads for free and Go-tier users in the US will roll out in the coming weeks. They appear at the bottom of answers when there's a "relevant sponsored product or service." Plus, Pro, Business, and Enterprise users won't see them for now. To power the rollout, OpenAI partnered with Criteo, the ad-tech firm responsible for those shoe ads that follow you around the internet for two weeks after one Google search.

Here's why the economics made this inevitable:

ChatGPT has 900 million weekly active users but only 50 million are paying.
Running the platform costs roughly $17 billion a year.
Revenue hit $20B+ in 2025 growing 10x since 2023 and still doesn't cover costs if infrastructure costs are included.

OpenAI CFO Sarah Friar addressed this directly in the past. The business model has to "scale with the value intelligence delivers." Subscriptions cover power users. The API covers developers. Ads, she wrote, follow the same logic. As people make decisions, ads should present relevant options to them as long as these options offer real value.

COO Brad Lightcap called the rollout "an iterative process" when they initially tested ads. OpenAI says ads won't influence ChatGPT's answers, won't share your conversations with advertisers, and will stay off sensitive topics like health and politics. Users under 18 are exempt.

Early advertiser commitments give a sense of the scale OpenAI is targeting. Various sources report that OpenAI is charging a reported $60 per 1,000 impressions, an unusually high rate, with a $200K minimum commitment. Shopify is already letting merchants advertise through its Shop Campaigns network, alongside early testers like Target, Williams Sonoma, and Adobe.

The free tier built ChatGPT's 900 million users. Ads are how OpenAI pays for them.

Indeed, there’s no such thing as a free lunch. Every query costs something. Someone has to pay for it. In the case of ChatGPT Free, users don’t spend a dime for every query but OpenAI shoulders the cost. Now the bill's arrived, and it comes in the form of ads.

FROM OUR PARTNERS

Ready to revolutionize your enterprise AI

Ready to revolutionize your enterprise AI? With Airia, you gain seamless AI management, from deployment to governance. Overcome adoption barriers and maximize efficiency—without compromising security or scalability. Airia empowers innovation, makes AI actionable, and helps your business achieve more with less.

Learn more

🎓 AI Skill of the Day: How to Use Multiple Claude Agents for Better Outputs

Most people use Claude like a chatbot. KJ Rainey uses it like a full-time employee who never sleeps. In his free Claude Cowork masterclass on YouTube, KJ breaks down how to unlock Claude's actual power using a dead-simple framework:

Give Claude a filing cabinet, not a flood. The reason most people get mediocre AI outputs? They dump everything into one chat. KJ's fix: organize your context into folders and markdown files, so Claude only pulls what it needs for each task. Better inputs, better outputs. Always.

Our favorite insight: Split your tasks into separate agents. Don't ask one Claude chat to generate ideas, write the draft, AND review it. Use three separate chats, each with a clean context window and only the ingredients it needs. One agent to ideate. One to write. One to review. The reviewer hasn't seen the original prompt, so it's not biased toward the output it just wrote. That's how you go from brown-soup outputs to something actually good.

Want more tips like this? Check out our AI Skill of the Day Digest for this month.

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

Click the image to watch this episode.

What if your AI chatbot deleted everything you told it the moment you closed the chat?

That's exactly what Eamonn Maguire built. The Oxford PhD and former CERN researcher is now Director of Engineering for AI at Proton, and he just sat down with us to explain how Lumo, Proton's privacy-first AI, encrypts your chats and wipes your data after every session.

New episodes air every week on: Spotify | Apple Podcasts | YouTube

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 675K+ readers here!

Clarity handles your customer support, answering questions, spotting recurring issues, and keeping responses compliant so your team can focus on the hard stuff.
Context.dev gives you one API to pull clean data from any website: text, images, products, and brand assets in seconds.
LTX 2.3 turns text or images into polished video clips using an open-source model you can run, fine-tune, and deploy yourself. Free.
TempoLens is a free Chrome extension that helps you read up to 500 WPM by flashing words one at a time or bolding the first half of each word. Built with ADHD and dyslexia in mind. Free forever.
ThetaEdge analyzes your stock portfolio and surfaces covered call opportunities with strike, expiration, premium, and assignment risk all in one view so you stop leaving monthly income on the table.
Tempmail Mail generates a disposable email address in seconds so you can sign up for anything without exposing your real inbox. First 9 addresses are free.

📰 Around the Horn

LLMs consistently recommend trendy buzzword strategies over context-specific ones, regardless of industry or prompt, new HBR research found.
SoftBank broke ground on a $500B AI data center campus in Ohio, targeting 10 gigawatts of power at a former uranium enrichment site.
Mark Zuckerberg is building a personal AI agent to help him do his job as CEO of Meta.
Tesla teased an imminent "Terafab" announcement, a US-based semiconductor factory Musk says is critical to scaling robotaxis and Optimus robots.
Tencent launched ClawBot, a tool that brings the OpenClaw AI agent directly into WeChat for its 1 billion+ monthly users.

Want absolutely EVERYTHING that happened in AI this week? Click here!