😼 OpenAI's President Just Told You Exactly What the Company is Betting Everything On

 Sign Up · Advertise

Welcome, humans.

LATER TODAY at 1:00PM ET / 10:00AM PT: Grant and Corey are going live for AI for Total Beginners… a walkthrough of our 5-Level AI Proficiency Stack that takes you from "I use ChatGPT sometimes" to "AI saves me 10 hours a week."

If you've never opened ChatGPT before, or you have but you're not sure you're getting the most out of it, this one's for you. Set aside an hour to at least get the basics down!

We'll cover Projects, Prompting, Skills, Automations, and Agents; no coding required. If you feel like a total noob in AI, this one’s for you. Come hang, ask questions, and level up your skills for 2026. Watch on YouTube | Join on LinkedIn

NOW: Ethan Mollick pointed out something brutal yesterday: despite AI giving everyone access to near-free image and video generation, this year's April Fools' posts were just as mediocre as ever. The tools got infinitely better. The ideas did not.

Turns out the bottleneck was never "I can't make a video." It was "I have nothing interesting to say." AI has democratized production but it cannot democratize taste.

Here’s what happened in AI today:

• 😼 OpenAI's co-founder Greg Brockman revealed the company is killing video generation to build one AI super app, said AGI is "70-80% here," and teased a model with two years of research baked in

• 📰 OpenAI closed a record $122B funding round at $852B valuation, but investors are already trying to dump shares and pivot to Anthropic

• 📰 Oracle fired an estimated 25,000 employees via 6am termination emails to fund its AI data center buildout

• 🍪 Cloudflare launched EmDash, a free open-source CMS positioning itself as the spiritual successor to WordPress

• 📰 A peer-reviewed Science study confirmed sycophantic AI is widespread and actively harmful, decreasing prosocial behavior while promoting dependence

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

Oh, and if you like robots, you’re in luck! The Neuron is getting ready to launch a sister publication all about robots! To get early access when it’s available, sign up here!  

😼 OpenAI's President Just Told You Exactly What the Company is Betting Everything On

Your favorite AI company just raised the largest venture round in history ($122B). So what's it doing with the money? According to co-founder Greg Brockman, who sat down with Alex Kantrowitz on Big Technology Podcast, the answer is: fewer products, more compute, one app.

Here's what happened:

OpenAI is killing Sora and video generation as standalone products, folding the research into robotics instead. Brockman: "Pursuing both branches is very hard for us."

The company is building a "super app" that merges ChatGPT, Codex, and browser into one product. Think: your personal AGI that knows you, your work, and your calendar.

A new pre-training run codenamed "Spud" represents two years of research. Brockman says it will solve harder problems, understand nuance better, and eliminate that frustrating moment where the AI "should be able to figure this out."

On AGI: Brockman says he's "70-80% there" on his personal definition. He expects full AGI "within the next couple years."

Codex is no longer just for coders. One employee built an Adobe Premiere plugin through it. The communications team hooks it to Slack and email to synthesize feedback.

When his team asked how much compute to buy after ChatGPT launched, Brockman's answer: "All of it."

Why this matters: The real thread connecting every one of these decisions is compute scarcity. OpenAI killed Sora because video generation runs on a different technical branch than the GPT reasoning models, and they can't afford to power both.

The super app exists because running multiple separate AI products burns compute that should be funneled into one system. Even the $122B raise is going straight into chips; Brockman called compute "a revenue center, not a cost center." Every GPU they build gets maxed out by demand.

This is the company that's supposed to have infinite resources telling you they have to make painful choices about what to ship. If OpenAI is compute-constrained at $852B in valuation, imagine what that means for everyone else.

Our take: Brockman took a direct shot at Anthropic's Dario Amodei, who recently called some AI companies' infrastructure bets reckless. "I just disagree," Brockman said. His bet is that everyone will be compute-starved by year's end, and OpenAI simply saw it first.

The most revealing moment came when he discussed an automated AI researcher coming this fall; essentially an AI that does the full job of an OpenAI research scientist, in silicon. Ask it to find AGI. And it tries.

FROM OUR PARTNERS

Why agentic AI is your next priority

• Agentic AI isn't just genAI with extra steps

• The buzz around agentic AI has reached a fever pitch. But most implementations are "agent-washing" — a chain of prompts or normal workflows with an LLM in the loop.

• This ebook breaks down agentic AI, how it differs from gen AI, and how it can improve CX on your site. Find out from the experts why you should prioritize agentic AI technology.

Read the ebook

🎓 AI Skill of the Day: How to Use AI Agents for Knowledge Work (Even If You're Not a Coder)

One of the most interesting moments in Greg Brockman's interview: he described OpenAI employees who aren't engineers using Codex to automate their work. The communications team hooks it to Slack and email, synthesizes feedback, builds internal tools. You don't need to know code to do this.

Here's how to start: instead of asking your AI to answer a question, ask it to build a tool that solves a recurring problem. The shift is from "help me with this task" to "build me something that handles this task forever."

Try this prompt in ChatGPT or Claude:

I have a recurring task at work: [describe the task].
I currently do it [how often] and it takes about [time].
Build me a simple tool or workflow that automates this.
Walk me through setup step by step, assuming I have zero coding experience.
If it requires any tools or accounts, tell me which ones and how to set them up.

The trick is being specific about the pain point. "Help me with email" gives you generic advice. "Every Monday I manually pull metrics from three dashboards, copy them into a spreadsheet, and format a summary for my manager" gives the AI enough context to build something real.

Favorite insight: Brockman's framing of the shift was perfect; computers were always supposed to contort to the human, not the other way around. The blank prompt box is intimidating, but the moment you start describing your problems instead of trying to phrase perfect questions, AI becomes dramatically more useful.

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

🍪 Treats to Try

1. EmDash (by Cloudflare) is a full-stack serverless CMS that sandboxes every plugin, includes built-in payments and AI agent skills, and comes with a WordPress importer —free/open-source.

2. Mercury Edit 2 predicts your next code edit using recent changes and codebase context, boosting acceptance rate 48% with sub-second latency —$0.25/M input, $0.75/M output (10M free tokens for new accounts).

3. Willow Voice launched Atlas 1, a speech-to-text model that holds 1.2% word error rate in clean audio and 2.1% in production (most competitors fall to 10-15% in real-world conditions), turning speech into formatted text 5x faster than typing —free to try.

4. Baton orchestrates multiple AI coding agents (Claude Code, Gemini, Codex CLI) in parallel with git-isolated worktrees so they never conflict —no pricing details.

5. Veo 3.1 Lite (by Google) is a cost-effective AI video generation model now available through the Gemini API and Google AI Studio —paid preview.

📰 Around the Horn

1. Q1 2026 venture funding hit $297B across 6,000 startups, shattering all records and jumping roughly 150% from the previous quarter.

2. Salesforce announced 30 new AI features for Slack including reusable "AI skills," meeting transcription, and desktop activity monitoring with suggested follow-ups.

3. Microsoft committed $6.5B across Singapore ($5.5B) and Thailand ($1B) for AI infrastructure through 2029.

4. NVIDIA invested $2 billion in Marvell Technology for NVLink Fusion (a way to connect different chip types into one AI supercomputing fabric); Marvell stock surged 13%.

5. NYC Health + Hospitals’ CEO said he's ready to replace radiologists with AI for first-read mammograms and X-rays, citing AI that's "actually better than human beings"; radiologists called it dangerously naive.

6. A Science study proved sycophantic AI (chatbots that excessively agree with you) is widespread across all 11 major models and actively decreases prosocial intentions. Your chatbot's flattery is making you a worse person, and there's now peer-reviewed evidence for it.

7. We watched a great overview of the new AI Studio with Logan Kilpatrick and Ammaar Reshi on Learn Prompting, a great resource to learn & apply AI skills; check it out!

FROM OUR PARTNERS

Are your LLM applications actually production-ready?

Move beyond prototypes with Datadog’s LLM Observability best practices guide. Learn to secure and optimize your AI at enterprise scale from tracing multi-call workflows and detecting prompt injections to evaluating model quality and controlling token costs before they spike.

View the Guide

🧩 Thursday Trivia

One is AI, and one is real. Which is Which? Vote Below!

A.

B.

A Cat’s Commentary

Trivia answer: A is AI, and B is real… but clearly that poster reads The Neuron lol. Then again… to quote Grant’s favorite line from Elf: “Ya never can tell, kid.”