😺 Are you ready for AI school?

PLUS: Someone tried to replace Sam Altman with ChatGPT

Welcome, humans.

Someone allegedly just applied to be “CEO of OpenAI” and got spectacularly rejected. The (presumably) fake rejection letter is going viral on LinkedIn and Reddit and honestly, it's comedy gold.

Ömer Öztok, CEO of AI security startup Sondra, said he shot his shot with a cover letter that promised to replace OpenAI's entire C-Suite with ChatGPT agents as his
“first action.” OpenAI's supposed response? That seemed “apocalyptic.” His backup plans included acquiring Google “just for the name” and training GPT-6 exclusively on his own tweets.

Look, we've all sent cringey job applications, but this guy went full send and turned his rejection into a marketing masterpiece. Plot twist: he's probably laughing all the way to the bank as thousands of people now know about Sondra. Sometimes the best way to get attention for your AI startup isn't a boring press release—it's applying to replace Sam Altman with a chatbot.

Here’s what happened in AI today:

We analyze what happens if AI makes everyone smarter.
Anthropic reported cybercriminals weaponizing Claude for extortion and fraud.
U.S. Special Operations Command pursued AI propaganda tools.
xAI sued former engineer for alleged trade secret theft.

Advertise in The Neuron here

P.S: Check out our latest podcast episode if you haven’t yet! This one’s one of our favorites: a chat with AI Studio Head Logan Kilpatrick on all the cool things you can do with AI Studio (including how HE uses it to build AI Studio!).

Listen and/or watch on: YouTube | Spotify | Apple Podcasts

What happens if AI actually make us all smarter?

We’re big fans of TBPN here at The Neuron, and last Friday, they had fintech legend, PayPal mafia alum, and now Affirm CEO Max Levchin on the show, who shared a fascinating insight: AI is actually raising the universe's average IQ (or at least the IQ of everyone online).

This tracks with what Ethan Mollick just wrote about “mass intelligence”, or the fact that close to a billion people now have access to near genius-level AI models.

Say what you will about economic bubbles or whether AI makes us lazy, but one thing's undeniable: these tools are dramatically raising the floor of what average users can accomplish.

Now here's where it gets wild: this might solve one of humanity's oldest problems…

Ever notice how every major platform eventually turns to garbage? Google drowning in SEO spam, Airbnb becoming a minefield of fees, social media optimizing for outrage?
12 Grams of Carbon recently wrote about the newly coined (and highly scientific) term for this phenomena, “enshittification”, and how it works.
See, services getting worse over time is not just the platforms' fault, but ours as users too. We're all stuck in a prisoner's dilemma where competitive incentives force everyone—platforms and users alike—to make choices that destroy value over time.
Back in 2014, Scott Alexander called this force “Moloch”—the demon of competition that makes us sacrifice everything good on the altar of staying competitive, and how its actually plagued us for as long as we humans have been around (long before the internet or ChatGPT slop).
His proposed solution at the time? Only a benevolent artificial superintelligence could coordinate humanity out of this mess (Ironically, he later co-authored AI 2027, predicting the race to build that very superintelligence might destroy us all. Whoops!).

Well, the conventional wisdom says AI is about to make Moloch, or “enshittification”, exponentially worse—more spam, more manipulation, more races to the bottom.

But what if we're thinking about this backwards? Let’s revisit Max’s point: what if AI IS actually raising humanity's average IQ? Not through some sci-fi brain implant, but by making everyone who touches these tools dramatically smarter at navigating the world.

Max's point was simple but profound. When AI helps us evaluate products in real-time, bad actors can't hide behind complexity anymore. His example: AI-powered search helps people understand that Affirm's transparent loans actually cost less than credit cards with compound interest traps.

Now multiply this by every decision we make daily:

Google's SEO slop problem? Now your AI assistant pre-filters search results, giving you just the information you ask for. Suddenly, spending millions on SEO manipulation becomes pointless—only genuine value surfaces.
Airbnb's bait-and-switch economy? Your AI agent spots patterns across thousands of reviews, calculates real total costs including those “surprise” cleaning fees, and warns you about hosts who mysteriously always find “damage.” The scammers lose their camouflage.
AI-generated slop flooding your feeds? As AI assistants get better at detecting synthetic content, they can flag it before it wastes your time, and the attention economy's race to the bottom hits a wall when fake engagement can't fool anyone anymore.

You can basically call this the “unshittification” hypothesis!

The catch: This only works if AI can reliably distinguish truth from fiction, and if humans keep creating genuinely valuable products, content, and services. Otherwise, we're just automating (and accelerating) Moloch’s race to the bottom.

But here's the thing—we don't need perfect detection. We just need to raise the cost of deception high enough that creating value becomes more profitable than gaming the system.

This is exactly what Meta's betting $72 billion on: so called ”personal superintelligence” that helps you navigate life better. Though ironically, if we all get smart enough to stop doomscrolling, Meta's current business model evaporates.

Now apply this to education. Joe Liemandt and MacKenzie Price at Alpha School aren't waiting for superintelligence. Their AI tutor helps kids master entire grade levels in only 20-30 hours right now. The results are almost unbelievable:

High schoolers who loved learning so much they canceled summer break.
A student jumping 50 SAT points after AI identified she needed third-grade multiplication practice.
Liemandt predicts that within 5-10 years, a sub-$1K tablet will teach any child everything they need academically in two hours daily.
The rest of their day? Building real businesses, developing actual skills, becoming genuinely capable humans (and Alpha School has a plan to take this national).

Check out our deep dive on two great interviews w/ the leaders who run Alpha School.

Now scale this up. When a billion kids get 10x better at learning, they become adults who are 10x better at evaluating products, spotting scams, and creating valuable products and services. The enshittification cycle breaks because the marks get too clever to con.

This might be AI’s ultimate twist: Instead of building artificial general intelligence to govern us from above, we're scaling human general intelligence from below. Every kid who learns to think critically, evaluate evidence, and solve real problems becomes another node in humanity's distributed intelligence network…making us all better for it.

So maybe we don't need a benevolent AI overlord to solve Moloch. We just need everyone informed enough to stop feeding the beast.

FROM OUR PARTNERS

Nothing slows you down like the endless back-and-forth of scheduling.

The constant context switching between apps, tabs, and tasks, the endless searching for a single slot that works for everyone — minutes tick by. Momentum dies.

Superhuman brings calendar and email together in one flow. AI triages your inbox, drafts replies in your tone, reminds you before you forget — and now, with the latest calendar feature, Find Time, it even books meetings without breaking stride. See your calendar and suggested slots side-by-side, pick one, save, and you’re done.

One workspace, zero wasted motion.

Get 1 month free →

Prompt Tip of the Day.

Transform ChatGPT-5 into your productivity sidekick with these nine simple prompts from Amanda Caswell at Tom's Guide.

The standouts: “Summarize this for me like I'm on a deadline” for quick document reviews, “convert the following notes into a prioritized to-do list” to organize brain dumps, and “find the hidden action items in this meeting transcript” when you spaced out during calls. Sometimes, simple commands are just better.

Check out all of our prompt tips of the day from August here!

Treats to Try.

*Asterisk = from our partners. Advertise in The Neuron here.

*VEED creates professional videos from text prompts, automatically adds subtitles, and generates avatars for your marketing content and YouTube videos.
Xpander handles all the backend infrastructure for your autonomous agents so you can build, test, and deploy them in under an hour instead of managing servers, databases, and deployment pipelines yourself.
Cloudflare Radar tracks which AI bots are crawling your website and what they're doing with your content.
Grok Code Fast 1 helps you code faster by understanding your entire project and calling tools automatically to fix bugs, write features, and handle complex tasks—free via partners like Cursor/Copilot for a limited time, then $0.20/$1.50 per million tokens.
Sara AI assistant remembers everything you tell it, recognizes your face and friends, and starts conversations with you while controlling your smart home devices.
For the devs:
1. Greptile AI review automatically reviews your pull requests by understanding your entire codebase, catching bugs that span multiple files and suggesting fixes with full context.
2. Sourcerer MCP searches your codebase semantically so you can find "user authentication logic" instead of reading entire files, saving massive token costs.
3. Skipper CLI agents gives command-line AI tools like Claude Code access to control your browser and desktop, automating workflows across applications.
4. sosumi.ai for Apple docs converts Apple Developer documentation into AI-readable Markdown and provides an MCP server so you can ask questions about Swift, SwiftUI, and other Apple frameworks.

Around the Horn.

Apple created a new AI chatbot called “Asa” for retail employees to help them learn about Apple products and improve their sales abilities through the company's internal training app.
Anthropic said cybercriminals have weaponized Claude for large-scale data extortion operations, North Korean operatives used it to fraudulently obtain Fortune 500 tech jobs, and criminals with minimal coding skills developed ransomware-as-a-service through AI assistance.
U.S. Special Operations Command apparently pursued AI tools to create propaganda that can "influence foreign target audiences" and "suppress dissenting arguments" in real-time influence operations.
Meta's AI leaders discussed using Google's Gemini and OpenAI's models to power Meta AI and other features in their social media apps as temporary measures while developing their own models.
Meta updated its chatbot training to prevent teenage users from discussing self-harm, suicide, disordered eating, or inappropriate romantic conversations and limited teen access to certain AI characters.
xAI sued former engineer Xuechen Li for allegedly stealing trade secrets about Grok after receiving $7 million in stock liquidation before joining OpenAI.
Meta's $14B+ Scale AI partnership showed early strain with executive departures and Meta's TBD Labs working with competing data vendors Mercor and Surge due to quality concerns.

FROM OUR PARTNERS

Anything

Introducing Anything

Anything lets anyone launch beautifully designed mobile and web apps from just a prompt. No coding or extra tools required. Anything’s frontier AI coding agent builds full apps end to end with database, payments, hosting, and one-click App Store submission all included. Go from idea to live in hours, not months.

Start building with Anything.