😺 Should YOU care about o3-Pro?

PLUS: Sam Altman says we hit the singularity... Web Audience

Welcome, humans.

ChatGPT went down yesterday morning, and the internet collectively lost its mind. Everyone literally ran to Reddit to check if it was just them, which perfectly captures our 2025 reality.

The memes were instant and brutal. People joked about being “millions forced to use brain” for the first time in months. Others posted what ChatGPT HQ looked like with technicians frantically fixing servers. The best meme showed everyone's panic face when asked to think without AI assistance.

But the funniest response? One guy claiming “it's working fine for me” while showing a screenshot of ChatGPT generating its own error message. Peak trolling.

The outage lasted several hours due to… plot twist… the launch of o3 Pro, which we'll dive into in today's main story. So everyone can relax; now you have an excuse to use even less of your brain with an even stronger thinking model.

TBH, this collective meltdown shows just how addicted we've become. People are spending ~$15B+ a year across the top AI tools, with ~$800M monthly on ChatGPT alone. Apparently we’re already at the point where we can't function for even a few hours without them…

Here’s what you need to know about AI today:

OpenAI released o3-pro and dropped the price of o3 by 80%.
Meta created a new lab focused on “superintelligence.”
Publishers are getting crushed by Google’s new AI Mode.
Sam Altman predicts the “AI Singularity” might already be here.

Advertise in The Neuron here.

OpenAI just released o3-pro, its “most reliable” thinker yet, and o3 got a major price cut…

Picture this: You're a developer who's been drooling over OpenAI's most powerful reasoning models but wincing every time you see the bill.

Well, Santa Sam Altman just dropped by with the gift that keeps on giving—and it's 80% cheaper than you'd expect. Christmas in July came a month early this year!

OpenAI launched o3-Pro today, and the price cuts along with it. The base o3 model got an 80% price slash from $10/$40 per million tokens (aka ~750K words) down to just $2/$8, while the new flagship o3 Pro clocks in at $20/$80—which is ~87% cheaper than the old o1-pro it's replacing.

Here’s OpenAI's official take on o3 Pro:

Expert evaluators consistently prefer o3-pro over standard o3 across science, education, programming, data analysis, and writing.
Rated higher for clarity, comprehensiveness, and instruction-following
Excels at math, science, and coding like o1-pro did.
Uses rigorous “4/4 reliability” testing (must answer correctly in all four attempts, not just one).
Has full tool access: web search, file analysis, visual reasoning, Python, memory, and more.
Available for Pro and Team users starting today, replacing o1-pro—making this OpenAI's most capable reasoning model widely available for the first time.

Swyx from Smol.AI shared a great visual (via Ben Hylak of Raindrop) for how to think of prompting o3 differently form other models.

In fact, if you want to understand o3 Pro and how it works, you should probably read Ben’s review. The TL;DR = o3 Pro's intelligence only emerges when you feed it massive amounts of context, but when you do, it produces insights that can fundamentally change how you think about complex problems.

THE CATCH: It's slower. Like, grab-a-coffee-while-you-wait slower. And you definitely don’t want to waste an o3-pro prompt on something small—just saying hi can cost you $80 and take 20 minutes to return an answer.

This weird, jagged intelligence issue (where AI can excel in one area and fail in another) was the crux of Apple’s recent illusion of thinking paper and its claims that “reasoning” models don’t actually reason and often overthinks problems.

The paper was more or less torched by the AI community (here’s the main counter argument), but is not without its merits. That said, o3-Pro seemingly crushed the problem that the researchers used to test other reasoning AI’s limits.

WHY IT'S IMPORTANT: When you can now run comprehensive benchmarks on o3 for $390 versus $971 for Gemini 2.5 Pro, the math gets pretty compelling pretty fast. Sure, Gemini might be the “smarter” model than old o3… but the highest intelligence, lowest cost model will likely win out. Y’know, for practical reasons.

Sam Altman confirmed the change in a post on X highlighting that the new pricing is intended to encourage broader experimentation. OpenAI is essentially betting that making their smartest models more accessible will unlock a wave of innovation they can't even predict yet. And on that note… check out Midweek Wisdom below.

FROM OUR PARTNERS

AI agents are already calling your APIs —the question is whether your stack can handle structured requests, provide proper context, and debug failures when things break.

Postman’s latest product reveals were designed by people who work with APIs all day, every day — making Postman an ideal partner in everything from designing agent-ready APIs to orchestrating intelligent workflows without managing infrastructure.

Think real-time API usage tracking, AI-powered automation, and native tools that plug directly into your stack—from VS Code to Slack to GitHub.

Get the full technical breakdown and start leveling up with Postman.

Prompt Tip of the Day

If a problem has you stumped, have the AI probe for overlooked factors. Ask “What might I be missing?”

Example: "Our project keeps getting delayed despite our best efforts. What might we be missing?"

Or better yet, have it brainstorm “at least 15 things I might be missing, ranked from most obvious to most unexpected, in a bullet point list. If there are less than 15 potential missing factors, include as many elements approaching 15 as there are.”

Often, an outside perspective (even an artificial one) can spot a missing angle you haven't considered.

This might be the best use of “artificial” intelligence—sparking real intelligence.

Treats To Try.

*Anything marked with asterisks is sponsored content. Advertise in The Neuron here.

*Shortform distills the best nonfiction in tech, business, and beyond—gain in-depth guides loaded with smart insights and analysis to help you learn faster and think sharper. Get 20% off.
Magistral is Mistral’s new open reasoning model (download) that shows its step-by-step thinking so you can verify how it solved your complex problem—try it here with the “think” button.
HuggingFace released AI Sheets, which transforms your spreadsheets so you can automatically analyze data, generate content, and build automations with thousands of built-in AI models—free to try.
Freepik now creates realistic videos with synced audio and camera movements from your text prompts using Google’s Veo 3 (demos).
Chronicle turns your bullet points into stunning presentation slides with drag-and-drop widgets.
Maze uses AI agents to investigate your cloud vulnerabilities in context of your actual environment and removes 90% of false positives so you only fix what attackers can actually exploit.
Personal AI creates AI teammates trained on your company data that work alongside your human team in messaging channels.
Linear helps you plan and track your software product development with issues, roadmaps, and sprints in one tool (raised $82M).
Simon Willison has an LLM Prices calculator that tracks how much a given model will cost you to run.

See our top 51 AI Tools for Business here!

Around the Horn.

It’s nice to see Claude Opus 4 on the board, but unfortunately for Anthropic, it clocks in below both of Gemini 2.5 Pro’s new releases, old o3, o4-mini high, Grok 3, and the new DeepSeek R1.

Sam Altman says OpenAI will release its open weight model later this summer (rather than the expected June delivery), but don’t worry, it’ll be a banger!
Meta just created a new lab focused on “superintelligence” (beyond AGI) and is throwing 7-to-9 figure compensation packages at top researchers to poach them from OpenAI and Google.
Salesforce will restrict third party apps from storing Slack messages long term—this will hurt companies like Glean, which searches across all your company apps to answer questions and automate work, and who just raised $150M.
The WSJ reported that publishers are getting crushed by Google’s new AI Mode—if you’re curious how AI Mode works, check out this report on their patents.

Midweek Wisdom

Sam Altman declared “we are past the event horizon” of AI singularity (where AI surpasses human intelligence), predicting it will feel gradual and manageable as we live through it, with 2026 bringing AI systems that discover novel insights and 2027 bringing real-world robots.
A hiring analyst shared hard data proving AI is already stealing jobs, as companies have cut hiring for “AI-capable roles” (where AI can perform the core tasks listed in this job description) by 19% since ChatGPT launched—tech and admin positions were hit hardest at 31% decline.
The Pulmonologist's demonstration of AI instantly diagnosing pneumonia from X-rays went viral ignited fierce debate among medical professionals about job displacement versus AI assistance, highlighting the tension between AI's superior pattern recognition and the irreplaceable human factor of patient care.
Thirty of the world's leading mathematicians secretly gathered to try and stump OpenAI's latest reasoning AI, only to watch it solve graduate-level problems in minutes that would take them months, with one researcher calling its scientific reasoning process “frightening.”
Astronomers trained a neural network on millions of black hole simulations to crack the secrets of our galaxy's supermassive black hole, discovering it's spinning at near top speed and defying established theories about how magnetic fields work around these cosmic monsters.